Journal of Petrology | Pages |
© 1998 Oxford University Press |
The North Nyasa Alkaline Province (NNAP,
Bloomfield, 1970) of central and northern Malawi consists of seven intrusions (Kasungu, Chipala, Chikangawa, Mphompha, Telelele Hill, Ilomba, and Ulindi) that lie along a north-south trend roughly parallel to the current rift valley (
Fig. 1). The dominant lithology is nepheline syenite, but alkali syenite and granite occur at Mphompha and pyroxenites outcrop adjacent to, and within, the Ilomba intrusion. Intrusions of gabbro, ijolite, carbonatite, nepheline syenite and syenite in adjoining Tanzania and Zambia have also been included in the NNAP
(Bloomfield, 1970).
All the published radiometric ages for the NNAP intrusions of northern and central Malawi, plus possible additional members of the province in southern Malawi (Tambani), Mozambique, Tanzania, and Zambia are listed in
Table 1. Except for Ngualla, which may be considerably older, the radiometric ages suggest that there was widespread emplacement of alkaline intrusions in this part of Africa in late Precambrian time. Rb-Sr whole-rock isochron and Pb-[alpha] zircon ages obtained for Ilomba (685 ± 62 Ma and 655 Ma) and Chikangawa (650 ± 40 Ma) are significantly older than K-Ar biotite ages for the same intrusions (Ilomba, 508 ± 12 Ma and 490 ± 12; Chikangawa, 410 ± 16 Ma).
Bloomfield, (1970) ascribed the difference in ages obtained by the Rb-Sr and Pb-[alpha] methods and the K-Ar biotite method to a late Pan-African phase of cataclasis which affected the NNAP intrusions. We are currently in the process of obtaining new radiometric ages for the NNAP. U-Pb zircon ages obtained to date fall between 710 and 750 Ma (S. A. Bowring & G. N. Eby, unpublished data, 1997), indicating a somewhat older age for NNAP magmatism than previously determined. The new K-Ar biotite ages (G. N. Eby & H. W. Krueger, unpublished data, 1996) cluster around 450 Ma, presumably dating the Pan-African metamorphism. The younger ages are correlative with the emplacement of post tectonic granitoids in southern Malawi
(Haslam et al., 1983) that marks the end of this period of tectonism and metamorphism. Field work on all seven Malawi NNAP intrusions revealed little evidence for metamorphism, but subsequent investigation of pyroxene chemistries
(Woolley et al., 1996) confirmed the presence of a metamorphic overprint. Thus, the available data indicate that the NNAP plutons were emplaced at ~750-710 Ma and subsequently metamorphosed (~450 Ma) during the tectonothermal Pan-African event. Table 1. Radiometric ages for NNAP and related intrusions.
The abundance of silica-undersaturated felsic rocks and the virtual absence of mafic rocks raises interesting questions regarding the petrogenesis of the NNAP intrusions. The younger Chilwa alkaline province (135-113 Ma,
Eby et al., 1995) of southern Malawi consists of carbonatites, nepheline syenites and peralkaline granites, and is also characterized by a paucity of mafic rocks. Thus this portion of the African craton has been intruded by several generations of silica-undersaturated feldspathic magmas.
This paper deals with the geochemistry and petrogenesis of the southernmost, Kasungu-Chipala, and northernmost, Ilomba and Ulindi, plutons of the NNAP. These intrusions are representative of the range of compositions found for the silica-undersaturated rocks of the NNAP. The Central African basement has been affected by three major orogenic cycles: the Ubendian (~2300-1800 Ma), the Irumide (~1350-950 Ma) and the Pan-African (~900-450 Ma). The two earliest orogenic cycles resulted in amphibolite to granulite grade metamorphism of the Proterozoic basement
(Ring, 1993). During the Pan-African orogeny the rocks were metamorphosed to upper greenschist and amphibolite facies, and it was this metamorphic event that affected the NNAP plutons. The country rocks intruded by the NNAP plutons are mostly high-grade gneisses and granulites.
Kasungu Mountain is approximately circular in plan view (2 km in diameter) and forms a prominent hill rising ~400 m above the surrounding plain. Chipala consists of a number of small bodies, petrographically similar to Kasungu Mountain, that lie ~7 km to the east-northeast of Kasungu Mountain. Both intrude biotite-hornblende orthogneiss. The intrusive rocks are coarse-grained massive nepheline syenites with a high color index [melasyenites of
Bloomfield, (1965)] which occasionally display a strong parallel orientation of the feldspars. This orientation is particularly marked on the northern and western sides of Kasungu Mountain, where the rocks are also strongly porphyritic. There are two, apparently interlayered, varieties of nepheline syenite. One consists of turbid perthite, nepheline, mafic patches of aegirine-augite, amphibole, and biotite, occasional scapolite, and accessory titanite, apatite, zircon, magnetite and/or ilmenite. The other contains two discrete feldspars, clear microcline and sodic plagioclase, both with simple margins and commonly with 120° triple junctions, nepheline, evenly distributed aegirine-augite, amphibole and biotite, accessory titanite, but no opaque phase. Vertical mafic dikes, up to 30 cm thick and commonly sheared along the margins, cut the nepheline syenites. The mafic dikes are fine-grained, equigranular, and contain alkali feldspar, amphibole, minor biotite and pyroxene, and accessory titanite. The dikes, in turn, are cut by thin nepheline-feldspar veinlets.
