The order of intrusion of the major rock units (Figure 1) is:

1. Gabbro; 2. Main oversaturated ring, i.e., inner quartz syenite, outer quartz syenite, microgranite ring dykes; 3. Porphyritic quartz syenite ring dyke and plug; 4. Nepheline monzonite, nepheline-bearing syenite, foyaite and microfoyaite cone sheets; 5. Leucomalignite core. Units 4 and 5 are composed mainly of nepheline-bearing, undersaturated rock types and IUGS names are used here except for samples showing trachytic textures with aligned alkali feldspar phenocrysts; for such rocks the term foyaite is retained here.

1. The equigranular gabbro consists essentially of plagioclase and pyroxene with lesser amounts of olivine and biotite (Figure 3(a)). Many samples contain small amounts of alkali feldspar (<5 vol. %) although these varieties and the “normal” gabbros do not show any recognisable geographic distribution.

2. Gabbros are intruded by the main oversaturated ring. The contacts between the latter and the basement are not exposed. The outer quartz syenites have variable quartz contents and the inner quartz syenites, in the northern part of the ring, are cut by offshoots from the outer quartz syenite ring.

3. The porphyritic quartz syenite (Figure 3(h)) occurs as two separate units: a discontinuous ring, and a single body in the undersaturated units.

4. Nepheline monzonite and nepheline syenite cone sheets intrude the oversaturated rocks and the gabbros, their centres of intrusion being to the N.W. of the main oversaturated ring centre. The contacts between the gabbro and the outer margins of the undersaturated intrusive units are often complex. The nepheline-bearing syenite (mapped as “white syenite”) consists of two crescents cut by the complex nepheline monzonite and foyaite “ring”. Typical nepheline-bearing syenites are equigranular and consist of alkali feldspar and amphibole with variable amounts of diopsidic pyroxene, plagioclase and nepheline (<5 vol. %) (Figure 3(c)). Some varieties are richer in nepheline and amphibole (Figure 3(d)).

The contact between the monzonitic rocks and the foyaite sheets is difficult to locate in the field. Foyaites studied are dominantly porphyritic types with alkali feldspar phenocrysts in a finer-grained groundmass (mapped as “grey syenite”). The foyaites consist of alkali feldspar, amphibole, nepheline (>5 vol. %), sometimes with pseudoleucites (~5 cm diameter). Feldspar phenocrysts and microphenocrysts show alignment, giving the rocks a typically foyaitic texture (Figure 3(e) and Figure 3(g)).

Figure 3.Representative Marangudzi hand specimens. (a) Orthocumulate gabbro A12, larger white and medium grey grains are cumulus calcic plagioclase and augite, respectively, small bright white spots are reflections from 2 - 3 mm biotite cleavage flakes; (b) Nepheline monzonite T308B, large white grains are mainly calcic plagioclase with some alkali feldspar, large dark grey/brown grains are Ti-rich biotites and hastingsites, medium grey augite, pale grey groundmass includes nepheline; (c) Nepheline-bearing syenite (“white syenite”) A31, large white grains are plagioclase, pale grey alkali feldspar, dark grey/green grains are amphibole; (d) “White syenite” G226, plagioclase-free, nepheline-rich “pegmatitic” sample, large dark grey/green grains hastingsite, white grains alkali feldspar, pale grey nepheline; (e) Foyaite (“grey syenite”) G183, large white grains aligned alkali feldspar (foyaitic texture), grey finer grained groundmass amphibole, alkali feldspar and nepheline; (f) Leucomalignite (“white syenite”) T330, dark grey elongate amphiboles, medium grey calcic pyroxene, white grains alkali feldspar, pale grey (greasy lustre) nepheline; (g) Foyaite A20, large elongate phenocrystic alkali feldspars in finer grained matrix of amphibole, alkali feldspar and feldspathoid; note similarity to G183 (Figure 3(e)); (h) Porphyritic quartz syenite A21, large phenocrysts of alkali feldspar with some xenocrystic plagioclase in finer-grained groundmass of amphibole, alkali feldspar and minor quartz, see white on black (cm + mm) scale bar on each image.

A series of more basic undersaturated rocks (mainly nepheline monzonites) occurs at the eastern margin between the foyaite and nepheline-bearing syenite “rings” (Figure 1). The nepheline monzonite unit varies from basic rocks rich in calcic plagioclase and pyroxene (Figure 3(b)) to those richer in alkali feldspar and amphibole; most are rich in biotite. This unit seems to be contemporaneous with the foyaite ring. The microfoyaites (“grey syenites”) intrude the nepheline-bearing syenite and main oversaturated ring. Overall, the age relations between the silica-undersaturated intrusive units are not clear and there is considerable overlap in the modal mineralogy between the different units. However, the textural relations and geochemical features, for each unit identified, are generally characteristic; indeed, different varieties could have developed in-situ within a single ring unit (see later).

5. The leucomalignite unit (“white syenite”) forms the core of the undersaturated rings. The rocks are heterogeneous nepheline-rich rocks including a coarse-grained pseudoleucite (up to 10 cm diameter)-bearing rock, and a variety free of pseudoleucites (Figure 3(f)). The leucomalignite unit appears to be the last intruded of the major units.

Magnesian olivine only occurs in the gabbro sheet and fayalite (or its pseudomorph) occurs in a few quartz syenites. Compositional relations (Table 3 and Supplementary file S.5) show very little Fe/Mg core/rim variation. CaO and NiO are uniformly low and show little change with changing Fo. For olivines from the different rocks, Fo ranges from 71.4 for the least evolved gabbro mesocumulates to 45.8 mol.% for the rim in the most evolved orthocumulate (T168). The most primitive average composition for the cumulus olivines is ~Fo₇₁ and three gabbros (100, 100A, G265) define the most primitive olivine composition (Figure 5). The olivines show only limited effects of magmatic fractionation with the most Fe-rich compositions shown by core/rim values for the most evolved orthocumulate T168 (points labelled c and r; Figure 5).

Table 3.Olivine electron microprobe (11 samples) and bulk chemical analysis (2 samples) from Marangudzi gabbros.

Note: n.a., not analysed; n.d., not detected; detection limit CaO ~0.04 wt.%; NiO ~0.08 wt%; standard deviations for last places in brackets.

Figure 5. Compositions of mafic minerals (olivine, clinopyroxene, orthopyroxene, amphibole and biotite) are plotted in the Ca (+ Sr, Ba)-Mg-(Fe + Mn) variation diagram as atomic %. The “Ca” component for the biotites mainly reflects their elevated Ba concentrations. Representative “tie-lines” for two-pyroxene pairs are shown for mesocumulate gabbro 100A and orthogabbro A12 and for the coexisting clinopyroxene-biotite-hastingsitic amphibole trio for nepheline monzonite T303. Note that orthopyroxene compositions are more Fe-rich than coexisting clinopyroxene and also that coexisting amphiboles are always more Fe-enriched than the coexisting biotites. Symbol labelled GNMPF is for the series Gabbro-Nepheline Monzonite-Nepheline-bearing syenite-Foyaite (see GNM), Bn = basanite T178. To preserve clarity, biotite from all the Marangudzi rock types has the same symbol.

