Analytical techniques

X-ray fluorescence (XRF) analyses of whole-rock samples and mineral separates of plagioclase, clinopyroxene and orthopyroxene were carried out at the Institut für Mineralogie, Petrologie und Geochemie at the University of Freiburg on a Philips PW 1450/20 machine using natural standards; accuracy and detection limits are of the order of 0·1 wt % for major elements and 1-10 ppm (depending on the specific element) for minor elements. Sample preparation was performed at the same institute. For whole-rock measurements, about 1-3 kg of material, depending on the grain size, was crushed and milled in agate mills, and powder and Li borate fusion discs were prepared to measure trace and major elements. Mineral separates were prepared from 63-125 µm fractions of the whole-rock samples; separations were made with heavy liquids, magnetic separators and a wave table. The resulting separates were again milled, and powder and fusion discs were prepared. Powders of six whole-rock samples were analysed for trace elements including La, Ce, Nd, Sm, Eu, Gd, Tb, Dy, Er, Yb and Lu by inductively coupled plasma mass spectrometry (ICP-MS) at the CRPG laboratory, Nancy, on a commercial basis.

Major and trace elements

Anorthositic rocks

Whole-rock analyses of anorthositic rocks typically reflect a cumulate mixture of various proportions of plagioclase, oxides and Fe-Mg silicates. Typically, it is difficult to derive genetic significance from whole-rock data, which in most cases only show that these rocks are cumulates (e.g. Haskin & Salpas, 1992). For example, a plot of wt % FeO vs wt % Al₂O₃ shows a linear relationship of decreasing Al₂O₃ with increasing FeO (Fig. 1). It is reasonable to assume that FeO in the rock is contributed by post-cumulate pyroxenes and oxides whereas the Al₂O₃ resides mostly in the plagioclase. Thus the linear pattern in Fig. 1 merely reflects different abundances of pyroxene and plagioclase. Similar trends are seen in other major oxides, indicating that the major element chemistry is controlled for the most part by the relative abundance of the minerals in the rock, rather than by the composition of the melt.

Figure 1. Al₂O₃ vs FeO^tot plotted for anorthositic and related rocks from the FBC, Lofoten Islands. The linear dispersion of data clearly shows a cumulate trend and does not reflect a fractionation trend.

Gabbro and ferrodiorite

In addition to the coarse-grained rocks, there are fine-grained mafic rocks, which we assume to represent chilled melts (Table 1). Of these samples, GM 405, 448, 406, 500, 508, 1120 and 1154 are dyke rocks, whereas GM 451, 443 and 420 are from bodies tens of metres in size. Eighteen additional analyses of evolved tholeiitic to ferrodioritic compositions from Lofoten dyke rocks have been given by Misra & Griffin, (1972). The fine-grained samples listed here are interpreted to represent melt rather than cumulate compositions because of their fine grain size, the lack of Eu anomalies (where analysed, see Fig. 2 and Table 2) and the lack of correlations between chemistry and the compositions of possible accumulating phases.

Table 1. XRF whole-rock data for selected gabbroic and ferrodioritic rocks from the Lofoten Islands; for comparison (see text), also an analysis of an LAC gabbro from Mitchell et al. (1995) is listed

Table 2. REE concentrations (in ppm) in selected gabbroic and ferrodioritic rocks from the Lofoten Islands

Figure 2. REE patterns of ferrodioritic rocks from Lofoten (whole-rock chemistry is reported in Table 1). (Note the absence of significant Eu anomalies, which is discussed in the text.)