The Ilomba intrusion outcrops within the Songwe Syenite complex. The Songwe Syenite consists of perthite, quartz (locally abundant), aegirine and occasional riebeckite. The Songwe Syenite locally shows signs of deformation with the development of a gneissic texture, indicating emplacement before, or syn-tectonic with, a deformation episode related to the Pan-African orogeny
(Ray, 1974).
The Ilomba intrusion (
Fig. 1) forms a horseshoe-shaped ridge which rises up the two arms to a summit ~335 m above the surrounding plain. The central core consists of nepheline microsyenite surrounded by an incomplete ring of biotite- and aegirine-nepheline syenite. Small areas of sodalite-nepheline syenite occur within the biotite-nepheline syenite, and the boundaries between the various types of nepheline syenite are gradational. These units are enclosed by an incomplete outer ring of coarse-grained perthitic aegirine syenite [perthosite of
Bloomfield et al., (1981)] that often shows a northwest to north-northwest trending foliation. Along the northwestern side of the western limb there is a gradation along strike from coarse-grained perthitic aegirine syenite to paragneiss
(Bloomfield et al., 1981). Small masses of this syenite also occur as xenoliths in the nepheline syenite. It is likely, therefore, that the coarse-grained perthitic aegirine syenite is part of the Songwe Syenite complex and is not a member of the Ilomba intrusion.
The biotite-nepheline syenite [biotite foyaite of
Bloomfield et al., (1981)] is coarse grained and consists of microcline perthite, abundant nepheline with some alteration to cancrinite, brown biotite, occasional aegirine-augite, sodalite, magnetite, calcite, titanite and zircon. Extensive fracturing of feldspar is ubiquitous and in many specimens large perthite prisms are surrounded and cut by vein-like aggregates of a fine-grained, granular feldspar which may be the result of deformation and comminution. Cross-cutting veinlets of sodalite and calcite are common and the sodalite often contains myriads of tiny feldspar inclusions.
The aegirine-augite-nepheline syenite [aegirine foyaite of
Bloomfield et al., (1981)] generally shows a strong foliation caused by aligned pyroxene, and to lesser extent aligned feldspar and nepheline, prisms. Large microcline perthites are often surrounded and cut by swirling areas of finely granular alkali feldspar which may have been produced by granulation and subsequent recrystallization. Nepheline is abundant, generally crowded with tiny pyroxene crystals and what appear to be small blebs of feldspar. The mafic minerals are aegirine-aegirine-augite and minor biotite. Minor minerals are sodalite and cancrinite, probably after nepheline. Accessory phases include titanite, magnetite, pyrite, alkali amphibole and occasional eudialyte.
Aegirine-rich syenite occurs as near-vertical sheets near the summit of Ilomba Hill [pyrochlore pyroxenite and aegirine pegmatite of
Bloomfield et al., (1981)]. These are melanocratic rocks of aegirine with a little alkali feldspar, abundant pyrochlore and titanite, generally distributed in lines or layers, and some apatite, zircon, calcite and amphibole. Betafite, eudialyte and an unidentified rare earth mineral were found in single specimens.
A large pyroxenite body outcrops at the eastern edge of the complex and pyroxenite is also found as small masses within the nepheline syenite in the summit area and north of the summit. The contacts of the large pyroxenite body are not exposed, but xenoliths of pyroxenite are found in the nearby nepheline syenite. The pyroxenites are believed to be part of an earlier basic suite intruded into the Songwe Syenite complex
(Bloomfield et al., 1981). The pyroxenites of the large pyroxenite body contain up to 90% pale pink to green, commonly zoned, diopside, up to 25% interstitial magnetite, tiny green spinels enclosed by magnetite, occasional colorless to pale green calcic amphibole forming ragged areas or rims on pyroxene and rare small brown patches of an unidentified material that may be after olivine. The pyroxenites outcropping within the intrusion are variably altered with the pyroxene becoming a deep green color with margins of amphibole and/or biotite. The intrusion lies 6·5 km east of Ilomba and crops out within the Songwe Syenite. It is a small body (0·8 km * 0·5 km in plan view) which forms a steep-sided hill. Biotite gives the rock a marked foliation which parallels the regional trend
(Ray, 1974). Thus this intrusion, like Ilomba, is pre- or syn-tectonic. The rock is coarse grained, contains two feldspars (microcline > albite), variable amounts of nepheline (occasionally reaching 50%), blue sodalite generally as separate grains but in the summit area also as irregular and discontinuous veins, biotite and minor muscovite in patches, and accessory magnetite and ilmenite. A few euhedral zircons were noted and zoned pyrochlore and bastnäsite were found in one sample.
Pyroxene, amphibole, biotite and nepheline were analyzed by electron beam methods. Both a Hitachi S2500 scanning electron microscope equipped with a Link AN10/55S energy-dispersive X-ray analysis system, operated at 15 kV and 1 nA current on vanadium metal, and a Cameca SX50 wavelength-dispersive electron microprobe, operated at 20 kV and 20 nA measured on a Faraday cage, were used to analyze the minerals.
Approximately 1 kg samples were collected for whole-rock geochemistry. The samples were slabbed and reduced to powder using a fly press and jaw crusher, a rotary pulverizer with ceramic plates and a final size reduction by hand-grinding in a ceramic mortar and pestle.