Table 4gives average clinopyroxene compositions for each of the major pyroxene-bearing units. The tholeiitic affinities of the main olivine-gabbro are reflected in the coexisting clino- and ortho-pyroxene (100A, G266 and A12). Clinopyroxene is comparatively homogeneous from core to rim for the gabbros and quartz syenites but is more variable in the silica-undersaturated units (deviation for wt.% MgO generally within 1% - 4% relative for the gabbro pyroxenes, 2% - 5% for silica-oversaturated “ring units”, and much higher values of 6% - 19% for the silica-undersaturated rings) (Supplementary file S.5).

Table 4. Representative Marangudzi pyroxene compositions for the main intrusive units.

Note: Based on Fe³⁺ calculated. I.E. Iron enrichment = Fe_tot/(Mg + Fe_tot + Mn); mg# = Mg/(Mg + Fe_tot + Mn); Fe³⁺ fixed to use all Na (acmite), then combined with Ca to make Ti tschermaks calcic ferripyroxene; all Ti combined with 2Al to make Tschemaks Al calcic px; remaining Ca makes Wo, all Mg makes En and all Fe²⁺ makes ferrosilite (formulae as in above table). Wo-En-Fs defines the pyroxene chemical system. *Oxide proportions calculated stoichiometrically.

Gabbro meso- and ortho-cumulate clinopyroxenes show lower Ca contents (~0.86 apfu) than those for the nepheline monzonite, nepheline-bearing syenite, foyaite and leucomalignite intrusive units (0.91 - 0.93 apfu) (Figure 6); the higher Ca relationship reflects the alkaline affinities of the latter units. Small clinopyroxene “fragmented” inclusions in nepheline-bearing syenite amphibole T301 are particularly Mg-rich and are locally altered to Mg-rich biotite, whereas larger separate pale green clinopyroxenes are less Mg-rich. T301 clinopyroxene has an average composition with Fe-enrichment factor lower than that for some of the nepheline monzonite biotites. Thus, the clinopyroxene in individual samples for each of the nepheline monzonite to nepheline-bearing syenite to foyaite sample series shows a scattered trend around Ca = 0.92 apfu, but average compositions for each intrusive unit (large symbols in Figure 6(a)) show a slightly decreasing Ca trend with evolution (ranging 0.3 - 0.7). The last intruded leucomalignite samples (Figure 6) have clinopyroxene compositions with similar higher Ca with Fe-enrichment factor ranging 0.4 - 0.65. In addition, the average clinopyroxene in the interior quartz syenite has a much lower Ca value than those for the porphyritic and outer quartz syenites, reflecting the presence of a coexisting orthopyroxene phase in the interior quartz syenite. Individual clinopyroxenes from the gabbros to nepheline-bearing rocks show scattered increasing Na and Mn (Figure 6(b) and Figure 6(d)); the averages for each of these units define clear evolutionary trends. Ti (Figure 6(c)) and Al (Suppl. file S.5) show similar scattered horizontal trends from gabbros to “foyaitic” samples with increasing Fe-enrichment factor, with the leucomalignites plotting above that trend. The average compositions for each confirm the clear horizontal trends for the gabbro to foyaitic rock series, with the average leucomalignite displaced to a slightly higher Ti (Figure 6(c)). Figure 6 also shows that each of the average quartz syenite compositions shows distinct positions for Na, Ti and Mn, with the average outer quartz syenite having the most evolved I.E. value.

Figure 6.Clinopyroxene atomic proportions (apfu) for Ca (6a), Na (6b), Ti (6c) and Mn (6d) versus the iron variation index I.E. (Fe + Mn)/(Mg + Fe + Mn) atomic). Individual rock samples are shown with small symbols and the average value for separate intrusive units is shown with a large symbol having the same colour as for the individual parent rock samples. See text for explanation.

The calculated molecular proportions of the clinopyroxenes have very small proportions of the CaTi- and CaFe³⁺-end members (<1 mole %), slightly larger CaAl (AlSiO₆) (1 - 4 mol.%), and low acmite (NaFe³⁺Si₂O₆ 1 - 2 mol.%), reflecting their miaskitic nature ([71]). Clinopyroxenes from the gabbro classify as “augites” [also see Table 4 in [19]] and have a mean mg# value of 78, which represents the mean composition of the cumulus clinopyroxenes. Clinopyroxenes from other units have higher Ca contents and are diopside-hedenbergite. The main compositional variations for the pyroxenes from each of the units are shown in Figure 5. Gabbro clinopyroxenes cluster with orthopyroxenes for three gabbros next to the En-Fs join. In Figure 5, lines of constant atomic Fe/(Fe + Mg) fan out from the Ca apex but for sample 100A the clinopyroxene has a slightly higher mg# than the coexisting orthopyroxene and the tie-line shown in Figure 5 diverges slightly from a line of constant mg#. The coexisting pyroxenes in gabbros G266 and A12 show a similar relationship, with mg# values for clinopyroxene being 5% - 10% more magnesian than for the coexisting orthopyroxene (Suppl. file S.5; tie-line for G266 not shown in Figure 5).

Compositions of the clinopyroxenes in the rocks forming the ring-type and plug intrusive units are also shown in Figure 5. The nepheline monzonite ring unit clinopyroxenes continue the trend of Fe enrichment (mg# 68-56) and have CaAl- and CaFe³⁺ end-member values similar to those for the orthocumulate clinopyroxene (Table 4 and Suppl. file S.5). Although some clinopyroxenes plot above the diopside-hedenbergite join, the recalculated Wo, En and Fs end-members have Wo < 50 (Table 4) and plot just inside the diopside-hedenbergite field ([19]). Note that the maximum mg# for clinopyroxene single spot analyses in nepheline monzonites average 63.6 (Table 4), which is significantly less than the mean value of mg#= 83 for the three most primitive cumulus clinopyroxenes in the gabbros (see above). The NaFe³⁺-component is significantly higher than for the gabbro clinopyroxenes, which explains the green colour of the more-evolved, alkali-rich character of the nepheline monzonite pyroxenes. The nepheline-bearing syenites containing plagioclase have pale green clinopyroxenes that continue these relationships (mg# 58-46), as do the foyaite unit pyroxenes (mg# 40-30), which are slightly richer in acmite; the nepheline-bearing syenite clinopyroxenes are diopsides and those in the foyaites are hedenbergites. The last-intruded nepheline syenite unit (leucomalignites) has more-primitive, magnesian pale-green clinopyroxenes in some rocks (overall range mg# 57-35; diopside to hedenbergite) with slightly elevated CaAl- and NaFe³⁺-clinopyroxene components (Suppl. file S.5).

Pyroxenes in the quartz syenite units are much less abundant than the associated amphiboles and some inner quartz syenites contain coexisting clino- and ortho-pyroxenes with the former being much more magnesian than the coexisting orthopyroxene, as shown by the tie-line orientation (e.g., A6, Table 4; Figure 5). The clinopyroxenes in this unit are less aluminous than those in the gabbros and the silica-undersaturated ring units, and have very variable Mg and Fe contents with mg# varying from 50 (porphyritic quartz syenite A21, salite) to 41.3 (inner quartz syenite A6, ferroaugite) to 3.6 (inner quartz syenite T17, ferrohedenbergite). The very Fe-rich clinopyroxene in T17 is accompanied by a Fe-rich silicate (~Fe_1.86Si_1.1O₄) that might be a fayalite pseudomorph.