These fine-grained rocks are chemically distinct from the gabbroic anorthosites. The most important differences are that the anorthositic rocks have higher Al₂O₃ and Na₂O contents. Some of these fine-grained rocks (GM 448, 500, 420 and 443) have low X_Fe [= FeO^tot/(FeO^tot + MgO)] and have major element compositions that are similar to those of evolved tholeiites (e.g. Carmichael et al., 1974), whereas others (GM GM 405, GM 406, GM 451, GM 1120) have higher X_Fe and have compositional affinities to ferrodiorites from other Proterozoic anorthosite complexes (e.g. high FeO^tot, TiO₂, P₂O₅; Emslie, 1980; Duchesne et al., 1989; Mitchell et al., 1995, , 1996). In terms of their trace element chemistry, the latter group shows enrichment in Zr and Ba and depletion in Ni and Cr compared with the other fine-grained rocks and with the gabbroic anorthosites. GM 448, GM 1120 and GM 1154 were sampled from dykes within the Eidsfjord anorthosite. As this is of importance in the discussion below, the analysis of the gabbro sample BM-10 from the Laramie Anorthosite Complex in Wyoming (Mitchell et al., 1995) is listed for comparison in Table 1.

Theory

The distribution of elements between a melt and coexisting crystals is given by the partition coefficient D_i, which is defined as

The partition coefficient of an element may vary with temperature, pressure, composition of the melt and composition of the crystal, although changes in pressure and melt chemistry seem to exert a smaller influence than do changes in the other parameters. Based on experiments in the simplified system CaMgSi₂O₆-NaAlSi₃O₈-CaAl₂Si₂O₈, Blundy & Wood, (1994) proposed a model that allows the calculation of the concentrations of all equally charged cations in the melt if their concentrations in the crystal, the partition coefficient for one element, the elastic moduli and the site sizes in the mineral are known. This model is based on the assumption that the structure of the crystal in equilibrium with a melt governs the partitioning between the melt and the crystal, and that the composition or structure of the melt is of much less importance. Because their experiments can account for the main network-modifying cations, this model system is also a good approximation of Fe-bearing basaltic compositions (see, e.g. Blundy, 1997). The D values and mineral analyses that were used in the calculations are given in Tables 3 and 4, and all other parameters are given in Table 5. Ionic radii are from Whittaker & Muntus, (1970) and Blundy & Wood, (1994); site radii were either adopted from Blundy & Wood, (1994) or calculated as detailed below.

Table 3. Calculated concentration of some major, minor and trace elements in the melt coexisting with plagioclase from the Lofoten anorthosites and the K_D values used (see text for explanation of the calculation procedure)

Table 4. Calculated concentration of some major, minor and trace elements in the melt coexisting with pyroxenes from the Lofoten anorthosites and the K_D values used (see text for explanation of the calculation procedure)

Table 5. Parameters used for modelling the anorthositic parental melts

The use of the Blundy & Wood, (1994) model is limited to 1+, 2+ and some 3+ elements because the partitioning behaviour of some 3+ and the 4+ elements is more complex, probably because of additional electron interactions. For this reason, and to better evaluate the validity of the results, we performed additional calculations using the partition coefficients from the works of Beattie et al., (1991) and Beattie, (1993), who investigated partitioning of various major, minor and trace elements between orthopyroxene, clinopyroxene and basaltic melts. For comparison, we also calculated concentrations of K₂O, Rb, Sr and Ba using the partition coefficients of Simmons & Hanson, (1978) and of Morse, (1988), and of Al₂O₃, TiO₂ and FeO using the partition coefficients reported by Nolan & Morse, (1986) that they derived for anorthosites or rocks related to anorthosites (Kiglapait troctolite). In the following, we will describe in detail which model and which values we used for the calculations with respect to the various minerals.

Calculation procedure

Problematic elements

Plagioclase mineral separates consistently have higher Fe than microprobe analyses. This is probably caused by the presence of small (micrometre-sized) inhomogeneously distributed grains of Fe oxides in the plagioclase. For FeO content of plagioclase, therefore, the microprobe analyses were used in the calculations because they are believed to give more accurate results. The estimate of Fe in the melt from clinopyroxene using the data of Beattie, (1993) gives unreasonably high results. We discount these results because there is a strong dependence of D_Fe on the X_Fe of the clinopyroxene (Beattie, 1993), and our clinopyroxenes are more Fe rich than those used in Beattie's, (1993) experiments.

