The Tertiary volcanic centre that dominates the constitution of Rhum, one of the Scottish Inner Hebridean group of islands, is immediately provocative to the visiting geologist. The higher mountains are entirely ultrabasic in composition and their near-horizontal, terraced layers remain nearly constant in composition to about 1 km above sea level. Any initial assumption of geological simplicity is quickly dispelled by those and more detailed features. Horizontal layers of ultrabasic rocks with coarsely crystalline texture do not fit happily on a Tertiary landscape, towering high above the adjacent Hebridean lava flows of similar age. And what of the rocks themselves? Thin layers entirely of calcic plagioclase feldspar compete for attention with thick olivine-rich layers, while any day of traversing the Hallival or Askival mountain slopes will reveal a wealth of superbly exposed question marks in the form of spinel–sulphide layers, slump structures, finger structures, ‘harrisitic olivine’ formations, undulatory ‘trough-like’ structures, ultrabasic veins and currently lesser features. Search as one may (and little is hidden from view), there is no marginal gabbroic envelope between the ultrabasics and the country rocks that would lead to a Skaergaard analogy, and no overlying gabbroic layered rocks that would elicit guidelines from the Bushveld or Stillwater intrusions. This is Rhum, and much of what is on that small island is unique and, indeed, still defies full explanation. It is the achievement of this issue, through its several contributors whose voices succeed mine, that the uniqueness of Rhum is welcomed, many of the problems challenged, and many of the secrets unravelled.

Several small outliers of Tertiary lavas and sediments rest with strong unconformity on a buried landscape eroded from Torridonian sediments and Tertiary granophyre. Erosion continued during the period of sediment and lava accumulation. Four formations are recognized; these are, in order of increasing age, the Orval Formation (hawaiite and basaltic hawaiite lavas), the Guirdil Formation (icelandite lavas, interbedded conglomerates), the Upper Fionchra Formation (tholeiitic basaltic andesite lavas, hyaloclastite deposits, basal conglomerate) and the Lower Fionchra Formation (alkali and transitional basalt, basaltic hawaiite and hawaiite lava flows, basal conglomerate); each is separated by an erosional interval. Clasts in the conglomerates reveal a history of erosion of a terrain exposing gneisses, Torridonian sediments, igneous rocks derived from the Rhum Tertiary Central Complex (including allivalites), and Tertiary lavas of local origin but also including, in the oldest conglomerates, tholeiitic basalts not now preserved on or near Rhum. Prior to and during lava and sediment accumulation, erosion on Rhum had cut down to a level similar to that of the present day, although not to the extent that high-grade thermally altered rocks, which are a marked feature of the Central Complex, were being eroded in any quantity. A sequence of east–west trending valleys, possibly initiated on the line of the earlier Main Ring Fault, drained the area of the Central Complex which then, as now, must have been high ground. Small lakes occasionally formed in the valleys allowing the accumulation of fine-grained sediment with plant remains, and promoting the formation of hyaloclastite deposits when buried by later flows. No source for any of the lava formations is preserved on Rhum; they are thought to have come from feeders north of Rhum, possibly near Canna, and to have ponded against the hills and valleys near and in the Central Complex.

The oldest tholeiitic lavas, not now found in situ, were followed by alkali and transitional flows compositionally similar to the Skye Main Lava Series but characteristically feldsparphyric; the most mafic also contain phenocrysts of magnesian olivine (with included Cr-Al-rich spinels) and aluminous spinel. Both the early alkalic/transitional basalts and the youngest hawaiites and basaltic hawaiites equilibrated at pressures < 9 kb; the tholeiitic basaltic andesites and icelandites equilibrated at relatively shallows depths.

Apart from a few N–S to NW–SE-trending basalt dykes, the lava formations represent the youngest Tertiary igneous event on Rhum.

