Most igneous activity in the ocean basins occurs at plate boundaries. Ocean-floor basalts are extruded along mid-ocean ridges that define divergent boundaries, whereas island-arcs are produced as the ocean floor is subducted at convergent plate boundaries. In contrast, oceanic island, oceanic plateau and seamount volcanism is distinctive in composition and eruptive style from the activity at plate margins. These sub-sea and subaerial land forms represent <1 % of all igneous rocks on Earth and ~2% of all volcanic activity in the deep ocean basins. However, enormous emphasis has been placed on them for understanding igneous processes and the evolution of Earth. Separated from contaminating continental crust, and unaltered by mountain-building metamorphic and tectonic processes (Basaltic Volcanism Study Project = BVSP, 1981, p. 161), oceanic island basalts (OIB) provide insights into mantle composition, magma formation processes and magma evolution (Fig. 1).

The significance of oceanic islands is apparent in any literature review. Their archipelagoes led to the formulation of fundamental plate tectonic ideas (Wilson, 1963), have influenced hypotheses for how convection drives plate tectonics (Morgan, 1971; Wilson, 1973; Hamilton, 2003), and today fuel a healthy debate about chemical layering in the Earth (Anderson, 1999; Allègre, 2002). Further, the gravitational instability of oceanic islands indicates that collapses could imperil major cities bordering the ocean basins (McGuire, 1996). Thus, studies of oceanic islands have both intellectual and hazard-alleviation significance for humans.

This is the first of two complementary papers on oceanic island basalts. It reviews the mineralogy, classification, tectonic significance and lithosperhic processes (e.g., differentiation) affecting OIB magmas. The second paper (Greenough et al., in press) examines the mantle processes affecting OIB magma generation, including the influence of source mineralogy and the effects of fluids and percentage melting on magma composition; it also provides isotopic and trace element evidence for five distinct types of mantle source regions. All aspects of oceanic island rocks, including petrography, volcanology, and geochemistry, are inter-related. Thus, no one scale of observation is entirely satisfactory to understanding their origin. This paper begins with hand-specimen scale features, and ends with the effect of differentiation processes on magmas.

Oceanic island rocks form two major families based on mineralogy and petrography: those that form from tholeiitic basalts and those from alkali basalts. Mantle-melting processes lead to chemical gradation between the two, but under low-pressure conditions in the lithosphere, each basalt type crystallizes different minerals and evolves in very different ways.

Alkali basalts and tholeiites are theoretically distinguished from each other using their modal mineralogy (Fig. 2; Table 1). However, many volcanic rocks are completely glass, or are extremely fine grained, making mineral identification impossible or difficult. To overcome this problem, a whole-rock major-element analysis is used to calculate the normative mineralogy, which estimates the abundances of minerals that would have formed if the rock crystallized (cf. Philpotts, 1990, pp. 88-92). All basalt chemical compositions can be plotted in the basalt tetrahedron, which has normative quartz, olivine, nepheline and augite at the apices. Because all basalts contain augite (modal or normative), compositions are projected onto the Ne-Q-Ol plane (Fig. 3). Alkali basalts are separated from tholeiitic basalts by the Ol-Ab join. This line has great significance, because, if Figure 3 was converted into a phase diagram, the Ol-Ab join would approximately coincide with a high-temperature line or "thermal divide". As any basalt cools and crystallizes minerals under low-pressure conditions, the left-over (residual) liquid's composition moves down-temperature, away from the thermal divide and Ol-Ab join. Thus, once in the lithosphere (low pressure), the composition of magmas cannot generally cross this line resulting in two series of magmas, tholeiites (Hy side of Ol-Ab join) and alkali basalts (Ne side).

The most fundamental characteristic of tholeiites is the presence of a low-Ca pyroxene (e.g., hypersthene), most commonly in the matrix, locally as phenocrysts (Fig. 2a), or as calculated in the norm. Thus, tholeiites fall in the Ol-Ab-Hy or Q-Ab-Hy triangles of the basalt tetrahedron (Fig. 3). Unfortunately, rocks plotting close the Ol-Ab join can have small percentages of normative Hy but no modal low-Ca pyroxene. The thermal divide and Ol-Ab normative mineralogy join do not precisely coincide. Various factors, including magma composition, control whether orthopyroxene or pigeonite occurs as the low-Ca pyroxene, but high temperatures yield orthopyroxene and low-pressure crystallization favours pigeonite. Only tholeiites can have quartz, which, if present, is usually restricted to the matrix. The most common phenocryst silicate phases are olivine, plagioclase and augite, typically, although not always, appearing in that order (Fig. 2b).