There are also discrepancies between estimations of K₂O in the melt made from plagioclase separates and those made from EMP analyses, especially in the case of the FBC. Based upon the composition of plagioclase separates, the calculated K₂O content of the FBC melt would be of the order of 2-4 wt %, whereas that of the Eidsfjord melt would be 1-2·5%. In contrast, calculations based upon microprobe analyses give more reasonable values of 0·9-1·4 wt% K₂O in both intrusions. We believe that these high K₂O values and related high concentrations of Ba and Rb are caused by the fact that the mineral separates included small amounts of late K-feldspar, which had formed from an interstitial liquid that was considerably enriched in K, Ba and Rb compared with the liquid from which the rest of the plagioclase and the mafic minerals crystallized. We therefore used the concentrations of K₂O in plagioclase as determined by the EMP analyses for the estimated liquid composition (Table 6). For the same reason, we do not report values for Ba and Rb in the calculated melt of the FBC.

Table 6. Reconstructed anorthositic parental melt compositions from the Lofoten anorthosites

Comparison of the results with Lofoten rocks and with MELTS calculations

For several reasons, we are confident that the calculated melts (Table 6) come close to the composition of the melt from which both the FBC and the Langoy anorthosites formed. First, the calculated liquids from the Lofoten anorthosites strongly resemble ferrodioritic rocks from Lofoten. Dykes of ferrodioritic and high-Al basaltic composition are found throughout the Lofoten Islands (Table 1) and additional data have been reported by Misra & Griffin, (1972). The calculated liquids of the FBC and the Eidsfjord anorthosite are both similar to GM 508 (although the FBC melt is more aluminous), a fine-grained basaltic dyke from the FBC. The Eidsfjord melt closely resembles the ferrodioritic rock GM 1154 from the margin of the Eidsfjord anorthosite (Table 1) in that they have both similar contents of many major and minor elements and similar Na/Ca and Al/Si ratios. The calculated melts are also similar to some of the dykes reported by Misra & Griffin, (1972) that are net-veined by anatectic material from the country rocks. This material has an apparent age (1795 ± 20 Ma, Rb-Sr whole rock, Misra & Griffin, 1972) similar to that of the mangerites and anorthosites. Pending more precise U-Pb geochronological work and because of the similarity between our calculated melts and some of these dykes, we propose that these dykes may represent basic melts coeval with the anorthosites and mangerites and that observed differentiation trends in these rocks may reflect the differentiation of the basic magma. A plot of major element oxides against X_Fe (= FeO^tot/(FeO^tot + MgO) in the fine-grained rocks from Lofoten is shown in Fig. 3. X_Fe is chosen as a monitor of differentiation in this figure because SiO₂ does not change much during crystallization of the ferrodiorites. There is a continuous trend of increasing Fe₂O₃, P₂O₅, K₂O and TiO₂, and of decreasing Al₂O₃, CaO and MgO with increasing X_Fe. Also shown in Fig. 3 are the calculated melts for the FBC and Langoy anorthosites. They clearly lie within the trend defined by the fine-grained rocks and closely resemble some individual dyke rock samples.

Figure 3. (a) and (b) Comparison of basic dykes, fine-grained gabbroic rocks and ferrodiorite GM 451 from the Lofoten Islands with the reconstructed melts for the Flakstadøy Basic Complex (FBC) and the Eidsfjord anorthosite. Data used in this plot are taken from Misra & Griffin, (1972) or are given in Tables 1 and 2. Crossed symbols indicate calculated melts.

Another reason why we have confidence in the veracity of our results is that we used the MELTS program of Ghiorso et al., (1994) to calculate the composition of the phases that would be in equilibrium with our inferred melts at the temperatures and pressures that we estimated for the formation of these anorthosites (Markl et al., 1998). The calculated phase compositions of plagioclase, olivine and clinopyroxene are very similar to those observed in the FBC and the Eidsfjord anorthosite (Table 7). Small discrepancies in the oxygen fugacities result from the fact that we have poor control of the ferric iron content of the pyroxenes and hence poor control of the ferric iron content of the calculated melts. This probably explains the predicted existence of olivine in the MELTS calculations at 4 kbar for the Eidsfjord melt, even though olivine seems to be lacking in these rocks.