Major and minor elements were determined as follows: Si, Ti, Al, Fe
(total), Mn, Mg, Ca, Na and K by inductively coupled plasma emission spectrometry, ferrous irontitrimetrically, total carbon and H2O+ using a C, H, N elemental analyzer, F by ion-selective electrode and Cl by silver chloride nephelometry. V, Cr, Ni, Cu, Zn, Rb, Sr, Ba, Y, Nb, Pb, Zr and Ga were determined by X-ray fluorescence (XRF) analysis of pressed powders with a precision and accuracy generally better than 3% for elements determined at concentrations of 10 ppm or higher. Sc, Co, Cs, La, Ce, Nd, Sm, Eu, Gd, Tb, Tm, Yb, Lu, Hf, Ta, Th and U were determined by instrumental neutron activation analysis (INAA). US Geological Survey standard rock AGV-1 is theirradiation standard and the recommended values from
Govindaraju, (1989) are used for this standard. Precision and accuracy were generally better than 5% for most elements except Cs, Gd and Tm, for which precision and accuracy were better than 10%.
Pyroxenes were separated from the pyroxenites (samples ME91, ME92 and ME93). Several thin slabs 2-3 mm thick were cut from each specimen. A jeweler's chisel was used to separate individual pyroxene grains from the surrounding matrix and the edges of the grains were removed. The grains were crushed in an agate mortar and pestle and further purified, to remove any included minerals, using standard electromagnetic and heavy liquid separation procedures. Four samples (0·04-0·13 g) were analyzed by INAA. Major element chemistry was determined by electron microprobe analysis on sections cut from the same specimens. Chemical compositions were determined for a number of minerals in the NNAP nepheline syenites, and some of these data have been reported
(Woolley et al., 1992,
, 1996). In this section we briefly consider the pyroxene and mica chemistry, as these minerals potentially contain information about the chemical evolution of the magmas.
Three groups of pyroxenes have been identified in the NNAP rocks
(Woolley et al., 1996). The first group consists of alkali pyroxenes ranging from diopside to aegirine with low total Al and an inferred igneous origin. The second group, distinguished by its high Alvi content, is interpreted as having a metamorphic origin. The third group, probably of igneous origin, consists of Al-rich pyroxenes in which Aliv is predominant and there is no trend of alkali enrichment. The data for all three groups are plotted in terms of Mg-Na-(Fe2+ + Mn) in
Fig. 2. In this plot there is no discernible difference between groups 1 and 2; group 3 (from the pyroxenites) forms a distinct cluster. Pyroxenes from Kasungu-Chipala show a limited range in chemical composition from diopside to aegirine-augite. The Ilomba pyroxenes show a much greater compositional range, from diopside to aegirine, reflecting the more extensive fractionation of the Ilomba magma. The Ilomba pyroxenes also show a much greater variability than that seen for pyroxenes from other nepheline syenites, and this is probably due to the metamorphic overprint. The compositional trend shown by the NNAP pyroxenes is similar to that for pyroxenes in nepheline syenites from Tenerife, Itapirapuã, and the recrystallized syenites from the South Qôroq Center (
Fig. 2).
Biotite occurs in all the intrusions and representative analyses are listed in
Table 2. Biotites from the biotite-nepheline syenite at Ilomba tend to be richer in Al than biotites from the other nepheline syenites at Ilomba and Kasungu-Chipala. Biotites from the nepheline syenites at Ulindi show similar elevated Al contents (
Fig. 3). These chemical distinctions also hold for other elements, i.e. higher Mn and lower Ti in the Ilomba and Ulindi syenites. Thus the elevated Al content is probably not due to metamorphism, as was the case for the pyroxenes, but rather reflects magma chemistry. The Ilomba and Ulindi biotites also have high Al content when compared with biotites from other nepheline syenites (
Fig. 3). Table 2. Representative analyses of mica.
Representative major and trace element whole-rock analyses are listed in
Table 3. Selected major elements are plotted vs SiO2 in
Fig. 4. The nepheline syenites of the Ilomba-Ulindi intrusions have significantly lower concentrations of TiO2, FeO(total) (not shown), MnO (not shown), MgO, and CaO (not shown), and significantly higher Al2O3 and Na2O + K2O than those of Kasungu-Chipala. The chemistry thus reflects the generally low mafic mineral content of the Ilomba and Ulindi nepheline syenites compared with those of Kasungu-Chipala. The Ilomba-Ulindi nepheline syenites have significantly lower F content than those of Kasungu-Chipala (
Fig. 5), and a number of the Ilomba nepheline syenites are Cl rich relative to F. Kasungu and Chipala are slightly different in terms of major element (and trace element,
Fig. 6) chemistry, suggesting slight differences in the compositions of the initial magmas. Table 3. Representative whole-rock analyses.
Selected trace elements are plotted vs SiO2 in
Fig. 6. Compared with the Kasungu-Chipala nepheline syenites, the Ilomba-Ulindi nepheline syenites are low in Sc. Rb is enriched in the Ilomba-Ulindi nepheline syenites compared with the Kasungu-Chipala nepheline syenites. The Ba and Sr contents of the Ulindi nepheline syenites are low relative to both the Kasungu-Chipala nepheline syenites and most of the Ilomba nepheline syenites.
Nb/Ta ratios for the Ulindi nepheline syenites vary from 25 to 35, and for the Ilomba nepheline syenites from 35 to 98 (
Fig. 7). These variations can be compared with the restricted range of Nb/Ta ratios (12-15) shown by the Kasungu-Chipala nepheline syenites (Nb/Ta = 13·1 ± 0·7) and mafic dikes (Nb/Ta = 13·8 ± 0·3).