Alkali basalts (Table 1) contain nepheline (Fig. 2d), commonly only in the norm, and plot in the Ol-Ne-Ab triangle of the tetrahedron (Fig. 3). Basanites contain modal nepheline. Less commonly, mantle melting yields magma with such low silica activity that plagioclase cannot form. These rocks, nephelinites, only contain nepheline + melilite. Ultramafic xenoliths primarily occur only in alkali basalts. The most common phenocryst phases are the same as those in tholeiites, but tend to form in the order olivine, augite and plagioclase. Augite is generally complexly zoned and pleochroic and has Ti-rich rims (Fig. 2c). Both alkali and tholeiitic basalts contain chromite as an early crystallizing phase, and form magnetite and/or ilmenite over a large range of temperatures.

Mineral compositions are closely connected to magma alkalinity (degree of silica saturation or undersaturation), the sequence of mineral crystallization and degree of magma evolution. Table 2 contains representative analyses of major silicate phases discussed below (see supplementary materials at http://www.gac.ca/JOURNALS/geocan.html for references and islands represented on plots).

Oceanic-island volcanism occurs at the ends of hotspot trails (e.g., Hawaiian-Emperor), which can, in places, be traced, using seamounts, across major parts of ocean basins (Fig. 1). The volcanically active trail-ends are preferentially located near divergent plate boundaries and are rare near convergent boundaries (Zhao, 2001). Wilson (1963) proposed that the trails reflect movement of oceanic plates over fixed mantle hotspots, an idea central to acceptance of plate tectonic theory. Many scientists think that hotspots mark locations where diapiric convection cells, called mantle "plumes", rise beneath lithospheric plates (Morgan, 1971; Wilson, 1973). Detailed studies of the angular relationships of hotspot trails show that the distance to the north magnetic pole, and by assumption the north geographic pole (= spin axis), has changed by up to 8 mm/y over the past 100 Ma, providing evidence for "true polar wander" or movement of hotspots relative to the geographic pole (Duncan and Richards, 1991). Also, hotspots probably move relative to one another and this is consistent with fixed hotspots being unlikely, if "plume" models for convection in the mantle are correct (Koppers et al., 2001). Nevertheless, these conclusions do not negate the hypothesis that hotspot trails dominantly reflect plate motions. Despite the popularity of the plume model, hotpots (oceanic islands) may simply represent the position of propagating cracks in the lithosphere (Anderson, 1999; 2000; Hamilton, 2002; 2003; see Greenough et al., in press, for a brief review of this model).

Given an internal heat source from radioactive decay, the size of the Earth, the low thermal conductivity of silicate rocks, and the temperature dependence of viscosity, mantle convection seems inevitable (Griffiths and Turner, 1998). Based on observations of the length of subduction zones, subduction and spreading rates, length and height of ocean ridges, and the topographic height of "hotspot" swells, there may be two modes of mantle convection. The dominant one is driven by cool, high-density, lithosphere sinking at subduction zones with passive rise of mantle below spreading ocean ridges, whereas cylindrical convection cells (plumes) may be responsible for oceanic islands (Davies, 1998). Some models suggest that plumes arise from the lower most, lower mantle, thermal boundary layer (the "D" layer) heated from below by energy leaving the core. Others see cells originating at the 670 km seismic discontinuity, or from mid-mantle depths below the discontinuity (Cserepes and Yuen, 2000). It has been argued that three-dimensional views of seismic wave velocities (seismic tomography) do provide support for hot, low-density (low-velocity) diapirs (plumes) rising from the core-mantle boundary to the lithosphere below oceanic islands such as Hawaii and Iceland (Ji and Nataf, 1998; Bijwaard and Spakman, 1999; Zhao, 2001). The results of these studies and those showing subducted ocean-floor slabs crossing the upper mantle – lower mantle boundary (670 km seismic discontinuity) have recently been questioned (Anderson, 2000; Hamilton, 2003). Thus, plumes may not exist and convection in the upper mantle may be separated from the lower mantle. As discussed in the complementary paper (Greenough et al., in press), this has important implications for models of the Earth's chemical structure and evolution.