Table 7. Comparison between mineral compositions that are observed in the Lofoten anorthosites and those that are calculated from the reconstructed parental melts using the MELTS program of Ghiorso et al. (1994)

DISCUSSION

Validity of the calculations

Choice of partition coefficients

Morse, (1992) cast doubts on the validity of the Blundy & Wood, (1991) formulation for augite-poor, anorthositic liquids. He argued that in the Kiglapait intrusion the formulation for Sr of Blundy & Wood, (1991) led to strong discrepancies between the calculated and his reconstructed liquid evolution trend. Blundy & Wood, (1992) argued in a reply that Morse's dataset was dependent on strictly closed system behaviour and pointed out that his results were in contrast to the results of all other studies they used for the derivation of their model. Furthermore, recent experimental work by Blundy, (1997) on the Kiglapait intrusion suggests that the calculated liquid composition of Nolan & Morse, (1986) and hence the partition coefficients calculated therefrom may not be correct. Hence, the discrepancy between the results calculated with the partition coefficients of Morse, (1988) and Nolan & Morse, (1986) and those calculated from Blundy & Wood, (1994) (Tables 3 and 4) are likely to be caused by errors in Morse's partition coefficients. Blundy, (1997) furthermore showed that the partition coefficients calculated after Blundy & Wood, (1994) are suitable for rocks and liquids of the anorthosite kindred, and he suggested that inversion using appropriate D values is a better means of calculating parental melt compositions rather than summation over exposed cumulates assuming closed system fractionation behaviour.

Validity and accuracy of the results

Because whole-rock analyses of cumulate rocks are sensitive more to variations in mineral abundance than to the composition of the melt from which they were derived, we have chosen to calculate the composition of the coexisting melt from the composition of the constituent minerals. This technique, though, is prone to some difficulties, which arise from the fact that anorthosites as cumulate rocks have had a long crystallization history and the melt from which the minerals crystallized certainly changed composition as crystallization proceeded. Calculations using cumulus and post-cumulus minerals will therefore give an average of the liquid composition that was in equilibrium with the anorthosites during its evolution. However, there are several arguments that support the validity of our results:

   (1) the calculated melts are similar to gabbroic to ferrodioritic rocks observed in the Lofoten area;

   (2) using the MELTS program and our calculated melt composition we can closely match the composition of the phases in the rocks;

   (3) the calculated totals come close to 100%, even though there is no reason why they should (without normalization!);

   (4) where element concentrations in the liquid are calculated from various minerals and from various sets of partition coefficients (e.g. Ca from plagioclase and from clinopyroxene, or Sr from plagioclase and clinopyroxene), the results show only low variance (Tables 3 and 4).

Based on the comparison between different samples and different sets of partition coefficients, we estimate that the calculated major element concentrations (SiO₂, CaO, FeO, Al₂O₃) are precise to about 1-2 wt %. The Fe content may be subject to larger uncertainties because of the effects of Fe³⁺/Fe²⁺ uncertainties. MgO, K₂O and Na₂O are probably precise to about 0·5-1 wt %. Sr gives a precision on the order of 30 ppm or 8-10 relative % when results from plagioclase and from clinopyroxene calculations are compared, and this may be representative for the trace elements. The calculated CIPW norm of these liquids is slightly Ne-normative when low ferric/ferrous ratios are assumed. At higher ferric/ferrous ratios and if a P₂O₅ content of 0·5-1 wt % in the liquid is assumed, the calculated melts are only weakly Ne-normative or not at all, and hence resemble the observed ferrodiorite compositions from Lofoten which areHy- or slightly Ne-normative.

The reconstructed melts from the FBC and the Eidsfjord anorthosite (Table 6) have similar SiO₂, TiO₂ and K₂O contents, and both are enriched in FeO^tot and Al₂O₃ compared with tholeiitic basalts. The calculated Eidsfjord melt is much more evolved than the FBC melt as suggested by all common compositional indicators (higher X_Fe, Na₂O, K₂O, lower CaO, lower An in plagioclase; see also Markl et al., 1998). Hence, we infer that the Eidsfjord anorthosite equilibrated with a more evolved melt than did the FBC.