Calculated temperatures and pressures for clino-orthopyroxene pairs ([80]) from three gabbros and one inner quartz syenite are presented in Table 4 and Suppl. file S.5. The gabbros fall at the lower end of the acceptable equilibrium test and the quartz syenite temperature is lower than those provided by the gabbro samples. However, the temperatures are plausible (gabbro average 925˚C ± 30˚C; quartz syenite 770˚C); see below for further discussion (Section 6).

Table 5. Average compositions of biotites from Marangudzi main intrusive units.

Red-brown biotite is the last-formed primary magmatic mineral in the Marangudzi gabbros and less-evolved silica-undersaturated rocks. [93] related the higher stability of micas compared with amphiboles to lower silica activity, lower water fugacity, and moderately high K₂O of such magmas ([53]). Mica compositions for each individual rock are given in Supplementary file S.5 together with average compositions for each intrusive unit; Table 5 gives the averages for each unit with petrologically significant parameters. Zoning is limited, as reflected by the MgO standard deviation (~5% - 6% relative), which is only slightly larger than found for primary clinopyroxene. Biotites from alkaline basaltic/ultramafic rocks are commonly enriched in Ba and Ti, leading to multivalent cation, multi-site cation exchange, combined with the presence of extra-oxygen replacing OH, F, Cl anions in the Wsite (denoted ^WO²⁻) ([41]; [13]; [91]; [40]).

[51] provided partial compositions for a Ti-rich biotite separate from Marangudzi alkali gabbro A12 (atomic Fe³⁺/ΣFe = 0.136). It is probable that ~30% oxybiotite molecule (^wO²⁻) is present ([40]); thus, the [61]) spreadsheet for biotite is used to estimate the ^wO²⁻ values for the average Marangudzi biotite compositions. Cell formulae data given for Marangudzi EMP analyses (all Fe as FeO) were calculated for [22+(^WO²⁻)/2] oxygens and 16 cations ([40]) to obtain Fe³⁺ and Fe²⁺ values (Table 5). Molar proportions of annite (Fe²⁺ end-member) confirm that annite contents increase with evolution. No tetrahedral Fe³⁺ or Ti⁴⁺is present and all samples contain octahedral Al. The biotites in gabbros have excess O (^WO²⁻) ranging from ~0.9 for the most magnesian sample to ~1.5 in the most evolved orthocumulate (T168); thus, for most of the Marangudzi biotites, the average oxymica proportion is 45% - 50%. However, the most evolved microfoyaite (A14) has 20% while a leucomalignite with a deuteric mica (A28) has 0% oxybiotite (Table 5); the samples show a coupled relationship between Ti and ^WO²⁻ ([40]).

Figure 7. Biotite atomic proportions (apfu) for Mg (7a), Na (7b), Ti (7c) and Mn (7d) versus iron variation index I.E. Symbol types are identified as in Figure 6. See text for explanation.

Figure 7(a) shows the plot of Mg vs I.E. to assess individual biotite compositions from different intrusive units. Thus, biotites from two nepheline-bearing syenites show that one is very primitive (T301 with I.E. 0.38 and 2.2% Ba) and the other is evolved (T175 with I.E. 0.73 and BaO < 0.05%). The former is present as small inclusions in amphibole and the latter occurs as small laths at amphibole rims. Sample T301 has lower I.E. than any of Ba-rich nepheline monzonite biotites and is more primitive than biotites in two of the evolved gabbros. One leucomalignite has a primary magmatic biotite (G260) with an I.E. matching that of the average nepheline monzonite while those for the deuteric biotite in A28 match that of an evolved microfoyaite biotite (A14). Nevertheless, the average biotites for an orthocumulus gabbro-to-microfoyaite series show a clear trend of decreasing Mg with increasing I.E. In Figures 7(b)-(d), the individual samples show little clear trend for the gabbros, while nepheline monzonite and nepheline-bearing syenite follow with a decrease for Ti (Figure 7(c)), and an increase for Mn (Figure 7(d)). For the intrusive unit averages, Na may show a small decrease with increasing I.E, Ti, a clear decrease, and Mn a clear increase.

Figure 5 shows that the biotite compositions define a continuous trend from gabbro through to nepheline-bearing syenites host rocks (mg#range 76% - 26%; Suppl. file S.5; Table 5). The mica compositions in the last-intruded leucomalignite unit are different in the two samples studied; G260 is definitely a primary magmatic biotite phase (mg# = 40.4%) but the siderophyllite in A28 is a hydrothermal phase (mg# 14.9). Crystallization of biotite as a major magmatic phase ended before formation of the foyaites, with plagioclase-bearing foyaite A17 lacking modal biotite even though clinopyroxene is present. Oversaturated rock biotites have mg# ranging 62% - 78% but their plotted symbol is not differentiated from that for the other biotites to avoid unnecessary complexity. The parameters given in Table 5show low model Fe³⁺/ΣFe values for biotites, with no clear difference for the silica-undersaturated and over-oversaturated rock series.

Figure 8. (a) Modified [30] variation diagram for average biotite compositions for each of the main Marangudzi rock units to define how the intersite-intervalence evolutionary trend is critically dependent on the presence of the oxybiotite molecule ([40]). The large dark green triangles mark the positions of hypothetical mica end-members, phlogopite (Ph, K₂Mg₆[Si₆Al₂O₂₀](OH)₄) and oxymica Al-Ti-kinoshitalite (BaMg₄Ti₂ [Si₄Al₄O₂₀]O₄, [55] 2005; [40]). (b) Relationship shown between the oxy-mica function (^WO²⁻, apfu) and deduced temperature for each of the rock types. The linear fit shown is for the average biotite compositions for the whole Marangudzi complex. See text for the explanation of the compositional variations.

[40] showed how high-temperature magmatic biotites might have Ti and Ba entry controlled by complex vacancy-free, multivalent cation interchanges. The [30] exchange mechanism: 2(K + Na) + 4M²⁺ + 4Si ↔ Ba + 3Ti + 4Al is used here to model that interaction;M²⁺includes all Fe as Fe²⁺, with Mg and Mn. The plotted position of each sample is not affected by varying Fe/Mg ratios and variations mainly reflect variable Al and/or Ti. Most of the Marangudzi samples plot in the same tight group with mica in leucomalignite G260 at the bottom (Figure 8(a)); all are primary micas. Micas from leucomalignite A28 are displaced upwards from that group and are hydrothermally-formed Al-rich siderophyllites. Biotites from the gabbros, minor basics, and nepheline-bearing syenites plot over similar ranges of composition to the nepheline monzonites plotting at the top of the group, reflecting their higher Al, Ti and Ba contents. The dark green triangles mark the positions of [30] end-member phlogopite (K₂Mg₆[Al₂Si₆]O₂₂(OH)₄, see annite) and an Al-rich K-Ti-kinoshitalite (K₂Mg₆[Al₄Si₄]O₂₂(OH)₄). Neither has vacancies and the compositional trend passes through the Marangudzi magmatic mica compositions and is only slightly lower than the A28 hydrothermal micas; thus, the magmatic biotites are clearly stoichiometric with full cation sites ([41]; [40]). Biotites from gabbros and nepheline-bearing syenites plot over similar ranges of composition, with nepheline monzonites (blue-green squares) plotting at the top of the compositional series reflecting their higher Al, Ti and Ba contents.

