Cr-rich spinel is a proven and powerful tool for deciphering petrogenetic history of basaltic lavas and for estimating host melt Mg# (e.g., Allan et al., 1988; Sack and Ghiorso, 1991a, 1991b; Allan, 1992, 1994; Allan et al., 1996). Here we use the algorithm of Allan (1992, 1994) to calculate magmatic Mg# from the Cr-spinel composition. This algorithm has been extensively tested on a wide variety of water-quenched alkaline and tholeiitic basalt where the quenched glass composition is known. It assumes that Cr-rich spinel behaves as an ideal reciprocal solution (Wood and Nicholls, 1978; Sack, 1982; Allan et al., 1988), that ideal mixing occurs between Mg and Fe²⁺ in the melt, and that the pressure effect on spinel-melt equilibrium is negligible (Roeder and Reynolds, 1991). The algorithm uses a measured or assumed distribution coefficient (K_D) for Mg-Fe²⁺ exchange between olivine and melt to allow algebraic derivation of the spinel-melt K_D from the experimentally well-established olivine-spinel K_D (e.g., Jamieson and Roeder, 1984). As olivine is totally altered in these samples, we assume a K_D for Mg-Fe²⁺ exchange between olivine and melt of 0.3 (Roeder and Emslie, 1970). Estimated melt compositions in equilibrium with individual Cr-rich spinel composition are given in Table 2 and in Figure 3, Figure 7, and Figure 8 for the analytical transects, with ranges defined from unzoned core compositions for individual samples given in Table 4. As most SDRS Cr-rich spinels have undergone at least partial re-equilibration, it is prudent to regard these inferred melt Mg# compositions as minimum estimates.

Cr-rich spinels within both Units 1 and 2 at Site 338 define trends of increasing Cr# with decreasing Mg#, with little variation in Fe³⁺# (Fig. 6). These trends are commonly seen in tholeiitic basalt undergoing low-pressure olivine and plagioclase fractionation, where melt Al₂O₃ and MgO (and Mg#) decline with increasing degree of fractionation (Allan et al., 1988). As a result, spinel Al₂O₃ and MgO directly decline, while spinel FeO and Fe₂O₃ increase. This effect is magnified by the reciprocal nature of the spinel solid solution, resulting from the strong coupling of Mg and Al and Fe²⁺ and Cr. Therefore, spinel Cr# will rise substantially as the Mg# decreases. Both units have Cr-rich spinel populations that appear to reflect spinels crystallized from a range of melt compositions, as evidenced by analyses 1, 2, and 3 in Table 2 derived from the same thin section. The simplest interpretation is that the higher Mg# Cr-rich spinels crystallized from more primitive parental magma; the original compositions of these high-Mg# spinels have been preserved by effective "armoring" afforded by the surrounding plagioclase and represent evidence of more primitive melt in the genesis of the host lava. This is consistent with the lower TiO₂ contents of the higher Mg# spinel (analysis 1). For both units, comparison of the analyzed whole-rock Mg# with melt Mg# inferred from the Cr-spinel compositions (Table 4) shows that these spinels crystallized from melt similar in composition to the values recorded by the whole-rock analyses and indicate that the whole-rock compositions have been only slightly affected by alteration. The data presented here confirm that melt with Mg# as high as 0.64 erupted within the Vøring margin SDRS magmatic system.

Samples analyzed from this series include examples of the most primitive lavas yet cored from the North Atlantic SDRS. Cr-rich spinels from these picrites are notable for their size (to well over 200 µm; Fig. 4, Fig. 8) and for their generally high Mg# (Table 2; Fig. 6, Fig. 8). Slight variation in Cr# at a given Mg# and TiO₂ exists within individual samples (see analyses 10-12 from Unit 21) (Table 2; Fig. 6,); variations of this size are common within single-specimen spinel populations (Allan et al., 1988). The high Mg# Cr-rich spinels from the Units 21 and 31B picrites provide direct evidence that highly magnesian melts with Mg#s at least as high as 0.70 are part of the East Greenland SDRS magmatic system. Comparison with their whole-rock Mg#s (Table 4) shows that these Mg# melt estimates are lower, likely reflecting whole-rock Mg-enrichment by accumulation of olivine (as argued by Thy et al., 1998) and alteration of mesostasis.