The magmatic evolution from basalts to ferrodiorites on the Lofoten Islands as recorded by olivine and feldspar

As noted above, the fine-grained rocks from the Lofoten Islands are continuous in composition from high-Al gabbro to ferrodiorite. There is evidence for melts with even higher MgO contents as indicated by the composition of magnesian olivines in the Moskenesøy cumulate (Fig. 4). The presence of an early Mg-rich basaltic magma is hence recorded by the mafic cumulates (Markl et al., 1998), whereas the more differentiated melts are recorded by the gabbros, anorthosites and ferrodiorites. The olivine composition in the Moskenesøy cumulate is Fo_82-83. Progressive Fe enrichment is recorded by the Ramberg gabbro on Flakstadøy (olivine with Fo₇₇), the anorthosites GM 263 and GM 265 with olivine of Fo₇₂ and Fo₆₈, respectively, and the anorthosite GM 139 with olivine of Fo₆₆. In the Eidsfjord anorthosite, olivine is absent, but the calculated equilibrium composition from QUI1F (Andersen et al., 1993) would have been Fo₅₈ (at lowera_SiO2 than actually observed). The apparent general fractionation trend reflected by the whole-rock (Fig. 3) and olivine (Fig. 4) compositions is also followed by the plagioclase compositions. Plagioclase evolved from An₆₉ in the Ramberg gabbro through normative An₆₃ and An₅₆ in the gabbros near Bjørnhaugen on Austvågøy (GM 420) and on Gimsøy (GM 440) to the FBC plagioclase (An_57-47) and the Eidsfjord anorthosite plagioclase (An_48-44). The ferrodiorite GM 451 has a normative ternary feldspar of An₁₇Or₂₃ (Markl et al., 1998).

Figure 4. The composition of olivine in various rock types from the Lofoten Islands. For rocks lacking olivine (Eidsfjord anorthosite, Eidsfjord ferrodiorite) the theoretical olivine in equilibrium was calculated using QUIlF (see text for explanation).

Comparison of the calculated melt compositions with rocks from other Proterozoic anorthosite complexes

Figure 5 plots the compositions of fine-grained rocks from Lofoten vs X_Fe together with a field for 50 analyses of gabbroic and ferrodioritic rocks from Harp Lake, Labrador (Emslie, 1980) and with 63 analyses from the Laramie Anorthosite Complex in Wyoming (Mitchell et al., 1995, , 1996). The fine-grained rocks from Lofoten have compositions very similar to those of the gabbros and ferrodiorites from both complexes (Fig. 5). The rocks from all three complexes clearly define similar evolution trends from high-Al gabbros to low-Al gabbros and to ferrodiorites in the majority of major and trace elements. The Lofoten rocks show the bend to high P₂O₅, K₂O and Zr at lower X_Fe than the Laramie and Harp Lake rocks (Fig. 5). These discrepancies, though, do not negate the similar evolutionary trend from basaltic to ferrodioritic compositions in both areas, as they are most probably due to contamination of the evolving ferrodioritic melt with crustally derived melts at various stages of their evolution. Similarly (but not shown in Fig. 5), the Lofoten rocks closely resemble high-Al-Fe gabbros from the Adirondacks in New York (Olson & Morse, 1990), ferrodioritic rocks from the Nain Complex in Labrador (Wiebe, 1990) and ferrodiorites from Rogaland, Norway (Duchesne et al., 1989). The Lofoten fine-grained rocks and the calculated melts evidently have equivalents in many large Proterozoic anorthosite complexes. Emslie, (1980), Olson & Morse, (1990), Wiebe, (1990) and Mitchell et al., (1996) have interpretedthese rocks as residual liquids of anorthosites, whereas Duchesne et al., (1989) have explicitly ruled out any comagmatic relationship.