The final rows of data given inTable 5 (and Suppl. file S.5) give tentative estimates for the crystallization temperatures of Marangudzi micas; both sets of estimates provide possible temperature ranges rather than absolute values. The top temperature values are obtained using biotite atomic parameters for Ti and mg#[Mg/(Mg + Mn + ΣFe)] ([49]); see Suppl. file S.5, biotite folder. The temperature estimates given below those values are based on Ti/Fe²⁺ of magmatic Ti-rich micas calibrated with Fe-Ti-oxide equilibrium ([64]). Also given in Table 5 are estimates of the pressure of formation of Marangudzi biotites based on the total Al content ([90]). Later work has shown that the [49] method can be widely applied, including to differentiated mafic igneous complexes, provided that a Ti-rich phase (ilmenite, rutile, or titanite) is present ([74]; [54]; [84]; [9]). Using the Luhr et al. calibration, biotites from 8 gabbros provide an average temperature of 1070˚C ± 130˚C, considerably higher than the Henry et al.’s average of 795˚C ± 20˚C; by contrast, the biotites from two altered gabbros (A4 and T123) have lower values of 705˚C and 709˚C, respectively. The more fractionated rocks all show closer temperature estimates for average biotites in each of the intrusive units (Table 5). Figure 8(b)shows the average of the two estimated temperatures for biotites versusthe extra-O factor (^WO²⁻); the latter is strongly dependent on Ti content. The linear fit is for the full dataset but individual intrusive unit plots show displaced sub-parallel trends; nepheline monzonites have average formation temperatures about 50˚C below the gabbro trend, with more-evolved nepheline-bearing syenites (lower mg#), least evolved microfoyaite and leucomalignite samples (higher mg#) about 40˚C lower still, reflecting the order of formation in the silica-undersaturated differentiation sequence.

Alkali basalt differentiation from liquidus to solidus conditions atP_H2O = 1 - 2 kbar (5.5 - 7 km) is ~1170˚C to 870˚C ([94]; [83]) with equivalent temperatures for nepheline syenites and quartz syenites of ~900˚C to 750˚C ([69]; [70]; [73]; [67]). Thus, biotites in the orthocumulate gabbros and nepheline monzonites precipitated as primary phases with nepheline-bearing syenite biotites forming closer to the solidus, and biotite re-equilibration probably continued subsolidus (see Section 6).

Primary, green- or green-brown calcic amphiboles are common in all Marangudzi ring-type units but also are present as minor phases which have re-equilibrated subsolidus in gabbros. [8] and [38] reported chemical compositions for Marangudzi amphibole separates with formulae calculated to 24 anions without considering an oxy-amphibole component.

With regard to possible cation vacancies and the presence of oxyamphibole ([33]) components, formula calculation for Marangudzi amphiboles was addressed by [44]. For analyses lacking water data, anhydrous, stoichiometric cell proportions are usually calculated on a 23 oxygen basis; this assumes the presence of exactly 2(OH, F, Cl)^– anions. In his consideration of the cell formulae for biotite, [40] stressed the requirement that, for calculating mica formulae, each site vacancy must be considered as a real component. The same approach is used here for amphiboles lacking water and Fe valence analyses ([44]), with formulae calculated for (23 + ^WO^2–/2) oxygens and 16 or 13 cation cells. Based on this approach, the total cation values (Σcats¹⁶) range 16.11 - 16.54 (average 16.26) for silica-undersaturated rocks and 15.96 - 16.09 (average 16.05) for quartz syenites. The very low 16 cation total for the latter indicates the presence of significant vacancies in the (Na + K) component; thus, the equivalent Σ¹³value [i.e., Σcats¹⁶ – (Ca + Na + K)] ranges 13.23 - 13.38 (average 13.31). However, for the apparent silica-undersaturated average amphibole cation total of 16.26 apfu with all Fe as FeO, the equivalent T + C cation total is 13.21 apfu; thus calculated 16-cation and 13-cation model Fe³⁺ contents agree within error, indicating that those amphiboles are stoichiometric [i.e., 24 anion T + C cation totals close to 13.000 (∆C = 0.0), [44]].

The full Marangudzi EMP amphibole dataset is attached as Supplementary file S.5 butTable 6 gives average amphibole compositions for each of the Marangudzi major intrusive units with wt.% EMP spot analyses for each individual Marangudzi sample entered with all Fe as FeO. Most samples show only limited compositional zoning, as is demonstrated by the standard deviations for the analysed MgO values. The atomic proportions calculated for all elements, as well as the model Fe³⁺ and Fe²⁺ based on a stoichiometric cell with (23 + ^WO²⁻/2) oxygens and 16 cations (i.e., ∆C = 0) are given in Table 6 (see Supplementary file S.5 for individual samples). The rows below the cell formulae data give the ^WO²⁻ factor for each sample obtained using the [63] spreadsheet, together with calculated values for I.E., mg# and Mg/(Mg + Fe). Amphibole names given in Suppl. file S.5 are mainly taken from those suggested in the Li et al. spreadsheet and are based on the IMA scheme ([33]; [76]). The least evolved amphiboles in nepheline monzonites and silica-undersaturated rock syenites are magnesio-hastingsites with some having slightly higher Ti, giving an additional Ti-rich prefix, however, most of the amphiboles are hastingsites.

Table 6.Average amphibole for each Marangudzi major intrusive unit.

Note: mg#^& = Mg/(Mg + Mn + Fe).

The full dataset includes separate analyses for core and rim of larger grains together with compositions for smaller grains in the matrix of porphyritic samples; thus, individual samples show significant ranges of I.E. but that does not necessarily lead to significant changes in Al, Na and Ca (see below and full dataset in Supplementary file S.5). Figure 9 shows plots for Si, Ca, Na, K, Ti and Mn vs I.E. for the average intrusive unit compositions. Figures 9(b)-(c) show data for Ca and Na variation in each individual sample with rim compositions marked with large open symbols (some joined to core compositions with labelled tie lines); thus, the plotted points are scattered, making it difficult to deduce clear trends. Figure 9(b) shows that the nepheline syenite and leucomalignite members show little clear trend but rims for nepheline-bearing syenites and foyaites occasionally show lower Ca contents. For Figure 9(c), Na shows no clear trends for nepheline monzonite and leucomalignite amphiboles (see Al not shown); with some rims in nepheline-bearing syenite and foyaite amphiboles having higher Na. The limited changes in Na and K contents in the less evolved ring units are probably controlled by coexisting alkali feldspar and nepheline buffering the alkali contents.

Figure 9. Amphibole atomic proportions (apfu) for Si (9a), Ca (9b), Na (9c), K (9d), Ti (9e) and Mn (9f) versus the iron variation index I.E. Individual rock data are shown with small symbols for Ca and Na to show the compositional within each intrusive unit and the average values for separate intrusive units are shown with a large symbol having the same colour as for the individual parent rock samples. Only average values are shown for Si, K, Ti and Mn to clarify trends within the silica-undersaturated rock types and to emphasise differences for the silica-oversaturated rock amphiboles. See text for explanation.