Less magnesian-Cr-rich spinels from these units and from the Unit 25 picrite are enigmatic, as they imply that melt of significantly lower Mg# could have been present in the picrites during or after eruption. As an example, the groundmass Unit 21 Cr-rich spinel shown in Figure 3B (with a calculated Mg# of 0.62) could represent re-equilibration with evolving interstitial melt after eruption. Alternatively, the Mg# discrepancy could reflect that the actual melt in the picrite was evolved before eruption, with the high-Mg# spinel inclusions severely out of equilibrium with the host melt and representing xenocrysts. An example similar to the latter case was described by Allan (1994) for a picritic lava from Site 839 in the Lau Basin, southwest Pacific. This sample contained very large olivine and Cr-rich spinel crystals that grew from a melt with an estimated Mg# of >0.75, with both residing in quench glass with an Mg# of 0.61. These crystals were sharply zoned at their margins, with the margins in equilibrium with the glass. Apparently, magma mixing occurred immedi-ately before eruption between highly primitive, phyric melt and evolved, aphyric melt. Cr-rich spinels from Unit 25 likely equili-brated with an evolved melt before or during eruption, although it is uncertain whether the interstitial melt was evolved on eruption or became evolved during cooling after eruption. It is highly unlikely that such large Cr-rich spinels in Unit 25 grew from the inferred evolved melt compositions given in Table 4; their low Mg#s likely reflect re-equilibration with evolving, interstitial melt or Mg-Fe²⁺ exchange during rock alteration. The coarse-grained groundmass of this sample (Larsen, Saunders, Clift, et al., 1994) is certainly consistent with extended lava cooling and interstitial melt evolution after eruption. In any case, Units 21 and 31B Cr-rich spinels provide evidence for highly primitive melt that existed at eruption or before eruption within the East Greenland SDRS system.

Cr-rich spinels within the olivine-phyric basalt of Unit 31A also record evidence for fairly primitive compositions, although the inferred Mg# is slightly below that given by the whole-rock composition (Table 4). This difference in estimated and measured Mg# is likely caused by two factors. First, the rock is highly altered, as shown by high LOI (5.12%; Table 1). Secondly, most of the Cr-rich spinels studied in this sample were not completely protected from interstitial melt, as shown by the common presence of thick magnetite jackets (see Transect J in Fig. 4). Although an analytical plateau is shown by Transect J in Figure 8, it is not well-developed, and the core of this crystal may have been affected by a Mg-Fe²⁺ exchange with evolving, interstitial melt or perhaps during rock alteration. The intergranular and subophitic nature of its coarse groundmass (Larsen, Saunders, Clift, et al., 1994) underlines the possibility of exchange with interstitial melt.

Cr-rich spinels within the aphyric olivine basalt Units 7 and 13 also provide melt Mg# estimates below that of the whole-rock analyses (Table 4), despite the presence of fairly well-developed plateaus in Transects C and D (Fig. 8). Spectacular zoning is visible in this crystal (Fig. 4); moderate decreases in Cr# (0.47-0.50 in both transects) are accompanied by slight increases in spinel Mg# expected because of crystal chemical reasons from Mg-Al and Fe²⁺-Cr coupling (Fig. 8; see Allan et al., 1988 for similar examples). These slight increases in Mg# are superimposed on the decreasing Mg# trends (from Mg-Fe²⁺ exchange with melt or during alteration) as the margins of the crystal are approached. This internal zoning, highlighted in the BSE images by the variation in Cr#, may reflect changes in spinel composition associated with magma mixing before eruption; supporting evidence is given by the broad range in the Cr# at a given Mg# of the Cr-rich spinels in this unit (Fig. 6). Evidence for magma mixing is also provided in Unit 13 by its compositionally mixed spinel population, most effectively shown by wide differences in TiO₂ content (see analyses 8, 9 in Table 2).