Figure 5. Element oxide vs FeO^tot/(FeO^tot + MgO) plots of high-Al gabbros and ferrodiorites from the Laramie Anorthosite Complex (Mitchell et al., 1995, , 1996, filled symbols), of high-Al gabbros, low-Al gabbros and ferrodiorites from the Harp Lake Complex in Labrador (Emslie, 1980; hatched field), and of high-Al basalts and ferrodiorites from the Lofoten Islands [this study and 18 samples from Misra & Griffin, (1972), open symbols]. (See text for discussion.)

Comparison of liquid evolution trends

Melt evolution as recorded by the fine-grained dyke rocks in the Lofoten Islands and the calculated melts is remarkably similar to evolution trends of tholeiitic andhigh-Al troctolitic liquids. Figure 6a compares the evolution of the liquids from Lofoten and the Laramie Anorthosite Complex with the trends reported from the Kiglapait intrusion in Labrador (Morse, 1981) and from the Skaergaard intrusion (Wager & Brown, 1967) and the Basistoppen sill in Greenland (Naslund, 1989). Figure 6b compares the Lofoten trend with the Skaergaard trend as calculated by Hunter & Sparks, (1987) and with thebroad trend spanned by the Rogaland monzonorites of Duchesne, (1990). With the exception of the Rogaland monzonorites, which were interpreted by Duchesne, (1990) to be lower-crustal anatectic melts, all trends in Fig. 6a and b show initial, slightly variable Fe enrichment at nearly constant SiO₂ followed by SiO₂ enrichment and Fe depletion. The extensions of the Lofoten Islands and the Laramie Complex trends to higher SiO₂ values cannot be reliably retrieved and are therefore shown dashed in Fig. 6a and b. Hunter & Sparks, (1987) and Toplis & Carroll, (1996) pointed out that the amount of Fe enrichment depends on the composition and on the f_O2 of the parental magma. This argument further supports that the Lofoten rocks and calculated melts record an evolution trend typical of tholeiites.

Figure 6. (a) Fractionation trends in terms of FeO^tot and SiO₂ of the Lofoten and Laramie Anorthosite Complex high-Al gabbros (this study; Misra & Griffin, 1972; Mitchell et al., 1995, , 1996) and, for comparison, of the Skaergaard intrusion (Wager & Brown, 1967), the Kiglapait intrusion (Morse, 1981) and the Basistoppen sill (Naslund, 1989). (b) Fractionation trend of the Lofoten rocks as in (a), compared with the trends in the Rogaland monzodiorites (Duchesne, 1990) and with the Skaergaard evolution as calculated by Hunter & Sparks, (1987).

Comparison with the calculations of Naslund, (1989) and Morse, (1981) suggests that the fraction of melt crystallized until the point where the Lofoten high-Al basalts or ferrodiorites attain their highest FeO^tot contents lies between 50 and 75 wt %. As the ferrodiorites are the liquids from which the anorthosites formed, there is a huge amount of these liquids to be crystallized after the formation of the anorthosites. During further fractionation and/or assimilation processes, these ferrodiorites may, for example, be used to produce the large amounts of mangerites and charnockites commonly observed to accompany the anorthosites and ferrodiorites in Proterozoic anorthosite complexes (e.g. Fuhrman et al., 1988; Emslie & Hunt, 1990; Emslie, 1991; Emslie et al., 1994).

Two types of high-Al basalts

The parental melts of Proterozoic anorthosite complexes are high-Al basalts that fractionated to ferrodiorites (e.g. Emslie, 1980; Olson & Morse, 1990; Mitchell et al., 1995, and references therein). It is important to note, however, that high-Al basalts associated with anorthosite complexes are chemically distinct from those that are characteristic of subduction environments (e.g. Myers et al., 1986). The subduction-related high-Al basalts are higher in Al₂O₃, have relatively high X_Fe (e.g. Yoder & Tilley, 1962; Singer et al., 1992) and much lower TiO₂ and Ni compared with the least evolved high-Al gabbros from Proterozoic anorthosite complexes (Emslie, 1980; Mitchell et al., 1995). In addition, they do not exhibit the strong Fe enrichment that is characteristic of the anorthosite-related high-Al gabbros. Many of the high-Al gabbros from Proterozoic anorthosites show strong positive Eu anomalies, which are not characteristic of high-Al basalts from subduction environments (e.g. Myers et al., 1986; Myers, 1988). Because of these geochemical differences, it is evident that the high-Al gabbros related to Proterozoic anorthosites have an origin distinct from that of the subduction-related high-Al basalts.