Each Marangudzi intrusive unit tends to have amphiboles with distinct mineralogical, chemical and textural features. In particular, the nepheline monzonite, nepheline-bearing syenite and foyaite units show overlapping extents of differentiation (Figure 5), perhaps related to local concentration of volatile components developing local pegmatite-like segregations similar to those shown by differentiated minor alkaline intrusions (e.g., [75]; [10]; [66]; [27]). This concept is tested by calculating average mineral compositions for each Marangudzi intrusive unit. Thus, in Figure 9, the average amphibole composition for each rock type is shown with a large coloured symbol, showing clear evolutionary trends of decreasing Ca and Ti (Figure 9(b) and Figure 9(e)) and increasing Na and Mn (Figure 9(c) and Figure 9(f)), coupled to increasing I.E. for the series nepheline monzonite-nepheline-bearing syenite-foyaite-microfoyaite. By contrast, the last intruded leucomalignite unit has amphibole compositions overlapping with those of the earlier intruded nepheline monzonites (also see Figure 5; Supplementary file S.5). The other panels in Figure 9 show average intrusive rock unit values with Si and K showing little variation (Figure 9(a) and Figure 9(d)) (see Al not shown) with increasing I.E. Each of the panels in Figure 9 shows the average composition for the different quartz syenite units (large open symbols) compared to those for the silica undersaturated rock types having similar I.E. values. Thus, quartz syenites have amphiboles with much higher Si and lower K (see Al, not shown) and slightly lower Ca, Na and Mn than the amphiboles in nepheline-bearing rock types, reflecting the critical compositional differences between the distinct parental magmatic compositions (see [38]).

The bottom rows in Table 6 show the estimated crystallization temperature, pressure, oxygen fugacities and magma water content based on experimental studies; the different petrogenetic approaches are based on: (i) amphibole Ti contents ([34]); (ii) statistical analyses of amphibole bulk chemical compositions ([82]); (iii) amphibole Si, Ti, total Fe and Na (equation 5, [79]) [Suppl. file S.5, amphibole folder]. The Putirka calculation consistently provides temperatures significantly higher than Helz values and an average of these is rather similar to the Ridolfi and Renzulli values ([82]). The data will be fully considered below (Section 6) but some difficulties are summarised here for average data. In particular, note that the [82] model pressure, temperature, oxygen partial pressure and silicate melt water content values obtained from their equations are based on analysed amphibole compositions obtained using published experimental work, mainly on intermediate and silica-saturated parent rocks at controlled physical conditions. Only two of the 21 experimental datasets were on silica-undersaturated alkali-rich basaltic parents and none were alkali-rich syenites. The average amphibole in Marangudzi nepheline monzonites and syenites is 5.5 ± 0.7 kbar and 895˚C ± 50˚C and for the average amphibole in quartz syenites is 1.4 ± 0.7 kbar and 795˚C ± 20˚C. While both sets of data for temperature estimates are plausible, the pressures for the silica-undersaturated rock units are higher than would normally be expected for sub-volcanic ring intrusion and, as discussed above, may be 2 - 3 kbar too high unless nucleation and early growth occurred in the lower crust with rapid transfer upwards of magma to ring-dyke intrusion at higher levels. The lower pressure estimates for the quartz syenites at ~1 kbar are consistent with the probable intrusion depth. In addition, the unexpectedly high ([82] oxygen fugacity estimates for amphiboles also may reflect the lack of nepheline-rich samples in the experimental database; thus, the model data for the Marangudzi nepheline syenite amphiboles (average ∆NNO +3.4 ± 1) are probably 2 - 3 orders of magnitude too oxidising, while the data for quartz syenite amphiboles (average −2.0 ± 0.5) are, as expected, closer to the QFM buffer ([1]). In addition, the range of calculated ∆NNO for the Marangudzi silica-oversaturated and undersaturated syenites covers 5 orders of magnitude; experimental work for basalt glasses on the FMQ buffer at the same temperature might have Fe³⁺/ΣFe values varying from 0.2 to 0.7 (e.g., [26]; [68]) but averages for analysed Marangudzi amphiboles show similar Fe³⁺/ΣFe for nepheline syenite amphiboles of ~0.23 and ~0.16 for quartz syenite amphiboles ([38]) consistent with both amphibole species having formed with similar QFM properties. In sum, it seems that only qualitative estimates for oxygen fugacity should be made for evolved potassic, silica-undersaturated alkaline igneous rocks using the method of [82].

For all of the physical and chemical formation conditions estimated here, it is assumed that the contrasting values for the different rock types are more reliable than the absolute values. Thus, the Cl-rich ferropargasite in altered gabbro A4 is clearly a hydrothermal, subsolidus phase with a crystallization temperature of ~600˚C at an oxygen fugacity slightly more oxidising than the QFM buffer; however, Ridolfi and Putirka temperatures for the amphibole in gabbro T123 suggest that it is a late magmatic phase at ~850˚C. All amphiboles in the nepheline monzonites, nepheline-bearing syenites, foyaites and leucomalignites give similar temperatures for each of the geothermometers quoted here, with an average of ~850˚C - 900˚C; ∆NNO oxidation estimates are probably 2 orders of magnitude too high, see earlier. The temperature for the nepheline monzonites is late-stage magmatic but for the nepheline syenites an early- to mid-magmatic crystallization temperature is indicated. The silica-oversaturated rock amphiboles all suggest a significantly lower temperature of 750˚C - 800˚C under oxygen fugacities similar to those for the nepheline syenites. As expected, where compositional data are available for both amphibole cores and rims, the latter temperatures are 20˚C - 30˚C lower.

Full compositional data are given for plagioclase and alkali feldspar in Supplementary file S.5. Average analyses for each intrusive unit are given here in Table 7. The feldspar An and Or proportions are displayed in Figure 10 as a function of I.E. for coexisting primary biotite in gabbros, nepheline monzonites and nepheline-bearing syenites and amphibole for foyaites, leucomalignites and quartz syenites, allowing assessment of how well-coupled the feldspars and mafic minerals are between the fractionated intrusive rock units. Standard deviations for CaO in plagioclase and K₂O in alkali feldspar reflect the zoning within each phase and, in the case of plagioclase, the most calcic compositions are usually very similar for cores in cumulus grains and for the central parts of matrix plagioclases. The bottom rows of Table 7 give the extreme variations for plagioclase (An variation) and alkali feldspar [Or and Ba-feldspar (celsian molecule Cn)].

Table 7. Average feldspar compositions in main Marangudzi intrusive units.

Continued

Figure 10. Plagioclase and alkali feldspar mole % An (10a) and Or (10b) values for the average feldspar compositions in each of the different intrusive units in the Marangudzi complex. Linear fits are shown for average An in plagioclase and average Or in alkali feldspar. See text for explanation.