For the Ilomba and Ulindi nepheline syenites Zr/Hf ratios show significant variation (generally Zr/Hf = 30-60,
Fig. 7), compared with the limited range shown by the Kasungu-Chipala nepheline syenites (Zr/Hf = 37 ± 3).
Green, (1995) reported similar ranges in Zr/Hf (38-52) for other alkaline suites.
For the Kasungu-Chipala nepheline syenites Th/U = 3·5 ± 0·6. Some of the Ilomba-Ulindi nepheline syenites have similar ratios, but many have much lower ratios. In the case of the Ulindi nepheline syenites there is an apparent trend towards lower Th/U ratios with increasing U (
Fig. 7), which suggests that the ratio is largely controlled by variations in the U concentration.
Chondrite-normalized rare earth element (REE) patterns for Kasungu-Chipala (
Fig. 8) show significant light REE (LREE) enrichment and tend to flatten at the heavy REE (HREE) end. (Ce/Yb)N ratios vary from 7·5 to 11. Chondrite-normalized REE patterns for Ilomba (
Fig. 8) and Ulindi (
Fig. 9) are generally slightly to significantly concave-up. A number of the patterns show small positive Eu anomalies which may be indicative of feldspar accumulation. For most of the Ilomba samples the (Ce/Yb)N ratio is <10, whereas the (Ce/Yb)N ratios vary from 10 to 44 for the Ulindi syenites. REE patterns similar to those of the Ulindi nepheline syenites have been previously reported for nepheline syenite dikes at Mount Royal in the Monteregian Hills
(Eby, 1984) and for peralkaline granites of the Oslo Rift, Norway
(Neumann et al., 1977).
INAA data for pyroxenes from the pyroxenite body at Ilomba are reported in
Table 4. Representative major element analyses for these pyroxenes have been given by
Woolley et al., (1996). The major element chemistry of the pyroxenes can be used to calculate some of the chemical properties of the melt from which the pyroxenes crystallized. On the basis of experimental and empirical work (see, e.g.
Grove & Bryan, 1983;
Eby, 1984;
Elthon & Casey, 1985;
Blundy et al., 1995), it is possible to calculate the TiO2 and Na2O content of the melt which coexisted with the pyroxenes. In these calculations the following Dcpx/melt values were used: Dcpx/melt = 0·50 for TiO2 (low Ti-Al-clinopyroxene in equilibrium with basanitic melts at low pressure;
Adam & Green, 1994) and Dcpx/melt = 0·15 for Na2O
(Kinzler & Grove, 1992). For the melt which coexisted with the pyroxene cores, TiO2 = 1·5-2·5 wt % and Na2O = 2·8-4·6 wt %. The Mg/Fe ratio of the pyroxenes can be used to calculate the Mg/(Mg + Fe2+) ratio (mg-number) of the melt.
Eby, (1984) used Monteregian pyroxene-olivine pairs to arrive at an empirical KD of 0·23 and
Grove & Bryan, (1983) experimentally arrived at the same KD value for the ratio (XMg/XFe2+)liq/(XMg/XFe2+)pyx. For the melt which coexisted with the pyroxenes mg-number varies from 65 to 78. These values can be compared with an mg-number of 74 for melts in equilibrium with mantle olivine of Fo90 composition. As there is a regular relationship between the mg-number and the TiO2 and Na2O content of the calculated melts it is possible to estimate these values for any particular mg-number. At an mg-number of 74, the coexisting melt would have had 1·8 wt % TiO2 and 3·1 wt % Na2O. Table 4. Trace element abundances for pyroxenes and calculated liquids.
The trace element compositions of melts coexisting with the pyroxenes can also be determined given appropriate partition coefficients.
Eby, (1984) empirically determined partition coefficients using clinopyroxene-matrix pairs from Monteregian Hills mafic dikes. Although the data set for the Monteregian Hills is internally consistent, comparison with other estimates of Dcpx/melt values [see review by
Green, (1994)] suggests that, relative to the middle REE (MREE) and HREE, the partition coefficients for the LREE were overestimated. Therefore, the REE partition coefficients have been recalculated according to the method of
Wood & Blundy, (1997). The Dcpx/melt for Sm is set equal to 0·36 (the empirically determined value for the Monteregian Hills dikes) and the other REE partition coefficients are calculated for a melt at 0·1 GPa and 1100°C. The Hf partition coefficient was set equal to 0·35, a value determined experimentally by
Dunn, (1987). A Cr partition coefficient was not determined for the Monteregian Hills pyroxenes. The Cr partition coefficient used here is from
Hart & Dunn, (1993). The results of these calculations, and the Dcpx/melt values used in the calculations are given in
Table 4. The compositions of the Kasungu-Chipala, Ilomba and Ulindi nepheline syenites are projected into the SiO2-NaAlSiO4-KAlSiO4 phase diagram in
Fig. 10. The Kasungu-Chipala nepheline syenites plot as a tight cluster on the low-pressure side of the 1 kbar minimum. As these are relatively mafic nepheline syenites, only 70-80% normative feldspar + foids, P-T conditions of crystallization should not be inferred from the phase diagram. The Ulindi nepheline syenites (and a few aegirine-nepheline syenites from Ilomba) plot near the 1 kbar minimum, suggesting that they represent fractionated liquids. Such a conclusion is in agreement with the trace element data which indicate significant pyroxene (low Sc, Co), feldspar (low Ba and relatively low Sr), and magnetite (low V, Co) fractionation. Extensive fractionation is also suggested by the high Rb abundances. The Ilomba biotite-nepheline syenites plot along the 1 kbar cotectic and extend towards the composition of the nephelines from the Ilomba nepheline syenites. These data are most reasonably interpreted as representing nepheline accumulation (and feldspar accumulation as indicated by high K2O). This conclusion is qualitatively supported by the high Ba and Sr concentrations, consistent with the inference of feldspar accumulation and the relatively higher (with respect to Ulindi) concentrations of V and Co, which suggest magnetite accumulation. In combination, the phase equilibria and trace element data suggest that the chemical composition of the nepheline syenites from Kasungu, Chipala and Ulindi may approximate magma compositions, whereas most Ilomba samples represent solids separated from magma.