The origin of the high-Al basalts associated with anorthosite complexes can be evaluated using the analyses of high-Al gabbros from the Laramie Complex, which have a wide range of compositions (Mitchell et al., 1995). One way to show this compositional range is on a plot of MgO (which monitors the crystallization of ferromagnesian silicates) vs Al₂O₃ (which monitors the crystallization of plagioclase) (Fig. 7). There are two apparent trends on this diagram. The first shows variable Al₂O₃ at slightly decreasing, but high values of MgO, whereas the second starts at high Al₂O₃ and MgO values and shows decreasing MgO with decreasing Al₂O₃. We interpret these trends to indicate that two major processes were involved in the magmatic evolution of these rocks. Trend (1) shows a transition from a magma that is of normal tholeiite composition to high-Al basalt, and Trend (2) is the differentiation trend from high-Al basalt to ferrodiorite (Figs 3 and 5).

Figure 7. MgO vs Al₂O₃ in fine-grained gabbros of the Laramie Anorthosite Complex (Mitchell et al., 1995). This diagram shows the evolution trend from tholeiitic basalts to high-Al basalts by incorporation of earlier formed plagioclase and contemporaneous fractionation of clino- and/or orthopyroxene (1), and the subsequent fractionation trend of the high-Al basalt to ferrogabbros and ferrodiorites (2). (See text for discussion.)

The existence of the first trend implies that there is a mechanism by which tholeiitic basalts may evolve to the high-Al basalts of anorthosite complexes. As suggested by Wiebe, (1990), we propose that resorption of plagioclase and contemporaneous fractionation of ortho- and/or clinopyroxene (which may be tentatively regarded as the megacrysts observed in many anorthosite complexes) during rise of the magma or during influx of new, hot magma is responsible for the formation of this type of high-Al basalts. We observe two justifications for this process: first, resorption of plagioclase combined with fractionation of, for example, Cr- or Ti-spinel will increase the melt in Al₂O₃ whereas MgO can stay relatively unchanged. Second, resorption of plagioclase also increases Eu in the melt, producing the positive Eu anomalies observed in many of the Proterozoic high-Al gabbros. The strongly positive Eu anomalies of the high-Al gabbros may also explain why ferrodioritic rocks often have low Eu anomalies or even lack them entirely (Mitchell et al., 1996; Scoates & Frost, 1996). It has long been argued that these rocks cannot be residual after extensive crystallization of plagioclase because they lack a negative Eu anomaly. Apart from the process of plagioclase resorption, the derivation of melts with strongly positive Eu anomalies from the mantle is difficult to reconcile with the known mantle mineralogy and geochemistry.

The process of plagioclase remelting may produce a variety of high-Al basaltic magmas. Depending on the amount and composition of the plagioclase and the fractionating pyroxenes and on the amount of fractionation in the tholeiitic basalt before resorption of plagioclase, the high-Al basalts can show various X_Fe, CaO or SiO₂ values. Thus, then the samples with the lowest Al₂O₃ and highest MgO should represent the least evolved magma. Such rocks are found in the Laramie Anorthosite Complex (e.g. sample BM-10), where they have compositions typical of tholeiites. Sample BM-10 has no Eu anomaly, high Ni and Cr contents, ~11 wt % MgO and ~14 wt % Al₂O₃ (Table 1). Extrusive rocks with nearly identical composition are found in Hawaii (Macdonald & Katsura, 1964) and on the Hebrides (Carmichael et al., 1974), the former representing an oceanic, the latter a continental tholeiitic province.