Information on the magnitude of hotspot swells (0.5-1 km high * ≤ 1000 km diameter), lithosphere thickness, and melting behavior of assumed mantle materials including xenoliths, yield models for the size and thermal structure of underlying convection cells. These explain changes in magma composition and volume produced by decompression melting as plates move over a hotspot (Watson and McKenzie, 1991; Sen et al., 1996). Figs. 10 and 11 show stages in Hawaiian volcanism (Clague and Dalrymple, 1987). In the pre-shield stage, the plate begins to override an ascending plume and decompression initiates melting. The initial magmas are alkaline (e.g., the subsea Loihi Seamount, southern end of the Hawaiian chain). Elements such as K, Rb, Sr, Nb, Zr, and La are highly concentrated in these incipient, small percentage melts because they are "incompatible" in minerals of the solid mantle residue and prefer melt over solid. In the second or shield stage, the plate is over the plume axis and large amounts of decompression occur. Melts form at the highest temperatures and lowest pressures resulting in large volumes of tholeiitic magma that form shield volcanoes such as Mauna Loa. "Incompatible" elements are diluted in these high-percentage melts. As the plate moves over the plume tail, melting percentages and magma production drop as alkali basalts and their differentiates appear during the post-shield stage. Following 0.3 - 2.0 Ma of dormancy with edifice erosion, small volume, explosive and highly alkaline magmas appear during a rejuvenation stage attributed to volatile-induced melting in a "wet" envelope at the plume tail's periphery.

Lithospheric thickness determines the height to which plumes can rise (or the depth where propagating cracks initiate melting), and so it also affects the depth, volume, percentage of melting and geochemical character of magmas. After new oceanic crust forms, the lithosphere cools and thickens as it moves away from the ocean ridge. Where a plume/crack occurs below/in young and thin (< 75 km) lithosphere, low-pressure melting creates large volumes of tholeiitic magma (e.g., Iceland and Hawaii). If the lithosphere is older and thicker (e.g., Canaries) melting occurs at greater depths resulting in low volumes of dominantly alkaline magma typical of the majority of oceanic islands (Haase, 1996).

Models for the plumbing system feeding flows are mostly based on Hawaiian studies. During the shield-building stage, magma originating from mantle depths ≥ 60 km penetrates a vertical conduit and is stored in a 3 to 7 km deep magma reservoir beneath the volcano's summit (Decker, 1987). During the post-shield stage, more alkalic magmas that have lower production rates are stored in deeper chambers (~15 km; Clague, 1987). Depending on supply rates, magma is periodically erupted from the summit and/or laterally intruded and erupted along rift zones. Rift zone orientations reflect the gravity-induced tensional stress field that is controlled by a volcano's internal structure and morphology (Fiske and Jackson, 1972). The volume (0.08 to 40 km³ ) and shape (interconnected sills and dykes versus spherical pluton) of magma chambers below Kilauea are debated but temporal geochemical variations suggest a small (2-3 km³ ), simple-shaped reservoir, whereas geophysical models "see" a large aseismic zone (chamber) that includes crystal mush (Pietruszka and Garcia, 1999).

The relatively non-explosive eruption of copious quantities of high-temperature, low-viscosity (highly fluid) basaltic lava results in the low slopes characterizing shield volcanoes. High temperature pahoehoe flows near vents have smooth "ropey" surfaces (Williams and McBirney, 1979, pp. 101-112). Lower temperature flows, and flows farther from vents are more viscous and have uneven, jagged or broken surfaces (aa and blocky flows). Pillow lavas form where lava is extruded under water. The outer selvages are glassy and the interior tends to be vesicular under low-water pressures. At great ocean depths (pressure), lower magma volatile contents, and/or higher magma viscosities, pillow interiors are less vesicular. Evolved (differentiated) and alkaline magmas tend to be more explosive reflecting higher viscosity and/or higher volatile content. Thus, post-shield and rejuvenation-stage alkaline magmas are more prone to forming localized, cone-shaped, fragmental deposits (Mullineaux et al., 1987). Magma interaction with groundwater or seawater, either before or in association with eruption, enhances the formation of fragmental rocks (Cas and Wright, 1987, pp. 42-45).