Chemical data for the pyroxenes from the various NNAP intrusions reflect the overprint of a later metamorphic event
(Woolley et al., 1996) related to the Pan-African orogeny. In addition, there is clear evidence of hydrothermal activity at Ilomba which led to concentrations of niobium and uranium for which this pluton was prospected
(Bloomfield et al., 1981). Thus the present geochemistry of the plutons is the end product of a number of processes, magmatic, hydrothermal and metamorphic.
For the Kasungu-Chipala nepheline syenites, the high field strength element (HFSE) distributions are regular, and Nb/Ta, Zr/Hf and Th/U (
Fig. 7) ratios are relatively constant. Thus, in the case of Kasungu-Chipala, the later metamorphic event apparently did not disturb the HFSE distributions. Conversely, for Ulindi and particularly Ilomba there is a wide variation in these elemental ratios. We see no textural or mineralogical evidence to suggest that the metamorphic event at Ilomba-Ulindi was of a different character from that at Kasungu-Chipala. Thus we conclude that the variations in HFSE cation ratios were not caused by metamorphism.
Nb/Ta variations
Green, (1995) noted that in alkaline suites Nb/Ta ratios tend to fall into two groups. One group is characterized by consistent Nb/Ta ratios close to chondritic or mantle values whereas the other comprises highly variable Nb/Ta ratios significantly greater than mantle values. For example, a suite of lavas from Mount Erebus, ranging in composition from basanite to anorthoclase phonolite, varies in Nb/Ta ratio from 16·0 to 18·6 with a mean of 17·2 ± 0·5
(Kyle et al., 1992). Similar limited ranges in Nb/Ta ratios and values near chondritic (or mantle) have been reported for both oceanic
(Palacz & Saunders, 1986;
Weaver, 1990) and continental
(Briggs et al., 1990;
Rogers
et al., 1992) alkaline suites. In contrast, high and variable Nb/Ta ratios have been reported for the Glass House Mountains suite in the Tertiary volcanic province of Eastern Australia [Nb/Ta ranges from 11-12 in trachyte to 35 in comendite
(Ewart et al., 1986)], Tenerife [Nb/Ta = 9-24
(Wolff, 1984)], the South Atlantic province [Nb/Ta = 15-68
(Weaver, 1990)] and Brava, Cape Verde Islands [Nb/Ta = 17-56 (J. A. Wolff, personal communication, 1997)]. Kasungu-Chipala falls into the first group, and Ilomba and Ulindi fall into the second.
Green & Pearson, (1987) investigated experimentally the partitioning of Nb and Ta between silicate liquids and a number of Ti-rich minerals (titanite, rutile, ilmenite and Ti-magnetite). With the exception of magnetite, all of these minerals have partition coefficients significantly greater than unity for both elements. Both rutile and titanite favor Ta with respect to Nb, and fractionation of these minerals would lead to an increase in the Nb/Ta ratio of the magma.
Green & Pearson, (1987) evaluated the effect of such fractionation and came to the conclusion that increases of the order of 20% in the Nb/Ta ratio might occur by fractional crystallization. Their experiments did not include silica-undersaturated liquids, and data obtained for glassy volcanic rocks (e.g.
Weaver, 1990) suggest that titanite fractionation can lead to Nb/Ta ratios of 50-60. Thus in the case of Ulindi titanite fractionation could lead to the observed high Nb/Ta ratios. Of note is the low absolute abundance of Ta in the Ulindi rocks (compared with Kasungu-Chipala), which are inferred to represent highly fractionated liquids. Fractionation of titanite would lead to a decrease in Ta abundances. Titanite was not found in the Ulindi samples, so if titanite fractionation did occur it pre-dated the emplacement of the Ulindi magma.
Nb/Ta ratios for Ilomba vary from 35 to 98. Titanite from a nepheline syenite southwest of the summit of Ilomba has been analyzed for Nb and Ta
(Woolley et al., 1992). The partition coefficients determined by
Green & Pearson, (1987) are used to calculate the Nb and Ta content of a coexisting silicate liquid. As these workers did not determine partition coefficients for silica-undersaturated liquids we used partition coefficients determined for a trachyte at 1000°C and 4 kbar. The difference between trachytic and phonolitic liquids, particularly in terms of the Nb/Ta ratio, is not expected to be large. The calculated Nb/Ta ratios for the silicate liquids in equilibrium with the Ilomba titanites vary from 82 to 196. These high ratios exceed those that have been observed for volcanic rocks.
Vard & Williams-Jones, (1993) investigated inclusions in vug minerals found in phonolite sills. They concluded that vug filling was the result of mineral precipitation from a primary orthomagmatic fluid and that this fluid was enriched in HFSE ions, particularly Zr, Hf, Nb and Ti. Thus it is possible for phonolite magmas to evolve HFSE-rich fluids. This observation, coupled with the presence of Nb hydrothermal mineralization at Ilomba, suggests that the most likely explanation for the high Nb/Ta ratios is the introduction of Nb to the rocks during late-stage hydrothermal processes.
