Late-stage microstructures developed during the last stages of solidification of the Skaergaard intrusion comprise a wide array of reactive and non-reactive intergrowths. Reactive microstructures ascribed to open-system behaviour include serrated grain boundaries between pyroxene and plagioclase primocrysts, fish-hook pyroxenes and mafic symplectites. They form by the addition of Fe and Ca and removal of alkalis and silica. Other microstructures include those formed by (internally generated) redox reactions between olivine, Fe–Ti oxides and pyroxene. Non-reactive microstructures include closely spatially associated granophyric intergrowths and previously undescribed ilmenite-rich intergrowths that are interpreted as a consequence of crystallization of separated conjugate immiscible liquids. The open-system reactive microstructures occur predominantly in the cumulates on the chamber floor, appearing in LZb and disappearing (to be replaced by a granophyre–ilmenite-rich intergrowth association) in the Upper Zone. They are not common in the Marginal Border Series. Their distribution mirrors that of efficient expulsion of trapped liquid and can be attributed to the gravitationally driven loss of a Si-rich immiscible component from the interstitial liquid. The loss of the Si-rich component causes the remaining Fe-rich liquid to react with the primocrysts. Bulk-rock major element data are consistent with little or no preferential loss of the Si-rich liquid from the mush, but the resolution of the available data is not sufficient to assess the effect of this relative movement on the liquid line of descent of the bulk magma. The first appearance of the paired conjugate non-reactive intergrowths in MZ* (Marginal Border Series) points to the early onset of immiscibility in the bulk liquid.

Re-examination of the Skaergaard intrusion in the context of its regional setting, combined with new data from exploration drilling, has resulted in a revised structural model for the intrusion. It is modelled as an irregular box, c. 11 km from north to south, up to 8 km from east to west, and 3·4–4 km from the lower to the upper contact. The walls of the intrusion are inferred to follow pre-existing and penecontemporaneous steep faults, and the floor and roof seem largely controlled by bedding planes in the host sediments and lavas, similar to regional sills. The suggested shape and volume are in agreement with published gravimetric modelling. Crystallization along all margins of the intrusion concentrated the evolving melt in the upper, central part of the intrusion, best visualized as an ‘onion-skin’ structure inside the box. The total volume is estimated to c. 280 ± 23 km³, of which 13·7% are referred to the Upper Border Series (UBS), 16·4% to the Marginal Border Series (MBS) and 69·9% to the Layered Series (LS). In the LS, the Lower Zone (LZ) is estimated to constitute 66·8%, the Middle Zone (MZ) 13·5% and the Upper Zone (UZ) 19·7%. The new volume relationships provide a mass balance estimate of the major and trace element bulk composition of the intrusion. The parental magma to the Skaergaard intrusion is similar to high-Ti East Greenland tholeiitic plateau basalts with Mg number c. 0.45. The intrusion represents the solidification of contemporary plateau basalt magma trapped and crystallized under closed-system conditions in a crustal reservoir at the developing East Greenland continental margin.

High-precision electron microprobe data for major and trace elements (nickel, calcium and manganese) in the margins and rims of groundmass olivine grains from the Majuagaa kimberlite (sensu stricto) in southern West Greenland are presented. Despite a range of olivine core compositions defining several zoning types most of the olivine grains have similar margins. The olivine cores are considered to be xenocrysts in the kimberlite magma whereas the margins represent cognate olivine crystallized from the kimberlite melt itself. We evaluate models of olivine margin formation by fractional crystallization, fractional crystallization with simultaneous digestion of xenoliths, and diffusion. Only fractional crystallization coupled with digestion of xenocrysts (primarily orthopyroxene), with subsequent minor diffusion, can account for the observed compositional profiles in the olivine margins. We propose that the digestion crystallization process, whereby the entrained xenolithic material reacts with the kimberlite melt, is important in the evolution of kimberlite magmas. We suggest that this type of reaction gives a distinct local flavour to the parental kimberlite melt and can account for the similarities and differences between the various kimberlitic (sensu lato) magma types present in southern West Greenland and worldwide.

The Skaergaard intrusion, Greenland, is the type locality for Skaergaard-type mineralizations. Mineralization levels are perfectly concordant with igneous layering, up to 5 m thick, internally fractionated, and contain crystallized sulphide droplets and precious metal alloys, sulphides, arsenides and telluride. Immiscible Cu-rich sulphide droplets, formed in a mush zone below the roof, scavenged precious metals. They were subsequently dissolved and transported to the floor in late-formed, immiscible, Fe-rich mush melts. Mineralized stratigraphic intervals of floor gabbro formed in ‘proto-macrolayers‘, owing to local sulphide saturation in melt concentrated between floating plagioclase and sinking clinopyroxene. The floor mineralization is divided into four stratigraphic sections. Formation of the Lower Platinum Group Element Mineralization (LPGEM) involved: (1) crystallization of the bulk liquid liquidus paragenesis and in situ fractionation; (2) sulphide saturation and formation of sulphide droplets in melt in the upper part of ‘proto-macrolayers‘. After further in situ fractionation, the following steps occurred: (3) the onset of silicate–silicate immiscibility and the consequent loss of buoyant and immiscible Si-rich melt; (4) dissolution of unprotected droplets of sulphide melt present in the Fe-rich mush melt; (5) compaction-driven upwards loss of residual mush melt enriched in, for example, Au. The LPGEM preserves upward increasing bulk Pd/Pt (∼6–13) owing to a continued supply of PGE and Au, with high Pd/Pt. The further development of the LPGEM ceased as the supply of precious metals to the floor waned. The Upper PGE Mineralization (UPGEM) subsequently formed from precious metals recycled in the floor. The UPGEM is characterized by increasing Au substitution in PGE phases, and a decrease in total PGE and Pd/Pt owing to upward fractionation in migrating mush melts and exhaustion of Pd and Pt. An upper Au-rich mineralization level (UAuM) was caused by late remobilization of Au and deposition on grain boundaries in fully crystallized gabbro. Cu concentrations (∼150 ppm) are not correlated with PGE and Au. Repeated Cu mineralization levels (CuM), attaining >1000 ppm, occur above the Au levels, caused by local mush layer sulphide saturation. PGE, Au and Cu distributions in the floor mineralization reflect sub-liquidus, but supra-solidus, processes and reactions in mushes at the roof, wall and floor. Constraints provided by a new model for the mineralization provide the basis for re-evaluation of the solidification processes in the Skaergaard intrusion. We have identified the importance of extensive in situ fractionation and intrusion-wide elemental redistributions in immiscible Fe- and Si-rich silicate melts. Our model characterizes the floor cumulates as bulk liquid orthocumulates containing an upwards-increasing proportion crystallized from Fe-rich, immiscible mush melt. The roof-rocks are complementary to the floor, with downwards increasing proportions crystallized from the conjugate Si-rich melt. Petrographic observations and the relative timing of crystallization support the hypothesis that crystallization was restricted to marginal mush zones. Bulk melt remaining in the magma chamber evolved not, as generally assumed, as a result of loss of crystals grown from the bulk melt, but as the consequence of mixing with recycled and evolved melt expelled from the mush by compaction. Redistribution of Fe in immiscible melts may be common to mafic intrusions and puts into question the validity of petrogenetic modelling of bulk liquids in mafic intrusions based only on consideration of floor cumulates.