In southeast Rhum, a Mesozoic/Tertiary sequence is preserved as a fault-bounded and rotated wedge. This is juxtaposed between Precambrian rocks (Torridonian sediments and Lewisian gneiss) and caught up in the complex structure of the Tertiary Main Ring Fault (MRF), which shows three distinct phases of movement. The Mesozoic rocks comprise fossiliferous limestone, sandstone and shale, which show differing degrees of thermal metamorphism depending on their relationships to the Layered Complex. On the basis of faunal and lithological evidence the Mesozoic sediments have been correlated with the Lower Liassic Broadford Beds. The Rhum sediments are overlain by sheared Tertiary basalts, the contact between them probably representing the original landscape unconformity. The presence of these younger strata caught up along the MRF provides crucial evidence for a major phase of central subsidence in the early history of the Rhum Tertiary volcanic centre.

The Cnapan Breaca felsite sheet contains strongly devitrified pyroclastic rocks including welded tuffs. These indicate formation in a high level, surficial or sub-volcanic environment. The gross structure and lithological associations of the sheet are similar to those of some ignimbrites. These observations support a resurgent caldera model for the evolution of the Rhum centre and raise some doubts about previous interpretations of the felsite-breccia association on the island.

Early Tertiary igneous activity on Rhum was preceded by doming and the formation of a major arcuate fault system, the Main Ring Fault (MRF), within which Lewisian gneisses, Torridonian sediments and younger rocks were uplifted by as much as 2 km. Doming and uplift are attributed to the diapiric rise of acid magma which ultimately formed the granophyres and felsites of Rhum. Felsite emplacement was accompanied and immediately preceded by the formation of explosion breccias and tuffisites. This phase involves massive gas escape along the MRF fractures; it marked a period of major subsidence within the MRF during which fossiliferous Jurassic sediments and relics of Tertiary lava flows were brought to low structural levels within the MRF. Finally, a further period of uplift, again of about 2 km, took place once more bringing gneisses and basal Torridonian sediments within the MRF to high structural levels. The driving force for this last phase of uplift may have been provided by a further uprise of acid magma or, more realistically, may have been directly connected with emplacement of layered ultrabasic rocks which now form the core of the Rhum centre.

The Lower Eastern Layered Series (LELS) comprises approximately units 1–5 of previous workers. Remapping has revealed considerable along-strike lithological variation in the units of the LELS. It is suggested, on the basis of field and geochemical evidence, that two layers formerly regarded as ‘conformable intrusive sheets of fine-grained olivine gabbro’, may be evolved alii valite layers rather than later intrusions. Xenolith suites, in these layers and elsewhere, indicate a component derived from the roof or walls of the magma chamber. Cryptic variation is more extensive in the LELS than in other parts of Rhum: olivine forsterite content varies from 85.6 to 70, and clinopyroxene Mg×100/(Mg+Fe) varies from 88 to 74. Postcumulus effects and subsolidus re-equilibration have altered the initial compositions of the mineral phases. The data are consistent with a model of repeated replenishment by picritic magma, although the replenishing liquids may have been slightly less magnesian than those subsequently available, during the formation of the upper ELS.

Lithological, major element, trace element and Sr isotope data from cyclic unit 10 of the Rhum Eastern Layered Series are presented. The lower 65 metres of the unit are peridotite, subdivided on textural and geochemical grounds into a lower homogeneous portion approximately 50 metres thick and an upper heterogeneous portion approximately 15 metres thick. The uppermost 16.5 metres of the unit are allivalite. There are steep geochemical gradients across the peridotite-allivalite boundary in Ni content of olivine and whole-rock Sr isotope composition.

Calculations are presented on the geochemical evolution of a Rhum picritic liquid undergoing olivine precipitation, both when the olivines remain suspended in the residual liquid as they precipitate, and when they are continuously fractionated. Quenched groundmass and olivine compositions from the Rhum dykes and the unit 10 peridotite olivines show good agreement with the suspension model but are inconsistent with the fractionation model. The Rhum chamber is thought to have been replenished with a picritic liquid from which olivine crystallized while held in suspension; however, replenishment by a highly olivine-phyric basalt is also possible. The peridotite probably accumulated rapidly as olivines were dumped out of suspension onto the chamber floor.