Th/U
Experimental studies
(Romberger, 1984;
Keppler & Wyllie, 1991;
Peiffert
et al., 1996) have shown that under hydrothermal to magmatic conditions U can form complexes with F, Cl and CO2 in the aqueous phase. Increasing concentration of any of these components leads to increased transport of U in the aqueous phase. Although U content is not correlated with the observed F, Cl and/or CO2 content of the nepheline syenites, this does not preclude open system transport of this element by a fluid phase. Particularly in the case of the Ulindi syenites there is a well-developed trend of decreasing Th/U ratio with increasing U content, a pattern consistent with U mobility.
REE
The concave-up shape of the REE patterns is unusual. As noted above, similar patterns have been reported for nepheline syenites and peralkaline granites, but the origin of this type of pattern in an evolved igneous rock is enigmatic. Several possibilities are considered.
Titanite fractionation. Both empirically (titanite-matrix,
Wörner et al., 1983) and experimentally
(Green & Pearson, 1983) determined partition coefficients for titanite show a strong preference of this mineral for the MREE.
Wolff & Storey, (1984) concluded that the fractionation of titanite from Tenerife phonolite magmas could explain the depletion of MREE in Tenerife pumice falls. Thus titanite fractionation could lead to a relative decrease in the abundance of the MREE, and a concave-up pattern, a result consistent with the increase in the Nb/Ta ratios for Ilomba and Ulindi.
Zircon accumulation. Zircon has very high partition coefficients for the HREE relative to the LREE
(Fujimaki, 1986). Fractionation of this mineral would lead to significant depletion of HREE in the melt, and conversely accumulation of zircon would lead to enrichment of the rock in HREE. From Eu to Lu, the partition coefficients increase rapidly so that a combination of an LREE-enriched magma and zircon accumulation can successfully explain concave-up REE patterns. According to the model of
Watson, (1979), a magma having the composition of the Ulindi nepheline syenites would be saturated in zircon. However, petrographic observation reveals that zircon is rare in the Ulindi nepheline syenites so zircon accumulation would not seem to be an important process.
Fluid transport. A fairly extensive literature now exists regarding the partitioning of the REEs into fluid phases under hydrothermal and magmatic conditions (e.g.
Cullers et al., 1973;
Flynn & Burnham, 1978;
Langmuir, 1978;
Taylor et al., 1981;
Bau, 1991;
Terakado et al., 1993). In general, the REEs are partitioned into the silicate-melt phase relative to the fluid phase. Fluid-melt partition coefficients are lowest for H2O-only fluids and increase with increasing concentrations of Cl, F and CO2 in the fluid phase. In the case of fluids containing Cl, there is generally little fractionation of the REEs, except over a limited Cl concentration range
(Terakado et al., 1993) in which there is some enrichment of HREE relative to LREE in the fluid phase. F-- and CO32--bearing fluids are more effective in transporting the REEs as fluoride and carbonate complexes. In these systems the HREE are enriched relative to the LREE in the fluid phase. Carbonate complexes of the HREE and U are stable under relatively alkaline and oxidizing conditions
(Langmuir, 1978). It has already been noted that in the case of the Ulindi nepheline syenites in particular there is evidence of U mobility, probably as a result of hydrothermal processes. There is a relationship between the U content of the rocks and the degree of concavity of the REE patterns as measured by the (Tb/Yb)N ratios (
Fig. 11). The (Tb/Yb)N ratio decreases as U increases, which would be in agreement with the transport of the HREE and U as carbonate or fluoride complexes.
Summary
Using the apparently undisturbed Kasungu-Chipala trace element distributions as a reference, the Ilomba and Ulindi rocks are characterized by relatively high Nb/Ta ratios and a relative depletion in the MREE. Qualitatively these observations can be largely explained by fractional crystallization of titanite, a mineral which is common in most of the NNAP plutons. Titanite was not found at Ulindi, but these rocks are chemically the most evolved in the NNAP and have very low Ca and Ti abundances. It is not unreasonable, therefore, to assume that titanite fractionation occurred earlier in the evolution of the Ulindi magma. However, the very high Nb/Ta ratios obtained for some of the Ilomba samples are probably due to the hydrothermal addition of Nb, and Nb mineralization occurs at Ilomba. For Ulindi we would suggest that hydrothermal fluids played a secondary role in the modification of REE patterns and U distributions. LREE patterns for Ulindi (
Fig. 9) are parallel but the patterns cross at the HREE end. Titanite fractionation will not explain the crossovers, but a secondary effect caused by the addition of HREE by hydrothermal fluids will, and this is consistent with the observation made previously concerning the inverse relationship between U concentration and the (Tb/Yb)N ratio.
The NNAP is typical of a number of alkaline provinces (such as the spatially related Chilwa Alkaline Province) in which felsic rocks form the dominant lithologies. In the NNAP the only exposed mafic rocks are the pyroxenites at Ilomba and occasional mafic dikes cross-cutting the felsic units in the other plutons. Does the absence of mafic rocks imply that the felsic magmas were the direct product of melting of metasomatized(?) lithosphere, or did differentiation occur at depth with only the low-density fractionates intruded to high level? The chemical data acquired for the NNAP provide some petrogenetic constraints, but a unique answer is still elusive.