Cr-rich spinels in the Lower Series samples are substantially smaller than those examined from the Upper Series lavas (Fig. 4), making their cores more vulnerable to the effects of posteruptive Mg-Fe²⁺ exchange. The more primitive Units 87 and 92 are characterized by coarse-grained groundmass (Larsen, Saunders, Clift, et al., 1994) and relatively high LOI (Table 1). For the olivine-phyric Unit 92, the relatively small disparity between the higher calculated melt Mg#s and the whole-rock Mg# (see analyses 20, 21 in Table 2) may result from a combination of olivine accumulation (the unit contains 19% altered olivine, with >8% classified as phenocrysts) (Larsen, Saunders, Clift, et al., 1994) and alteration. Other Unit 92 Cr-rich spinels that give lower melt Mg# estimates have likely experienced some Mg loss from posteruptive magmatic or alteration exchange. Unit 87 spinels have clearly been affected by Mg loss, leading to estimates of melt Mg# that are too low to be consistent with Cr-rich spinel growth (Table 4).

In contrast to the more primitive Lower Series samples, Units 60 and 68 are substantially fresher with finer-grained groundmass (Larsen, Saunders, Clift, et al., 1994). Estimates of melt Mg# derived from unzoned Cr-rich spinel inclusions of Unit 68 (see analysis 18 in Table 2) are in good agreement with the whole-rock analysis, supporting petrographic observations that indicate a relative lack of olivine accumulation and alteration effects.

Unzoned Cr-rich spinels from Unit 60 instead give estimates of melt Mg# substantially below that defined by the whole-rock analysis (see analysis 17 in Table 2, Table 4). This sample is the only one in this study to have occasional relict cores of fresh olivine, with two very different olivine analyses from this unit reported by Demant (1998) (Fo₈₃ and Fo₇₅). The high TiO₂ content (1.5%-2.6%, varying inversely with spinel MgO) of these small spinel inclusions (most are below 20 µm in size) (Fig. 4) is notable, and it implies that they grew or re-equilibrated with melt more evolved than that represented by the whole-rock composition. The highly phyric character of the sample (5%-8% olivine, 5-7% plagioclase, and 3% clinopyroxene pheno-crysts in a very fine-grained matrix), and the fact that all phenocrysts are rounded and give petrographic evidence of being partially dissolved, is consistent with the variable olivine and Cr-rich spinel compositions in implying a mixed magma with a disequilibrium mineral assemblage.

The results given here provide evidence that highly primitive melts erupted during the formation of the SDRS, with erupted melt Mg#s at least as high as 0.70. Another important result of this study is the conclusion that the majority of lavas erupted in the North Atlantic SDRS were too evolved to have precipitated Cr-rich spinel, especially those of the Vøring Margin. Studies of water-quenched, glassy tholeiites indicate that melt Mg# and Cr are generally >0.60 and 250 ppm, respectively, in tholeiitic melts crystallizing Cr-rich spinel (e.g., Allan et al., 1987, 1988), though this is dependent as well on fO₂ (Roeder and Reynolds, 1991; Sack and Ghiorso, 1991a). Unequivocal evidence for truly primitive, near-primary melt exists only among the picrites of the Upper Series at Hole 917A, a tem-porally-limited part of the southeastern Greenland SDRS section. Further studies in progress may extend this evidence to cover the picrites of the Lower Series as well. The melt Mg#s inferred by the Leg 38 Cr-rich spinels indicate that these Vøring Margin melts underwent considerable fractionation after their generation in the upper mantle (for consideration of parental melts in the North Atlantic Volcanic Province see Fram and Lesher, 1993, 1997). The conclusion is that the North Atlantic SDRS are largely comprised of evolved basalts that have undergone considerable fractionation and heat loss since their generation by mantle partial melting. Recent seismic studies (Larsen et al., 1988; Dahl-Jensen et al., 1997; Korenaga et al., 1997) indicate that the Layer 3 crustal thickness underlying the SDRS basalts is abnormally thick, both in terms of total thickness and in relation to Layer 2 thickness, consistent with large amounts of fractionation of SDRS basalt parental lavas at the base of the crust or within the deep crust. This abundance of evolved SDRS basalt is both at proximal- and distal-cored SDRS sites from the inferred location of plume impact near Site 988 (Larsen, Duncan, Allan, et al., 1996), and it represents a result similar to that found by Larsen et al. (1989) for the break-up-related, East Greenland basalt of the Scoresby Sund region.