Lapilli from the August 1966 eruption of the carbonatite volcano Oldoinyo Lengai consist of carbonate-cemented aggregates of (i) mono- and poly-mineralic fragments of ijolitic rocks, (ii) single grains and clusters of euhedral nepheline, Ti-andradite, and Ti-magnetite, and (iii) corroded pyroxene and wollastonite grains surrounded by coronas containing combeite, melilite, Ca-silicates (possibly larnite and rankinite), and rounded bodies of submicrometre intergrowths with complex bulk compositions dominated by Na,K,Ca-phosphate-carbonate and alkali-iron-sulphide-carbonate. The (ii,iii) materials, together with abundant Na-carbonate, sylvite and fluorite occurring as cement and shells in the lapilli, are attributed to mixing and incomplete reaction of ijolite and carbonatite magmas during the explosive eruption. The rounded submicrometre intergrowths are interpreted as the quench products of two types of immiscible liquids whose properties should be studied by controlled synthesis.

From G. F. Becker's and L. V. Pirsson's early enunciations linking the dynamics of magma chambers to the rock records of sills and plutons to this day, two features stand at the centre of nearly every magmatic process: solidification fronts and phenocrysts. The structure and behaviour of the envisioned solidification front, however, has been mostly that akin to non-silicate, non-multiply-saturated systems, which has led to confusion in appreciating its role in magmatic evolution. The common habit of intruding magmas to carry significant amounts of phenocrysts, which can lead to efficient fractionation, layering, and interstitial melt flow within extensive mush piles, when coupled with solidification fronts, allows a broad understanding of the processes leading to the rock records of sills and lava lakes. These same processes are fundamental to understanding all magmas.

The spatial manifestation of the liquidus and solidus is the Solidification Front (SF); all magmas, stationary or in transit, are encased by SFs. In the ideal case of an initially crystal-free, cooling magma, crystallinity increases from nucleation on the leading liquidus edge to a holocrystalline rock at the trailing solidus. The package of SF isotherms advances inward, thickening with time and, depending on location — roof, floor, or walls — and the initial crystallinity of the magma, is instrumental in controlling magmatic evolution. Bimodal volcanism as well as much of the structure of the oceanic crust may arise from the behaviour of SFs.

In mafic magmas, somewhere near a crystallinity (N) of 55% (vol), depending on the phase assemblage, the SF changes from a viscous fluid (suspension (0<N<25) and mush (25<N<55%)) to an elastic crystalline network (rigid crust (55<N<100%)) of some strength containing interstitial residual melt. With thickening of the roofward SF of some mafic magmas, the weight of the leading, viscous portion repeatedly tears the crust near N ∼ 55–60%, efficiently segregating the local residual melt into zones of interdigitating silicic lenses. This is SF instability (SFI), a process of possible importance in continental crust initiation and evolution, in producing silicic segregations in oceanic crust, and in recording the inability of the viscous part of the upper SF ever to detach wholly in typical (<∼ 1 km) sheet-like magmas. These granophyric and pegmatitic segregations, individually reaching 1–2 m in thickness and 30–50 m in length, form thick (∼ 50–75 m) zones that can be misconstrued as sandwich horizons where the last liquids might have accumulated. In effectively splitting the magma chemically and spatially, SFI is, in essence, a form of chaos (i.e. silicic chaos).

Differentiation of initially crystal-free, stationary magmas is limited to processes occurring within SFs, which operate in competition with the rate of inward advancement of solidification. Local processes operating on characteristic time scales longer than the time for the SF to advance a distance equal to its own thickness are suppressed. Enormous increases in viscosity outward within the viscous, leading portion of the SF efficiently partition the distribution of melt accessible to eruption. Eruptible melts lie essentially inward of the SF and are thus severely restricted in silica enrichment. The silica-enriched SFI melts are thus generally inaccessible to collection and eviction unless the host SF is reprocessed or “burned back” through, respectively, later regional magmatism or massive, late-stage re-injection. And because of large viscosity contrasts between SFI melts and host basalts, once freed, SFI melts are literally impossible to homogenize back into the system and may collect and compact against the roof to form large silicic masses. Unusually voluminous, bulbous masses of silicic granophyre present along, and sometimes warping, the roofs of large diabase sills may reflect collections of remobilized blobs of SFI melts. These bulbous masses may be later added to the continental crust through solid state creep.

In sheets made of phenocryst-rich, singly saturated magma, most phenocrysts are able through settling or floating to avoid capture by the advancing SFs. Significant differentiation is possible through extensive settling of initial phenocrysts and upward leakage of interstitial residual melt from the associated cumulate pile, which over-thickens the lower SF, greatly tipping the competitive edge against suppression of melt leakage by advancing solidification. Dense interstitial melts may similarly drain from roofward cumulates of light phenocrysts. The variation in crystal size and modal abundance in these cumulate piles are intimate records of prior crystallization, transport, and filling.