Kasungu-Chipala
The Kasungu-Chipala intrusions consist largely of mafic nepheline syenite, and represent the least evolved silica-undersaturated syenites in the NNAP. Two nepheline syenite types (one-feldspar, two-feldspar) have been identified at both intrusions. In terms of rock chemistry these two types are indistinguishable. The REE patterns for all of the samples are similar (
Fig. 8), with the mafic dikes having the lowest total REE abundances. None of the samples have Eu anomalies, which suggests that plagioclase fractionation did not play an important role in the evolution of the Kasungu-Chipala magma(s). Thus it is concluded that only minor crystal-liquid fractionation, involving pyroxene and plagioclase, occurred during emplacement of the magmas, as indicated by slight decreases in MgO and Sr with increasing SiO2.
For Kasungu-Chipala Nb/Ta (13·1), Zr/Hf (37) and Th/U (3·5) ratios are within the range of values observed for mantle-derived melts. In addition, on the Y/Nb vs Yb/Ta plot (not shown) of
Eby, (1990) the Kasungu-Chipala nepheline syenites plot in the oceanic-island basalt (OIB) field. Except for the Th/U (4·1, 2·4) ratio the mafic dikes have similar Nb/Ta (14·0, 13·6) and Zr/Hf (43, 37) ratios to the nepheline syenites and plot with them in the OIB field. The data are consistent with the interpretation that the Kasungu-Chipala magma(s) were derived from an OIB-like source and that the initial melt(s) underwent high-pressure (plagioclase-absent) fractionation before emplacement at high level. As suggested above, there may have been minor pyroxene and plagioclase fractionation during the emplacement of themagmas.
Ilomba-Ulindi
The Ilomba sodalite-nepheline syenites and biotite-nepheline syenites, on the basis of their phase chemistry (
Fig. 10), high Ba and Sr contents (
Fig. 6), and low total REE abundances, with many patterns showing small positive Eu anomalies (
Fig. 8), are inferred to be the result of crystal accumulation. The aegirine-nepheline syenites plot between the nepheline-feldspar mixing line (
Fig. 10) and the composition of an undersaturated liquid at the thermal minimum at 1 kbar, and thus may represent magma-crystal mixtures. All of the Ulindi samples plot in a tight cluster around the 1 kbar thermal minimum (
Fig. 10), and presumably approximate liquid compositions. The Ulindi samples have high Al2O3 (
Fig. 4), which is also reflected by the biotite chemistry (
Fig. 3), and a similarly high Al content is observed for many of the biotites from Ilomba. The Ulindi samples are also characterized by high Rb and low Ba and Sr (
Fig. 6). The trace element data could be interpreted as indicating substantial feldspar fractionation from the Ulindi magma. However, negative Eu anomalies are not observed for Ulindi (
Fig. 8) and the only observed Eu anomalies are slightly positive. One possible conclusion is that the observed high Al and Rb contents, and low Sr and Ba contents, are a characteristic of the initial Ulindi magma. The chemistry of this initial melt would have been significantly different from that of the melt which gave rise to the Kasungu-Chipala intrusions.
Using the composition of pyroxenes from the pyroxenites at Ilomba, it is possible to calculate a partial composition for the melt from which the pyroxenes crystallized. The TiO2 and Na2O concentrations for this melt are similar to those of average basanite and nephelinite
(LeBas, 1989). In comparison with tholeiitic and alkali olivine basalts, the absolute concentrations of these two oxides, at similar mg-number, are significantly higher than those observed in either of these two basalt groups. Thus basanitic-nephelinitic magmas were apparently present at the time of intrusion of the Ilomba and Ulindi plutons and could have served as mafic precursors for the nepheline syenite magmas.
Calculated Sc (19-30 ppm), Co (42-56 ppm) and Cr (240-470 ppm) values for the melt from which the pyroxenites crystallized are similar to those for OIBs with mg-numbers in the range 65-74 (G. N. Eby, unpublished data, 1997). The Yb/Ta ratios (0·25-0·72) for the calculated liquids are in the same range as those for basalts derived from OIB sources
(Eby, 1990). The Sm/Hf ratios (1·11-1·21) fall at the low end of the range for OIBs. It is inferred, therefore, that this magma was derived from an OIB-like source.
The REE data for the pyroxenes and the calculated coexisting melt are plotted in
Fig. 12. The calculated patterns have a steep slope and high (Ce/Yb)N ratios (31-40). Also plotted in
Fig. 12 are similarly calculated REE patterns for Monteregian Hills magmas, which were inferred to represent partial melts of a garnet lherzolite source
(Eby, 1984). The steeper patterns of the Ilomba pyroxenites could be interpreted as the result of even smaller amounts of partial melting of a garnet lherzolite.
The Nachendezwaya carbonatite complex crops out 2 km north of Ilomba on the Tanzania side of the Songwe river. This carbonatite complex has not been dated, but
Gittins, (1966) inferred that it was the same age as Ilomba. Thus there may have been carbonatitic magmas spatially and temporally associated with the Ilomba magmas.
Nelson et al., (1988) reported chemical and isotopic data for a sövite from the Nachendezwaya complex. The Sr and Nd isotopic ratios are typical of a depleted mantle source and fall in the same range as those of other carbonatites. However, in terms of elemental chemistry, this sample is anomalous in several respects when compared with other carbonatites. First, it has a low Nb content relative to Ta and thus an anomalously small Nb/Ta ratio (~2), which is confirmed by additional analyses of sövite from the complex (G. N. Eby, unpublished data, 1997). Second, it has a very low Zr/Hf ratio (~6). The foyaites and ijolites associated with the Nachendezwaya carbonatite have Nb/Ta (15·5) and Zr/Hf (53) ratios (G. N. Eby, unpublished data, 1997) typical of silica-undersaturated alkaline rocks.
