An important geological breakthrough of the past decade has been the realization that mantle plumes play an important role in the breakup of continents. Recent models have centered around the role that plume head impingement on continental lithosphere plays in crustal thinning and breakup, accompanied by massive volcanic activity (e.g., White and McKenzie, 1989; Richards et al., 1989; Campbell and Griffiths, 1990). Evidence for this volcanic activity in the early Tertiary North Atlantic volcanic province is supplied both by land exposures and by seaward-dipping reflector sequences (SDRS) that are on the continental shelf and slope. The volcanic nature of the North Atlantic SDRS has been confirmed by drilling (Fig. 1) on Deep Sea Drilling Project (DSDP) Leg 81 (Rockall plateau), DSDP Leg 38 and Ocean Drilling program (ODP) Leg 104 (Vøring Plateau), and ODP Legs 152 and 163 (East Greenland margin). Characterizing the thermal structure of the impinging plumes is critical toward understanding their effects, with estimates of the plume core "excess" temperatures (as compared to the surrounding ambient mantle temperatures) varying widely (100º-300º above the surrounding ambient mantle; see Schilling, 1991; Ribe et al., 1995; and White et al., 1995). In addition, an understanding of the thermal environment of melting is necessary to define the nature of the plume head as it impinges on the continental lithosphere. Fundamentally, does voluminous basaltic volcanism result from an abnormally hot mantle, from large amounts of decompression melting associated with lithospheric thinning, or both? Integration of the drilling results with other onland studies indicates that decompression melting of a rapidly emplaced, hot plume head was the most likely source for the North Atlantic SDRS (Larsen and Saunders, 1998).

Samples from Legs 163 (Duncan, Larsen, Allan, et al., 1996) and 152 (Larsen, Saunders, Clift, et al., 1994; Larsen and Saunders, 1998) define a nearly complete stratigraphic record of the volcanic evolution of the East Greenland SDRS along a transect at 63ºN, with inception of subaerial basaltic volcanism dated at 61-62 Ma (Sinton and Duncan, 1998). These samples are valuable for studying melt generation, transport, and storage processes during continental breakup. In addition, samples from Site 988 (at 66ºN on the East Greenland margin), together with samples from Legs 104 (Eldholm, Thiede, Taylor, et al., 1987), 81 (Roberts, Schnitker, et al., 1984), and 38 (Talwani, Udintsev, et al., 1976) allow an examination of geochemical gradients away from the proposed sites of plume impingement.

A key geochemical parameter for characterizing the crustal and mantle thermal regime during continental breakup is the determination of melt Mg# (Mg/[Mg + Fe²⁺]) in the rift magmatic system (Fe²⁺ is assumed throughout this paper to represent 90% of total Fe, crudely approximating the quartz-fayalite-magnetite oxygen buffer). Melt Mg# reflects a combination of the extent of partial melting (mantle temperature) (Takahashi and Kushiro, 1983; Klein and Langmuir, 1987; Kinzler and Grove, 1992; Niu and Batiza, 1991) and crystal fractionation (effects of cooling, primarily in the lithosphere or at its base; Basaltic Volcanism Study Group, 1981). Picrites were recovered during Leg 152, leading to possibilities that near-primary melts were sampled (Larsen, Saunders, Clift, et al., 1994; Larsen et al., 1998). These highly magnesian lavas may have originated from either abnormally hot mantle or from high amounts of shallow mantle melting during continental breakup (McKenzie and Bickle, 1988; Larsen et al., 1998). Determination of true magmatic melt Mg# compositions for these and other SDRS lavas, therefore, would place direct constraints on plume head thermal influence and the resulting shallow magmatic thermal regime during initial rifting, subsequent continental breakup, and early seafloor spreading.

A significant problem in the interpretation of whole-rock, major element data from all recovered SDRS samples results from their altered nature and from strong textural evidence that many have accumulated olivine. Both effects will skew whole-rock compositions to higher MgO and Mg# values. Leg 152 lavas are pervasively altered (typically >40% alteration, with loss on ignition (LOI) values to 9.5 wt%) with olivine and mesostasis completely altered and plagioclase and pyroxene usually partially altered as well. This alteration involved postemplacement hydrothermal circulation of seawater at temperatures up to 170º (Demant et al., 1988). All other cored SDRS lavas have also been affected by alteration, with complete alteration of olivine and mesostasis (Raschka and Eckhardt, 1976; Desprairies et al., 1984; Desprairies et al., 1989; Viereck et al., 1989; Parson et al., 1989; Duncan, Larsen, Allan et al., 1996). During this alteration, Mg can be exchanged for Ca on a one-to-one basis, resulting in strong depletions in rock CaO and strong enrichments in rock MgO (see Allan, 1992, and references therein). Evidence for this exchange is buttressed by the sediment pore-water geochemistry at Sites 917 and 918 from Leg 152, where dissolved Ca increased and Mg decreased as basement was approached (Larsen, Saunders, Clift, et al., 1994; Gieskes et al., 1998). In addition, accumulation of olivine during magmatic processes has also likely raised bulk-rock Mg# in the more olivine-rich Site 917 samples (Thy et al., 1998; Larsen, Saunders, Clift, et al., 1994). As a result, the compositional nature of the most primitive SDRS melts is uncertain.

In contrast to olivine and other silicate phases, Cr-rich spinel is robust in resisting alteration, and its composition is highly sensitive toward changes in host melt composition (Allan et al., 1988). In addition, the kinetics of re-equilibration of Cr-spinel with changing melt are slow in comparison with olivine, allowing zoned spinel cores and spinel inclusions to record evidence of earlier, more primitive melts than those that were erupted (Allan et al., 1988). In the SDRS lavas, Cr-spinel provides the only available tool for producing direct estimates of melt Mg#. In this paper, we use Cr-rich spinel to estimate the Mg# of the most primitive melts in these SDRS sequences, using the algorithms of Allan (1992, 1994). These results provide first-order constraints on the extent of magmatic processes during the early opening of the North Atlantic.