Magmas in transit erode SFs and thoroughly charge the magma with crystals, facilitating fractionation and differentiation, especially if the body occasionally comes to rest. The key to protracted differentiation through fractional crystallization is not crystallization in stationary, closed chambers, but the repeated transport and chambering of magma or the periodic resupply to chambers of phenocryst-rich magma. This is punctuated differentiation, which may be the general case. Close corollaries are that thick, closed sheets of initially crystal-free, multiply-saturated magma undergo precious little overall differentiation, and that deciphering the sequence and crystallinity, including in transit phenocryst entrainment, growth, and sorting, of the filling events is central to unravelling intrusive history.

Variations in temperature, whether on phase diagrams or in actual magmas, are intrinsically linked to commensurate variations in space and time in magmatic systems. The spectrum of all physical and chemical processes associated with magma is accordingly strongly partitioned in space and time.

The idea of a magma chamber as a vat of low crystallinity melt crystallizing everywhere within and differentiating through crystal settling is unrealistic. A magma chamber formed of any number of crystal-laden inputs, encased by inward-propagating, dynamic solidification fronts, and where significant differentiation is tied to the dynamics of late-stage, interstitial melt within extensive mush piles is more in accord with the rock record.

Three small picritic dykes, intruded at a late stage in the evolution of the Rum basic–ultra-basic complex, Inner Hebrides, shed new light on the nature of the magmas responsible for the main complex. The magmas are of transitional (mildly alkalic) type, generated by relatively small-fraction (6–7%) melting of a depleted mantle source. Melting is tentatively deduced to have commenced at ±100 km, straddling the garnet–spinel transition. Of the three samples, one (M9) is remarkable for the preservation of very primitive characteristics, with olivines of Fo93 containing highly aluminous spinels; these appear unique within the British Tertiary Volcanic Province. Sr, Nd and Pb isotopes indicate only minor ([les ]4%) contamination with Precambrian crustal lithologies, reflecting the rapidity of ascent of the magma batches. The Rum picrites have 187Os/188Os ratios and trace element characteristics comparable to those of recent picrites from Iceland, suggesting minimal temporal change of the depleted parts of the Iceland plume over 60 Ma. Movements of the Long Loch Fault may have been instrumental in causing decompression melting of the spreading Iceland plume-head and facilitating ascent of the melts to near-surface levels.

Granites within suites share compositional properties that reflect features of their source rocks. Variation within suites results dominantly from crystal fractionation, either of restite crystals entrained from the source, or by the fractional crystallisation of precipitated crystals. At least in the Lachlan Fold Belt, the processes of magma mixing, assimilation or hydrothermal alteration were insignificant in producing the major compositional variations within suites. Fractional crystallisation produced the complete variation in only one significant group of rocks of that area, the relatively high temperature Boggy Plain Supersuite. Modelling of Sr, Ba and Rb variations in the I-type Glenbog and Moruya suites and the S-type Bullenbalong Suite shows that variation within those suites cannot be the result of fractional crystallisation, but can be readily accounted for by restite fractionation. Direct evidence for the dominance of restite fractionation includes the close chemical equivalence of some plutonic and volcanic rocks, the presence of plagioclase cores that were not derived from a mingled mafic component, and the occurrence of older cores in many zircon crystals. In the Lachlan Fold Belt, granite suites typically evolved through a protracted phase of restite fractionation, with a brief episode of fractional crystallisation sometimes evident in the most felsic rocks. Evolution of the S-type Koetong Suite passed at about 69% SiO2 from a stage dominated by restite separation to one of fractional crystallisation. Other suites exist where felsic rocks evolved in the same way, but the more mafic rocks are absent. In terranes in which tonalitic rocks formed at high temperatures are more common, fractional crystallisation would be a more important process than was the case for the Lachlan Fold Belt.

A two-parameter Margules method has been developed for modelling the thermodynamic mixing properties and hypersolidus phase relations of albite-water (ab-w) melts. The method is based largely on phase-equilibrium and calorimetric data that are either currently available or readily obtainable; P-V-T data are not required.

The new modelling method has been applied to calculate: (1) thermodynamic mixing properties for ab-w melts at 2-5 kbar, 815°C; and (2) hypersolidus phase relations for the ab-w system at 2·5 kbar. Results are uniformly positive—activity-composition relations are similar to those predicted by equations based on P-V-T data, excess enthalpies are in good agreement with data obtained from calorimetry, and calculated phase relations are fully compatible with phase-equilibrium data for ab-w melts acquired at 2·5 kbar.

It is concluded that the Margules method for thermodynamic modelling of ab-w melts is a practical alternative to the P-V-T modelling method.