Five mesocumulates define the most calcic plagioclases overall with An averaging 80.3 ± 1.6 (100, 100A, G265, A4, T398) followed by three mesocumulates with 75.2 ± 2.2 (T123, T205, G266); four orthocumulates, as expected, have a more evolved An of 66.8 ± 7.3 (A8, A12, T168, T203) [Suppl. file S.5 and Table 7]. Orthocumulate A12 plagioclase has thick sodic rims (Figure 10). Table 7 also gives averaged analyses of plagioclase phenocrysts in minor basic intrusions (~An₅₄ for microgabbro, T101 and basanite dyke, T178), both slightly more evolved than for the orthocumulate gabbros. However, the basanite has large, very calcic plagioclase cores (An₉₃) which are probably antecrystic. The average plagioclase compositions for the basaltic composition rocks are displayed in the feldspar ternary variation diagram (Figure 11), in which Ba and Sr are added to the Ca end-member; the most An-rich point in Figure 11 is for mesocumulate T123 while the orthocumulates A12 and T168 have An₆₃ and An₆₁, respectively.

Figure 11. Compositions for average plagioclase and alkali feldspars for the different Marangudzi units plotted in the variation diagram for Ca (+ Ba + Sr) (An)-Na (Ab)-K (Or) as mole %. Coexisting plagioclase-alkali feldspar tie lines are shown for gabbros (100A, A12), nepheline monzonites (A19, T310, T303, T174) and nepheline syenites (A31, A17). The high “An” parameters for some nepheline monzonite and leucomalignite alkali feldspars mainly mark their enrichment in Ba and Sr, rather than Ca (see text).

Alkali feldspar compositions are available for four gabbros (100A, G266, A12 and T168) (Table 7, Figure 10), a microgabbro and a basanite. All are late-crystallized phases located in the interstices of plagioclase subhedral grains and are potassic, varying from Or_88.5 for sample 100A, which coexists with calcic plagioclase (An₈₁), to Or _76.5 for T168, which coexists with evolved plagioclase (An_60.3). Sr composition estimates are available for bulk alkali feldspar separates from Ne-bearing syenites A26 (1600 ppm) and A31 (590 ppm); foyaites A1 (660 ppm), A20 (720 ppm), microfoyaites A14 (30 ppm); and leucomalignite A28 (800 ppm) [<0.3 Sr feldspar]. However, BaO (wt.%) values of 1.00 (100A), 1.29 (G266), 1.18 (R12) and 2.36 (T168) confirm the enrichment of Ba at this stage of the magmatic evolution (Cn in mol. %, 1.83, 2.54, 2.05 and 4.31 respectively) [Suppl. file S.5, Table 7].

The nepheline monzonite and nepheline-bearing syenites units both contain plagioclase (usually cumulus) and alkali feldspar (interstitial) (Table 7). Plagioclase only occurs in one foyaite (A17) but also occurs in microfoyaite G133 (Table 7 and Figure 10). In nepheline monzonites, the most calcic plagioclases (An87, T308 and A19) match that for the gabbro; the most evolved sample (T174) has an average plagioclase An₄₄, slightly less calcic than that for orthocumulate A12. The plagioclase in the nepheline-bearing syenites analysed extends this trend to An₃₂ (A31). Both the Sr and Ba contents of the plagioclase in the nepheline monzonites and nepheline-bearing syenites show some fluctuation but the most evolved rocks have Sr and Ba at the detection limit. The average compositions of interstitial alkali feldspars in the nepheline monzonites vary from Or₈₇ (T308) to Or₆₁(T303) but the variable Ba and Sr contents show BaO up to 5.5% (T310) and SrO up to 0.59% (T174). The nepheline-bearing syenite alkali feldspars have less variable compositions (Or_67-58) with maximum SrO of 0.17% and BaO of 2.3% (Table 7). Figure 11 shows a substantial scatter of points for K-rich alkali feldspars in microgabbro T101 and nepheline monzonites A19 and T310, mainly reflecting their variable and exceptional Ba enrichments (Suppl. file S.5, Table 7). Figure 10(a) shows that the Marangudzi plagioclase feldspars from the gabbros, nepheline-bearing syenite and foyaitic rocks fall close to a trend of decreasing An (and increasing Ab, not shown here) with evolution. However, the nepheline monzonites are displaced upwards from that trend, reflecting that the “An” content for some rocks is strongly enriched with Ba (and Sr), which are included in that component.

Alkali feldspar in foyaite A17 and microfoyaite G133 show the typical alkali feldspar alignment. The compositions of both the plagioclase and the alkali feldspar for A17 overlap with those for the evolved nepheline-bearing syenites for most of the end-member components but the plagioclase in G133 is less calcic with much lower Sr and Ba contents (Table 7). All the foyaites show that the alkali feldspar compositional trend continues with low An and then high Ab and Or showing relatively little change through the series from A1 to T304. Sr and Ba components decrease to very low concentrations for the most evolved foyaites. Microfoyaite A14 has a coarsely perthitic alkali feldspar allowing analyses of the Or and Ab-rich components for two-phase Ab-Or phases (Table 7); these analyses confirm that the Or-rich solvus flank is shallower than that for the Ab-rich flank. Figure 11 shows how the foyaitic plagioclases extend the trend from the nepheline-bearing syenites towards more Na-rich compositions; the Or content remains low throughout the plagioclase fractionation series. The alkali feldspars in the foyaites mainly plot close to the Na-K (Ab-Or) sideline, reflecting their low contents of Ca, Sr and Ba (Figure 11).

The magnesian compositions of the mafic minerals in the leucomalignites have shown that they do not fit into the same evolutionary pattern as the other silica-undersaturated rocks and the same is true for the alkali feldspars. Their compositions (Suppl. file S.5) show that the pseudoleucite-free rocks A27 and 730 have strongly zoned alkali feldspars with the compositions for A27 and sample 730 averaging Or₆₇ and Or₇₀, respectively. Both are very rich in Sr and Ba with low Ca and the elevated “An” contents shown for those alkali feldspars clearly reflect the high Ba and Sr content rather that for Ca. The feldspars in both rocks enclose euhedral nephelines and mafic phases; the zoning trend is exceptionally well developed in A27 with SrO 2.0 - 0.5 wt.% and BaO 12.6 - 2.5 wt.%. The CaO content is uniformly low and Na shows a small decrease. There is an inverse relationship between orthoclase content, which ranges from Or₄₉ to Or₇₉ in order of decreasing Ba; the combined Or + Sl + Cn is effectively constant at 82 ± 2 mol.%. In Figure 11, the two leucomalignites with Sr- and Ba-rich alkali feldspars plot well away from the Na-K side line.

Figure 11 shows that leucomalignite alkali feldspars are very close to being Or end-members. These relationships are consistent with the feldspar in A28 mainly being derived from pseudoleucite which, in turn, is likely to have formed from primary leucite which typically has very low Na, Ca and crystal/liquid D values for both Ba and Sr ([37]; [48]). Leucomalignite feldspars in G260 and T330 have intermediate features, with some compositions having very low Sr (<0.1%) and Ba (<0.2%) and others with much higher concentrations (0.4% and 1.0%, respectively). Thus, it is probable that some feldspar (and nepheline) grains with high K and low Sr and Ba might be derived from pseudoleucite disaggregation [leucomalignites A28, G260 and T330 (average Or₉₈)]. Figure 10(b) shows a clear trend of decreasing Or with evolution ranging from gabbros to nepheline monzonite to nepheline-bearing syenites, foyaitic rocks and also includes the quartz syenites (as did the An trend in Figure 10(a)); this reflects a genetic relationship between different silica-undersaturated Marangudzi rock units. The leucomalignite is undoubtedly part of that same genetic complex but the displacement of the average sample from the general trend is probably related to the exceptionally K-rich feldspars forming from breakdown of earlier-formed leucite.

