Electron microprobe point analyses of the SDRS Cr-rich spinels are given in Table 2 and Figure 5 and Figure 6. These analyses are keyed to the BSE images in Figure 2 and Figure 4 either by analytical number or by transect letter label, with the analytical point in the transects given by distance in micrometers from the point of transect origin. Arrows show the direction of analysis for the analytical transects mapped in Figure 2 and Figure 4, which are analytically shown in Figure 7 and Figure 8. In general, Cr-rich spinels in this study are classified as chromium spinels or magnesiochromites, using the criteria and terminology of Sigurdsson (1977). They approximate the formula (Mg, Fe²⁺)(Al, Cr, Fe³⁺)₂O₄, and have minor TiO₂ (most have 0.4%-0.8%, but rarely are as high as 2.6%), NiO (= 0.31%), and MnO (= 0.50%). Microprobe analyses of Cr-rich spinel are found in Table 3 (see "ASCII Tables" in the Table of Contents).

Figure 5 shows all Cr-rich spinel analyses obtained in this study, plotted by mapped unit number. In examining the data, it is apparent that an enormous range in Cr-rich spinel composition exists within single units, exhibited particularly by Mg/(Mg + Fe²⁺) (Mg#) at a given Cr/(Cr + Al + Fe³⁺) (Cr#). The origin of this variation can be understood by considering the examples given in Figure 3. In Figure 3A, a large decrease in the Mg# is found away from the spinel core, accompanied by little change in Cr# or Fe³⁺/(Cr + Al + Fe³⁺) (Fe³⁺#). This exchange of divalent cations without movement of the less mobile trivalent cations is common where Cr-rich spinel is rapidly re-equilibrating with a quickly evolving interstitial melt before, or more typically, after eruption, and it is equivalent to what happens under spinel metamorphic re-equilibration (Allan and Dick, 1996; Allan, 1992, 1994; Sack, 1980; Engi and Evans, 1980; Roeder et al., 1979). These exchanges may also occur during extensive alteration of the lava, as demonstrated by Allan (1992). In the example shown in Figure 3C, the margins of the analytical transect show more profound exchange, with the sharp increase in Fe³⁺ causing the increase in Fe³⁺#. Interestingly, the moderate increase in Cr# is caused by a sharp decrease in Al₂O₃ at the margins (from 27% to 10%), likely associated with both the increase in Fe³⁺ and the strong compositional couple between Al³⁺ and Mg²⁺ in the spinel structure (Sack and Ghiorso, 1991a; MgO decreases from 12% to 5%). The sharp increases in Fe³⁺# with small declines in Mg# shown in Figure 3C are associated with similar processes of exchange and chromian magnetite formation on Cr-rich spinel margins during extended lava cooling, where interstitial melt fO₂ may increase and Mg# decreases substantially.

As mentioned above, Cr-rich spinel inclusions in olivine and especially plagioclase can be protected from the effects of changing melt composition and alteration, avoiding the formation of magnetite jackets and extensive re-equilibration to lower Mg#. Nevertheless, exchange with melt, especially by the divalent cations, may occur by diffusion through olivine, as was documented by Scowen et al. (1991) for Cr-rich spinels within evolving magma in the Kilauea Iki lava lake. In the Hole 917A lavas, Cr-rich spinel inclusions completely enclosed within olivine may lack magnetite jackets, but many do show, nonetheless, evidence for re-equilibration by exhibiting decreasing Mg# decline toward their margins. Although plagioclase is usually quite effective at "armoring" Cr-rich spinels against re-equilibration, some Hole 338 spinel inclusions also show signs of re-equilibration, perhaps aided by cracks within the plagioclase.

As the primary intention of this paper is to identify primitive melts within the SRDS system and the melt compositions existing during eruption and not to study the effects of evolving interstitial melt after eruption or Cr-rich spinel magmatic and metamorphic alteration, Figure 6 shows a cleaned data set consisting only of Cr-rich spinel analyses from unzoned crystal cores. For spinels with profiles shown in Figure 3A and Figure 3C, the entire analytical transect was not considered, as it is not clear from the calculated Mg#s whether any relict core composition is preserved. For other spinel grains where strong zoning was observed, only the analyses of the unzoned interiors were plotted, where a significant plateau in Mg# was present. Examples of these plateaus are given in Figure 8, with the interiors of Transects E, G, and J representing marginal cases because of having a poorly-developed plateau. The data from the transect shown in Figure 3B was also included in Figure 6, as the interior of the Cr-rich spinel was unzoned. Comparison of this groundmass Cr-rich spinel with Cr-rich spinel inclusions from the same sample show it to have markedly lower Mg# (see Transects F and G in Fig. 8), either reflecting a mixed magma or complete re-equilibration of the spinel interior with an evolving interstitial melt after eruption. These two possibilities exist for all unzoned Cr-rich spinel cores that contain a comparatively low Mg# in samples where a broad range of spinel Mg# at a given Cr# exists.