Non-explosive shield volcano eruptions rarely result in human deaths but lava flows, the most common type of volcanic hazard (e.g., Hawaii), can be very destructive (Mullineaux et al., 1987). More daunting is the tendency for oceanic islands to undergo catastrophic collapse. The seafloor surrounding most oceanic islands shows debris-flow deposits stretching hundreds of kilometres from islands (e.g., Moore et al.,1994; McMurtry et al., 1999; Moore and Clague, 2002). Various factors contribute to the collapse (Elsworth and Voigt, 1996; Carracedo, 1996). These are: 1) flows extruded into seawater become quenched and contribute to over-steepening of the volcanic edifice. The result is a tensional stress regime, induced by gravity, which radiates from volcano summits, 2) porous subaerial fragmental deposits can be intruded by impermeable dykes. In humid climates the porous rocks between sub-parallel dykes in rift zones become saturated water reservoirs, and 3) intrusion along rift zones facilitates catastrophic collapse because magma has no shear or tensional strength and it also decreases the strength of surrounding rocks by heating trapped water and raising pore-fluid pressures. The potential for catastrophic collapse has attracted much attention because resulting giant tsunamis could imperil major cities bordering the ocean basins.

Once magmas form in the mantle, various processes, including differentiation, can cause them to evolve. The products of these various processes are highly dependent on the starting magma composition. Basalts and rarer intermediate and felsic rocks can be categorized into three series based on Na₂O + K₂O at particular SiO₂ contents (Fig. 12a; averages used appear in supplementary materials with representative averages in Greenough et al. (in press; their Table 1). Three reference suites from Iceland, Moorea and Tristan da Cunha illustrate (respectively) the tholeiitic series containing basalt to rhyolite, the alkaline series with alkali basalts to trachytes and a strongly alkaline series, beginning with basanite and ultimately yielding phonolite; few oceanic islands contain tholeiite-series rocks. The alkaline series is most common (Fig. 12b). Variations in composition within each series reflect differentiation in magma chambers under volcanoes. Crystals (minerals) have a different bulk composition than the magma they form from. For example, early crystallizing olivine has higher MgO, lower SiO₂ and little Na₂O andK₂O compared to the host magma. Its formation causes the residual liquid (magma) to have lower MgO and higher SiO2 and Na₂O+K₂O content than the original magma. Magma chambers are not exposed, so processes that remove crystals from magma must be inferred from erupted flows. Whether minerals grow in magma and are gravitationally separated (i.e., minerals denser than magma, sink), or nucleate and grow on the floor, walls and roof of chambers from convecting magma, is debated (see Hawkesworth et al., 2000) but most petrologists are of the view that crystal fractionation, in whatever form, is the dominant differentiation process. Intermediate composition lavas tend to be less common, or missing, in some suites, a phenomenon called the Daly gap. The Daly gap has been related to high viscosities (Thompson et al., 2001) and/or higher densities (Renzuli and Santi, 2000) of intermediate magmas, which prevent them from reaching the surface. Evolved magmas can have lower viscosities than intermediate ones because of increased alkalis and volatile contents. This, together with volatile exsolution, which creates buoyant magma, allows evolved magma to erupt.

Various approaches yield information about how magmas evolve. For example, olivine is a common phenocryst phase in both Kilauea and Mauna Loa tholeiites. The effect of olivine formation and removal on magma composition can be calculated, and matches glass compositions observed at both volcanoes. Projection of these liquid compositions onto experimentally determined phase diagrams reveals that the slightly higher silica content of Mauna Loa magmas causes evolving liquids to intersect the olivine -pyroxene cotectic. Therefore, pyroxene is the next mineral to start forming after olivine. Plagioclase comes onto the liquidus after pyroxene. In contrast, olivine formation and removal causes Kilauea magmas to intersect the olivine - plagioclase cotectic. Thus the order of phenocryst appearance in Kilauea lavas is olivine, plagioclase, augite whereas in Mauna Loa lavas it is olivine, augite, plagioclase (Basaltic Volcanism Study Project, 1981, p. 165).