Quartz syenites mainly have microperthitic alkali feldspars but in some perthites are coarser while others are cryptoperthites. The compositions available (Supp. file S.5, Table 7) are mainly for unaltered plagioclase cores and separate grains, and for cryptoperthitic alkali feldspars, but separate exsolved phases were analysed in some coarsely exsolved syenites. The average plagioclase in porphyritic quartz syenite A21 is An₃₂ but the most Ca-rich centres are up to An₅₀ with rims as sodic as An₁₉. Alkali feldspars average Or₆₃ with core/rim ranges of Or₇₂ to Or₅₃and Ba-feldspar (Cn) from 1.8 to 0.3 mole %. The separate plagioclase grains in the inner quartz syenite and outer porphyritic quartz syenite units are less calcic than those in the porphyritic quartz syenite (An₂₆ and An₇, respectively). Cryptoperthite in plagioclase-free inner quartz syenite ranges Or_52-29 and the coarsely alkali feldspar in porphyritic quartz syenite T17 has compositions Or₉₃and Ab₉₈ for the separate components. Ba (and Sr) contents for all the feldspars in the quartz syenite units are much lower than those of the silica-undersaturated rocks.

Tie-lines for averaged plagioclase-alkali feldspar compositions for each parent rock are shown in Figure 11 for gabbros and the silica-undersaturated differentiated rock series. Thus, tie-lines for mesocumulate gabbro 100A and orthocumulate gabbro A12 show that the latter rock is more evolved with a lower An content but with similar alkali feldspar compositions; this leads to a slightly “flatter” tie line for A12. The two feldspar pairs for nepheline monzonites A19, T310, T174, and T303 show similar evolutionary shifts to lower An plagioclases but also show a trend towards lower Or contents; the slopes of the tie lines decrease with differentiation as would be expected for coexisting equilibrium plagioclase and alkali feldspar ([12]; [6]). Nepheline-bearing syenite A31 has a 2-feldspar tie-line reflecting a more evolved composition, as does that for foyaite A17. Tie lines are not shown for Marangudzi quartz syenites but the compositions summarised above would plot in the same general regions as those for nepheline-bearing syenite A31 and foyaite A17.

Suppl. file S.5 and Table 7 give data for the temperatures provided by geothermometry calculations for all of the two-phase assemblages (including those for coarse-grained perthites) [[80]; Equation 27a, “global regression calculation”]. The least evolved gabbro (100A) has an estimated temperature of ~1000˚C (average mesocumulate value 968˚C) but the orthocumulates (A12 and T168) are more evolved and are lower with ~900˚C. Average equilibration temperature for feldspar pairs in the nepheline monzonites is also 900˚C (Table 7). The more evolved feldspar pairs in nepheline-bearing syenites show averages of 845˚C and only foyaite and microfoyaite, with coexisting separate feldspars having lower estimates of 825˚C (foyaite A17) and ~770˚C (microfoyaite G133). In contrast, the coarsely exsolved alkali feldspar in microfoyaite R14 suggests a temperature of ~420˚C consistent with its formation by post-magmatic hydrothermal recrystallization (see earlier). The feldspar pairs in the least-altered quartz syenites (porphyritic quartz syenites A21, A16 and inner quartz syenite A6) suggest equilibration close to the solidus (~800˚C) but inner quartz syenite and porphyritic quartz syenite have lower temperatures of ~640˚C and 550˚C, respectively, indicative of post-magmatic processes. The variation in temperature reflects the trend of changing tie-line slopes displayed in Figure 11. Feldspar data will be assessed further below (Section 6).

Nepheline occurs in some gabbros as small interstitial grains and then increases in amount as the rock types become more silica-undersaturated. The full nepheline dataset is given in Suppl file S.5 and Table 8gives averages for the silica-undersaturated intrusive units. Nepheline in orthocumulate T203 is the most Ca-rich nepheline found (CaO 3.3 wt.%) with less in nepheline monzonites ~2.5%. Nepheline-bearing syenite to foyaite averages range 1.9% - 1.2% with evolved microfoyaite (A14) having <0.5% CaO. The heterogeneous leucomalignite nephelines have CaO varying from 0.5 to 1.5 wt.% CaO. The basanite dyke (T178, 2.4% CaO) is similar to that in the nepheline monzonites. Sr and Ba were always below the detection limit of ~0.015%.

Atomic formulae for nephelines were calculated to 32 oxygens (Table 8) with end-member formulae determined for a stuffed tridymite model ([4]; [39]). Thus, Ca is calculated as the tridymite-structured Ca-nepheline molecule rather than as anorthite, which most authors use. The Ca-Ne molecule has a value double that of An with an excess SiO₂ (denoted Q’ here), half of that reported for an An-calculation.

Table 8.Nepheline compositions in Marangudzi mafic rocks and silica-undersaturated syenites.

The proportions of the end-members in the Marangudzi nephelines were calculated using [45]. Stoichiometry parameters ∆Al^cc and ∆T^ch, the ∆Al^cc/∆T^ch ratio, and the mol. % values for the various nepheline end-members are given in Table 8. Ks contents increase rather irregularly from Ks_12.5for nepheline in the least-evolved basic rock to Ks₁₈in the most-evolved microfoyaite; alkali feldspars in the same rocks range Ks_80-60 but note that lower Ks contents are present in nephelines which coexist with alkali feldspars enriched in Ba coupled with lower K. Leucomalignite nephelines are slightly more potassic ranging Ks₁₇ to Ks₂₃ but these nephelines are still less K-rich than the ideal nepheline formulae (Ks₂₅), even though the alkali feldspars in the leucomalignites are almost K-end-members (~Or_96-99) in leucomalignites A28, G260 and T330 (Suppl. file S.5).

[39] and [45] recommended that the excess SiO₂ parameter Q^(Si-Al) is the most reliable, as its calculation does not depend on Na, which is vulnerable to loss in the electron microprobe beam. This recommendation is followed here. The stoichiometry parameters ∆Al^cc and ∆T^ch values are mostly <0.6, which [39] suggested is the limit for considering acceptable analyses. The coupled ∆Al/∆T ratios (Table 8) are close to 1.135, which is the ideal value for stoichiometric stuffed-tridymite structure nephelines ([39]). The Ca-Ne end-member in Marangudzi nephelines varies from 16.5 mol.% in gabbro T203 to ~5% in evolved foyaites; these represent late-magmatic minerals. Microfoyaite A14 has exsolved perthitic alkali feldspars, pointing to sub-solidus hydrothermal reactions and the nepheline in that rock has the lowest Ca-Ne contents, reflecting this post-magmatic stage. In addition, magmatic nephelines showing negative ∆Al/∆T ratios close to 0.5 might have small amounts of a primary cancrinite molecule in solid solution ([39]). The high delta values for the nepheline in some foyaites (A1, A20, A14) suggest the presence of primary cancrinite in solid solution. On that basis, calculation of these compositions using Spreadsheet 2 ([39]; [45]) provides estimates of 4 - 6 mol.% vishnevite molecules (ideal Na-cancrinite). In addition, the magmatic-textured nepheline in nepheline monzonite T308 shows lower delta values of ~0.1 (Table 8) as well as a low model vishnevite content of 1 mol.%; however, the associated intergrowth between early plagioclase and nepheline in this sample has delta values of over 0.5 (vishnevite ~5 mol.%), perhaps consistent with a very-late stage, localised alkali-exchange/recrystallization of primary plagioclase to form such a texture ([42]; [72]). The coexisting alkali feldspar and nepheline compositions show typical magmatic tie lines when plotted in the Ne-Ks-Qz residua system (see Suppl. file S.5; [88], 1954 and [46]).