The lower part of the peridotite is a poikilitic adcumulate; it is suggested that this formed by convective circulation of melt in the pores of the pile of cumulus olivines. In the latter stages of adcumulus growth, more Fe-rich and isotopically contaminated magma entered the top of the cumulus pile causing cumulus olivines to re-equilibrate and giving the intercumulus plagioclase a higher Sr isotope ratio than lower down. The olivines in the allivalite show steep stratigraphic gradients in major element composition but not in their Ni content. They also show substantial variation in major element composition laterally within the allivalite. It is suggested that these features are a consequence of postcumulus re-equilibration of olivine with migrating intercumulus magma.

Sr and Nd isotopic compositions together with whole rock Sr and Nd abundances and olivine compositions are presented for cumulates from unit 10 of the Eastern Layered Series of the Rhum intrusion, Northwest Scotland.

The results show an Nd–Sr isotopic correlation consistent with contamination by high 87Sr/86Sr, low 143Nd/144Nd crust. There is a strong correlation between the degree of contamination, Sr and Nd whole rock abundance and Fe enrichment in cumulus olivine. The allivalites are the most contaminated, while the basal peridotites are uncontaminated. This suggests that the cumulates crystallized from a primitive uncontaminated magma and the allivalites from an evolved contaminated magma. Subsequently, movement of contaminated feldspathic intercumulus liquid has perturbed the isotopic, trace element and olivine composition of the upper peridotites.

Finger-like protrusions of peridotite are developed on Rhum where peridotite is overlain by allivalite. These structures, which were described by Brown as ‘upward-growing pyroxene structures’, are found in the following environments: at the main intra-unit junctions; along the upper surface of subsidiary peridotites in certain allivalites; and along the lower surface of allivalite blocks in some peridotites.

The structures generally take the form of parallel-sided or tapering protrusions with circular cross-sections. The tops of fingers are conical or hemispherical in shape. Typical dimensions are: finger amplitude, 2–5 cm; finger diameter, up to 3 cm; and finger wavelength, 5–10 cm. Peridotite in the finger is modally and texturally similar to the underlying layer, varieties range from feldspathic peridotite to dunitic peridotite. In the field the fingers apparently cut through layering, laminae and lamination without any associated disruption of the planar structures.

Two contrasting mechanisms of formation are discussed: vertical deformation of crystal mushes, and metasomatic replacement. On balance, we prefer to interpret the fingers as evidence for the replacement of pre-existing allivalite by secondary peridotite. Replacement was achieved by pore magma from the underlying peridotite migrating upwards into the overlying allivalite, in response to compaction. This pore magma was able to resorb plagioclase but crystallize olivine and pyroxene in its place.

Late-stage veins have been studied in the Rhum layered intrusion as part of a wider investigation into postcumulus processes – in particular, to study compositions of residual liquids and their effects on host cumulates. Studies have concentrated on Unit 10 of the Eastern Layered Series (ELS), though brief investigations have also been made in the Western Layered Series (WLS).

The Unit 10 veins are generally concordant with the layering, although locally they are transgressive with dilational relationships. Two principal types can be recognized: ultrabasic and basic. Detailed microprobe studies have shown that the liquids responsible for the veins have substantially modified the compositions of the cumulus minerals in the wall-rock peridotite up to 1 cm from the vein margin.

The style of veining in the WLS is more complex than in the ELS, and several generations of vein can be identified. Both normal and harrisitic cumulates have undergone localized chemical modification near to the veins. Modification occurs regardless of the vein orientation.

The ELS and WLS veins are geochemically distinct from the numerous gabbro intrusions associated with the layered series, and it is concluded that they represent fractionated intercumulus liquids injected into early formed fractures in the cumulates. The observed chemical modification is explained by a model involving mixing of the intrusive vein melt with trapped intercumulus melt in the adjacent peridotite, followed by diffusion-controlled re-equilibration of the cumulus minerals. This type of highly localized compositional modification along fractures is termed channelled metasomatism.

In the allivalite of cyclic unit 9 in the Eastern Layered Series of the Rhum intrusion there is an undulatory contact between a relatively dense pyroxene-rich rock and a less dense pyroxene-poor rock. The three-dimensional form of this contact resembles that developed between two fluids deforming under the influence of gravity. Minor layers, laminae and lamination in these rocks in places cut across the contact between the pyroxene-rich and pyroxene-poor rock types. They are, therefore, inferred to post-date the formation of the contact and hence to have formed within a porous crystal pile on the chamber floor. Possible mechanisms of formation are discussed but their origin remains unclear. These observations cast doubt on the common assumption that the combination of granular (cumulus) textures and layering structures is sufficient evidence on which to infer equilibrium of granular crystals with a contemporary magma.