Plagioclase compositions and density contrasts

Proterozoic anorthosites are characterized by plagioclase of the composition An₄₅-An₅₅ and rarely contain plagioclase above An₆₀. This observation may be explained by the experimental constraints of Fram & Longhi, (1992) and Longhi et al., (1993), according to which more calcic plagioclase is not a liquidus phase of anorthositic parental melts at high pressures. However, we propose an additional process that may control plagioclase composition in Proterozoic anorthosites. At low pressures the density contrast between more calcic plagioclase and basaltic melt is too small to permit efficient accumulation at the top of the deep-seated magma chamber. Figure 8 shows results of calculations with the MELTS program of Ghiorso et al., (1994), where the densities of the least fractionated high-Al gabbros, of tholeiites, of the calculated melts, of a ferrodiorite and of various plagioclase compositions were calculated at 10 and 4 kbar. Figure 8 shows that the density contrast during early stages of fractional crystallization is around 0·1-0·15 g/cm³, until the plagioclase reaches a composition of about An₆₀ and, especially, until the melt reaches high values of Fe and of X_Fe. Thus, plagioclase of An_>60-65 probably is not able to float in primitive tholeiitic magmas.

Figure 8. (a) The density of tholeiitic liquids like BM-10 and primitive high-Al gabbros, of the reconstructed anorthositic melts and of ferrodioritic liquids (calculated using MELTS; Ghiorso et al., 1994) plotted against the FeO^tot/(FeO^tot + MgO) ratio of the liquid as a monitor of differentiation. Also shown is the density of plagioclase that can be in equilibrium with these melts at pressures between 4 kbar (An₇₅) and 10 kbar (An₆₅). (b) The density contrast between plagioclase and the coexisting liquid. It should be noted that the plagioclase compositions found in anorthosites record a density contrast of 0·15 or higher, indicating that at lower contrasts or high X_An^plag the plagioclase is not able to accumulate at the top of a deep-seated magma chamber.

The original magma to start with: a tholeiitic basalt

The arguments above imply that the parental melts to anorthosite complexes are not of extraordinary composition; they are tholeiitic. The controlling parameter in the formation of Proterozoic anorthosites is probably not the melt composition, but the specific crystallization history (resorption of plagioclase in melts), and this in turn may be strongly influenced by the tectonic setting. The most obvious way for resorption to happen is as a result of movement of a high-temperature tholeiitic magma through a deep-seated magma chamber containing intermediate plagioclase. Because intermediate plagioclase will crystallize from relatively unfractionated tholeiitic melt only at high pressures (Fram & Longhi, 1992; Longhi et al., 1993), this leads directly to the two-stage model of anorthosite formation (e.g. Emslie, 1978; Duchesne, 1984; Longhi & Ashwal, 1985; Wiebe, 1992; Ashwal, 1993; Emslie & Hegner, 1993; Longhi et al., 1993; Emslie et al., 1994; Scoates & Frost, 1996). Plagioclase stable at high pressures could be remelted either because of a pressure decrease (Longhi et al., 1993) related to convective lithospheric thinning (Corrigan & Hanmer, 1997) or because of injection of new hot magma into the magma chamber in a plume-related environment (e.g. Hoffman, 1989). Under high-pressure conditions at the crust-mantle boundary, the stable crystallizing plagioclase would be of intermediate composition and would float in the crystallizing magma; continuous fractionation of these high-Al melts can eventually lead to anorthositic cumulates buoyant enough to rise to high crustal levels as crystal-rich mushes (Fram & Longhi, 1992; Wiebe, 1992; Ashwal, 1993; Longhi et al., 1993; Emslie et al., 1994; Lafrance et al., 1996) whereas the melt entrained in this crystal mush continues to fractionate. During ascent and emplacement at mid-crustal levels, the successively more evolved liquids may be removed at various stages of fractionation because of filter-pressing. Our results strongly support that these liquids are the ferrodiorites observed in most anorthosite complexes.