Mass-balance calculations and trace-element modeling are widely used to infer how magmas evolve (e.g., Maaløe et al. 1986; West et al., 1988; Le Roex, 1990; Thompson et al., 2001). Generalizations on processes can be made from variation diagrams showing representatives of the three magma series (Fig. 13). Selected major elements and trace elements are plotted against Mg# (= Mg/(Mg + (0.9*Fe^total )) atomic) which decreases during differentiation reflecting the preferential incorporation of Mg over Fe in ferromagnesian minerals. Nickel, Cr and Co behave as "compatible" elements (i.e., are concentrated in early crystallizing minerals) in all three series with the concentrations falling as Mg# declines. Declining Ni and Co are most commonly ascribed to olivine precipitation. However, globules of high-density, immiscible sulfide liquid can also form in the early stages of magma differentiation. Sulfide liquid removes Ni more rapidly than olivine because of the extremely high-sulfide liquid/silicate magma partitioning coefficients. Some siderophile and chalcophile elements such as Au and the Platinum Group elements (PGE's) are more strongly partitioned into immiscible sulfide liquid than Ni. Thus, the removal of sulfide liquid may explain the compatible behavior (concentrations fall with differentiation) of these elements in many hot-spot magmas (Fryer and Greenough, 1992; Tatsumi et al., 2000). Chromium decreases are commonly attributed to the early precipitation of chromite, and later, the formation of augite.

Strontium tends to increase in basalts and then plunge in more differentiated rocks (Mg# > 0.40) of alkaline and strongly alkaline islands, (e.g., Tristan da Cunha, Fig. 13); plagioclase concentrates Sr. The change from incompatible to compatible behavior heralds the onset of significant plagioclase (or feldspathoid) precipitation/removal. Similarly, Ba behaves incompatibly in mafic magmas but concentrations either level off or fall in highly differentiated rocks. Declining Rb and Ba in evolving, alkaline Hawaiian rocks has been ascribed to feldspar precipitation (West et al., 1988) because both elements have high partition coefficients in K-feldspars (e.g., sanidine).

The rare-earth-elements (except Eu), and high-field strength elements (e.g., Zr, Hf, Nb, Ta) generally behave incompatibly (Fig. 13). However, Nb and Ta are compatible in some Jan Mayen and Hawaiian suites, apparently reflecting Fe-Ti oxide precipitation (Maaløe et al., 1986; West et al., 1988). Phosphorous initially rises and then falls in Tristan da Cunha lavas, but remains constant, or falls, in the Moorea and Iceland suites (Fig. 13). Although P can be incompatible in entire suites (West et al., 1988), apatite comes on the liquidus by intermediate to late stages of differentiation in many suites as indicated by declining P concentrations (e.g., Thompson et al., 2001; Maaløe et al., 1986). Note that because the REE partition into apatite, the proportion of apatite precipitating must be large enough to remove P but small enough to keep the bulk partition coefficient for the REE less than one because their concentrations increase with differentiation. Titanium may initially increase during differentiation, but it becomes compatible in all types of magmas above an Mg# of ~ 0.5 (Fig. 13). Minerals that remove Ti include augite and ilmenite/ titanomagnetite.

An important question is whether the alkaline and tholeiitic series magmas are related by differentiation. Classic experiments by Yoder and Tilley in the 1960's showed that Ne-normative alkali basalts are probably not related to tholeiitic basalts by differentiation because of a low-pressure thermal divide approximately coinciding with the Ol - Ab join in the basalt tetrahedron (Fig. 3). To cross this divide, basaltic magmas would have to increase in temperature as they differentiate (see Petrography and Mineralogy Section). Both series show similar maximum Mg# values and concentrations of differentiation-sensitive transition metals (Ni, Cr, Co, Sc). This precludes fractional crystallization even at high pressures as a process relating the two types of basalts (Dostal et al., 1982). However, Renzulli and Santi (2000) show that under high FO₂ conditions, slightly Ne-normative basalts can yield highly evolved peralkaline, Hy- and Q-normative trachytes through a two-stage fractionation process occurring in deep and then shallow-level magma chambers.

The issue of whether oceanic islands form from mantle plumes or represent propagating cracks in the lithosphere is of fundamental importance and requires a re-evaluation of seismic tomography results. Although the importance of crystal differentiation appears established, the impact of assimilation and effect of volatiles on magma compositions requires further consideration. Studies of the trace-element composition of minerals using new analytical methods such as laser ablation – inductively-coupled-plasma mass-spectrometry, promise refinements in our understanding of differentiation processes as well as alternative methods for assessing relationships between magma chemistry and source region processes and compositions.