Feldspathic peridotites on Rhum contain teschenite segregations and veins. Bulk rock and mineral compositional trends define a path of alkali and iron enrichment related to discontinuous isolation of evolving intercumulus melts. Compositionally zoned and skeletal crystals of actinolite, zircon, apatite, sphene, epidote and chlorite are enclosed by clear analcime, and appear to have grown directly from a water-rich fluid at some undercooling, and below dry equilibrium solidus temperatures. The occurrence of late stage residua with undersaturated alkalirich characteristics in the Rhum ultrabasic complex suggests that the patental magma was alkaline rather than tholeiitic, and had similarities to the Plateau Magmas on Skye.

A microscopic and electron microprobe investigation has been made of six samples of the allivalite–chromitite–peridotite band at the junction of Units 11 and 12, Eastern Layered Series, Rhum. Analyses are presented of olivine, plagioclase and pyroxene, the plagioclase showing reverse zoning in the chromitite. Sulphide droplets occur within and above the chromitite, and contain pyrrhotite, pentlandite, cubanite, bornite, digenite, chalcocite, native copper and electrum, as well as chromite zoned to magnetite, spinel and ilmenite. The variation from aluminous chromite to chromite in the chromitite is confirmed by analyses of five new traverses. The rocks formed during and just after the time when new magma entered the magma chamber. The variation in the chromite compositions is ascribed mainly to magmatic variation in composition due to the mixing process.

The terminology of cumulates is discussed with particular reference to postcumulus processes as exemplified by the Rhum layered intrusion. The history of the terms orthocumulate, adcumulate, heteradcumulate and harrisitic cumulate is reviewed, and some modifications to the classification scheme proposed by Irvine (1982) are suggested. Possible mechanisms of adcumulus growth are considered and, on balance, the Rhum evidence confirms the view that postcumulus cementation occurred at a relatively early stage, while the intercumulus melt was in effective contact with the main body of magma.

During the postcumulus stage of solidification in layered intrusions, fluid dynamic phenomena play an important role in developing the textural and chemical characteristics of the cumulate rocks. One mechanism of adcumulus growth involves crystallization at the top of the cumulate pile where crystals are in direct contact with the magma reservoir. Convection in the chamber can enable adcumulus growth to occur to form a completely solid contact between cumulate and magma. Another important process may involve compositional convection in which light differentiated melt released by intercumulus crystallization is continually replaced by denser melt from the overlying magma reservoir. This process favours adcumulus growth and can allow adcumulus growth within the pore space of the cumulate pile. Calculations indicate that this process could reduce residual porosities to a few percent in large layered intrusions, but could not form pure monomineralic rocks. Intercumulus melt may also be replaced by more primitive melt during episodes of magma chamber replenishment. Dense magma, emplaced over a cumulate pile containing lower density differentiated melt may sink several metres into the underlying pile in the form of fingers. Reactions between melt and matrix may lead to changes in mineral compositions, mineral textures and whole rock isotope compositions. Another important mechanism for forming adcumulate rocks is compaction, in which the imbalance of the hydrostatic and lithostatic pressures in the cumulate pile causes the crystalline matrix to deform and intercumulus melt to be expelled. For cumulate layers from 10 to 1000 metres in thickness, compaction can reduce porosities to very low values (< 1%) and form monomineralic rocks. The characteristic time-scale for such compaction is theoretically short compared to the time required to solidify a large layered intrusion. During compaction changes of mineral compositions and texture may occur as moving melts interact with the surrounding matrix. Both compaction and compositional convection can be interrupted by solidification in the pore spaces. Compositional convection will only occur if the Rayleigh number is larger than 40, if the residual melt becomes lower in density, and the convective velocity exceeds the solidification velocity (measured by the rate of crystal accumulation in the chamber). Orthocumulates are thus more likely to form in rapidly cooled intrusions where residual melt is frozen into the pore spaces before it can be expelled by compaction or replaced by convection.