Finally, excess Q^(Si-Al) values ([39]) do not show clear variation with rock type; indeed, apart from two of the more evolved nepheline monzonites, the whole series from gabbro through to evolved foyaites have Q^(Si-Al) falling within the range 4.5 to 6 mol.%. The nepheline in all of rocks is the last main phase to crystallize under magmatic conditions, which, in turn, suggests similar solidus temperatures and similar compositions for each fractionated residual melt. However, leucomalignite rock types show more variation (Suppl. file S.5). Thus, the least evolved leucomalignite R730 has Q^(Si-Al) = 5.2 mol.%, followed by the similar textured rock A27 with a value of 3.8 mol.%; neither of these rocks contains pseudoleucites. Leucomalignite A28 has abundant large pseudoleucites with some having the coarse-grained K-rich alkali feldspar and nepheline grains disaggregated to define large felsic patches; this sample has the lowest Q^(Si-Al) of 1.6 mol.%, reflecting a source from the pseudoleucite rather than direct crystallization from the magma. Leucomalignites G260 and T330 have white patches of disaggregated pseudoleucites, and the nephelines have intermediate Q^(Si-Al) values of 2.5 mol.% reflecting the mixed source of mineral components.

Sodalite and associated nepheline occur interstitially between larger alkali-feldspar laths. Coarser, more-even-grained nepheline-rich rocks in the foyaite ring-unit have larger rounded grains of sodalite (1 - 2 mm across) with “trains” of fluid inclusions. Compositions for sodalite are given in Suppl. file S.5 and averages for each intrusive unit in Table 9. The most reliable analyses obtained with a larger beam (3 - 5 μm) have close to 8 atoms Na pfu with only very small concentrations of K and Ca. The presence of Ca and low analytical totals could be explained by the presence of some Ca-cancrinite ([39]). Indeed, separate grains of Ca-bearing cancrinite (9.3 wt.% CaO) have been analysed in microfoyaite A14 and similar composition phases have also been found in leucomalignites A27 and sample 730.

Other hydrated aluminosilicates formed during deuteric alteration processes are analcime and phillipsite (nepheline monzonite T308) (Table 10); [46]) showed that at 1 kbar P(H₂O) analcime formed stably and rapidly at temperatures below 500˚C, well below the magmatic solidus, Ca-rich analcime is unstable above ~400˚C ([43]), while many other zeolites form at even lower temperatures (e.g., phillipsite, natrolite, <200˚C, [20]).

Table 9.Microprobe analyses (ED) of sodalites in Marangudzi silica-undersaturated syenites.

Table 10. Marangudzi oxide average analyses for magnetite and ilmenite in each intrusive unit.

Titaniferous magnetites are the main oxide minerals in the gabbroic and syenitic rocks. These can show fine exsolution lamellae of ilmenite and electron microprobe analyses of the Fe-oxides were obtained using a defocused beam (3 - 5 μm) or a rastered beam. Separate grains of magnetite and ilmenite occur in mesocumulate gabbro G205, evolved orthocumulate T168, and in nepheline monzonites, foyaites, leucomalignites, and quartz syenites (Suppl. file S.5). Multiple ilmenite spot analyses are only available for orthogabbro T168, nepheline monzonite T.174 and some quartz syenites. Magnetite and ilmenite compositions are given in Table 11 together with averages for each of the intrusive units. Atomic formulae are recalculated to 32 oxygen, 24 cations (magnetite) and 24 O, 16 cations (ilmenite) to obtain Fe valence data based on strict stoichiometry ([21]). The mol.% proportions of the end member molecules were calculated as follows: Spinel phase: Magnetite (M²⁺X³⁺₂O₄, M²⁺ = Fe²⁺ + Mg + Mn + Ni + Zn; X³⁺ = Al, Fe³⁺, V, Cr); ülvospinel (M²⁺₂TiO₄): rhombohedral phase ilmenite (M²⁺TiO₃); and hematite X₂O₃. In coexisting mineral pairs, spinel concentrates Al and ilmenite concentrates Mn.

In the gabbro to foyaite series, titano-magnetites tend to show a steady decrease in Al-spinel contents with little systematic change in the model Fe³⁺/Fe²⁺ ratio; V- and Cr-spinels decrease rapidly from the gabbros to the nepheline monzonites and are uniformly lower in silica-undersaturated rocks. The ülvospinel contents peak in the nepheline-bearing syenites and have extremely low values in the foyaites (Table 11). The Fe³⁺-spinel end-member (magnetite) is significantly higher in the foyaites than in the other nepheline syenites, reflecting a more-oxidised Fe valence. Overall, the coexisting ilmenite shows very limited variation except for Mn.

Table 11.Microprobe analyses of deuteric aluminosilicates.

Note: n.d., not detected; detection limit ~0.03wt.%.

The compositions of coexisting spinel and ilmenite are used to estimate “equilibration” temperature and oxygen fugacity ([60]); values are reported in Table 11 and Suppl. file S.5. For the silica-undersaturated series, an average ilmenite composition is used for each of the individual intrusive units. Average temperatures for the Fe-oxide pairs in the different rock types are mainly ~600˚C with log oxygen fugacities (log fO₂) of ~−22 bar, showing that re-equilibration of the Fe valence has occurred below the magmatic solidus. Coexisting ilmenite-magnetite pairs in the nepheline monzonite, silica-undersaturated “syenites” and quartz syenites generally plot close to the low temperature extension of the quartz-fayalite-magnetite (QFM) buffer curve.

Apatite is the main accessory mineral in the basic rocks and compositions are given in Table 12(a) and Suppl. file S.5. Apatite in the gabbros is likely to be a cumulus phase but most have similar compositions except for Cl and F. Unfortunately, Cl was not determined in apatite from plagioclase mesocumulate A4, although its halogen/hydroxyl site is full with 2.06 atoms pfu of fluorine. All the other compositions available have both F and Cl present (F > Cl), with halogens totalling 1.4 - 2.1 atoms pfu. Titanite is present as an accessory mineral and compositions are given in Table 12(b) for a nepheline monzonite (T178), two leucomalignites (G260, T330) and a porphyritic quartz syenite (T36). For each analysis, sufficient Al and Fe are present to combine with Ti to fill the corner-sharing octahedral sites.

Table 12.(a) Apatite analyses; (b) Microprobe analyses (ED) of titanite.

(a)

Note: n.d., not detected; detection limit Fe ~0.03 wt.%.

(b)

Note: Detection limit Fe ~0.03 wt.%.