Some workers (e.g.
Rudnick et al., 1993;
Green, 1995) have suggested that carbonatitic metasomatism of the mantle may play a role in the generation of alkaline (OIB-like) magmas. Data from xenoliths indicate that carbonatitic metasomatism would lead to an increase in the Zr/Hf ratio of the mantle source
(Rudnick et al., 1993), and
Green, (1995) has also suggested that an increase in the Nb/Ta ratio might be expected. In particular, the high Zr/Hf and Nb/Ta ratios shown by many of the Ilomba samples might be a reflection of this type of process. However, the very low Nb/Ta and Zr/Hf ratios for the Nachendezwaya carbonatite, if they are representative of the composition of the carbonatitic melt at depth, would seem to preclude this possibility.
A petrogenetic model
It has been proposed on the basis of experimental
(Hay & Wendlandt, 1995) and geochemical
(Hay et al., 1995) studies that Kenya rift plateau-type flood phonolites originated by partial melting of alkali basaltic material at lower-crustal pressures. In terms of major element chemistry the Kasungu-Chipala and Ulindi nepheline syenites are similar to the Kenya phonolites. The REE patterns for the least evolved phonolites resemble those of Kasungu-Chipala both in terms of total REE abundances and pattern shape, but the Ulindi nepheline syenites have distinctly different REE patterns. However, Nb/Ta (21·6), Zr/Hf (55) and U/Th (2·5) ratios for the Kenyan phonolites are significantly different from those of Kasungu-Chipala. On the basis of the current geochemical data, partial fusion of an alkali basalt at lower-crustal pressures might produce a magma broadly similar to that of Kasungu-Chipala, but seems unlikely to produce a magma similar to that of Ulindi. We also note that unlike the Kenyan phonolites, there is no evidence of an earlier period of alkali basalt magmatism that injected and/or underplated the lower crust in Malawi. Thus, although we cannot rule out a lower-crustal fusion model, it is suggested that such a model is not the most likely one for NNAP magmatism.
Kasungu-Chipala nepheline syenites and mafic dikes have REE patterns that show moderate LREE enrichment and a flattening at the HREE end. The similarity to REE patterns produced by the partial melting of an alkali basalt has been noted above. The Kasungu-Chipala nepheline syenites and mafic dikes also have trace element concentrations and elemental ratios typical of alkali basalts derived from an OIB-like source. These observations are consistent with the derivation of the Kasungu-Chipala magmas by high-pressure (feldspar absent) fractionation of an alkali basalt magma. This magma ponded near or at the base of the crust until differentiation produced a magma of sufficiently low density to allow its ascent through the crust.
The emplacement of the Ilomba and Ulindi complexes is spatially, and perhaps temporally, associated with the emplacement of the Nachendezwaya carbonatite complex. The pyroxenites at Ilomba are apparently cumulates from a basanitic-nephelinitic magma, and nephelinites are often associated with carbonatite magmatism. The Ulindi nepheline syenites are significantly enriched in Al, show strong LREE enrichment, a relative enrichment in HREE with respect to MREE, and no negative Eu anomalies. In terms of trace element geochemistry, it is possible that the Ulindi syenites were derived from a basanitic-nephelinitic magma through fractionation of nepheline, pyroxene, and titanite. Titanite seems the most likely choice to explain both the relative depletion of MREE and the increase in Nb/Ta ratio. The non-involvement of feldspar in the fractionation process is supported by the absence of negative Eu anomalies and the high Al content of the Ulindi nepheline syenites. The implications of this observation are that either the source lacked significant feldspar or crystal fractionation occurred at high pressure outside the plagioclase stability field. The strongly LREE-enriched REE pattern calculated for the basanitic-nephelinitic magma suggests that the melt equilibrated with a garnet-bearing residue, a conclusion that also is supported by the low Sc content.
In the case of both Ulindi and Kasungu-Chipala it appears possible that the nepheline syenites were generated by crystal fractionation at high pressure from more mafic magmas. The precursor magmas were different, however, alkali basalt for Kasungu-Chipala and basanitic-nephelinitic for Ulindi, suggesting different degrees of partial melting and perhaps different depths of melting. In the case of Kasungu-Chipala in particular the trace element data are typical of an OIB-like source. Whether this source was metasomatized continental lithosphere, as suggested for the Monteregian Hills
(Eby, 1984) and Chilwa Alkaline Province
(Woolley & Jones, 1987), or asthenosphere is still unresolved. Mr G. Jones (Natural History Museum) and Mr K. Parish (Oxford) assisted with the analytical work. The XRF analyses were done at the University of Oxford and the INAA was done at the University of Massachusetts Lowell Radiation Laboratory. T. Andersen, B. J. G. Upton and J. Wolff provided constructive and helpful reviews of the manuscript. Partial support for this project was provided by a NATO research grant (No. 880038).
INTRODUCTION
GENERAL GEOLOGY AND PETROGRAPHY
Kasungu-Chipala
Ilomba
Ulindi
GEOCHEMISTRY
Analytical methods
Mineral chemistry
Whole-rock geochemistry
Pyroxene major and trace-element geochemistry
DISCUSSION
Conditions of crystallization
Elemental variations
Petrogenetic considerations
ACKNOWLEDGEMENTS
REFERENCES