Geology and origin of the Kimbo ruby deposit
(The John Saul Mine, SE Kenya)
(The John Saul Mine, SE Kenya)
1. Introduction
The region of Southern Kenya that extends between the Taita Hills to the north and the Umba valley to the south has become for a few decades a major gemstone production area (e.g. Keller, 1992). Numerous ruby and sapphire deposits, of a great geological variety, are present (Simonet, 2000a). In the heart of this region, the ruby deposit known as the John Saul Mine, one of the earliest found, is now the largest ruby mine in Africa (Emmett, 1999). Although this deposit and the neighbouring mines of the Mangare area have been described in gemmological litterature, few have been studied in detail. The John Saul Mine is the deposit that has been the most under the attention of geologists (e.g. Pohl, 1985; Key and Ochieng, 1991; Mercier et al, 1999a). The work of Key and Ochieng (1991) regards a mineralization that was exploited during until 1995, and that is currently exhausted. The study of Mercier et al (1999a), that concerns the whole of the Mangare area ruby deposits, is aimed at understanding the mineralisation's origin, and is more instructive. In this paper, we focus on one of the corundum-bearing veins present at the John Saul Mine, known as the Kimbo vein, which has been exploited since 1995.
2. Regional geology
The John Saul Mine is located at E 38°30'00" and S 3°51'30". The deposit is part of the Mozambique Belt, which Pan African evolution has been reviewed by Key et al (1989) and more recently by Mosley (1993). To this orogen are associated most of Kenya, Tanzania and Madagascar gemstone deposits (Dissanayake and Chandrajith, 1999). In Kenya, Mosley (1993) divides the Mozambique Belt into three main zones separated by major shear zones having a NNW orientation. The western zone, in which is located the John Saul Mine, is characterised by W to NW verging thrust sequences. Metamorphism culminated in granulite facies (4 to 10 kbar and 500 to 700 C), with a high PCO2 underlined by the ubiquity of graphite. P-T paths are clockwise owing to crustal thickening resulting from the mozambiquian collision.
Key et al (1989) and Mosley (1993) distinguish several major phases in the post-kibarian evolution of the Mozambique Belt in Kenya, linked to the oblique collision between two east and west fragments of Gondwana (Burke and Dewey, 1972; Key et al, 1989; Kröner, 1991; Mosley, 1993). Before 820 My erosion and crustal extension followed the Kibaran orogeny, and limited construction of oceanic crust occured in suprasubduction basins. The Samburuan-Sabachian event, around 820 My, was the result of the oblique collision between the two Gondwana fragments, and was accompanied by metamorphism reaching granulite facies and by the production of cylindrical folds, later tightened. The thermal peak was followed by the developpement of major recumbent folds and by low angle thrusting. Around 620 My, the Baragoian event was linked to ENE-WSW compression across the orogen. It is characterised by the appearance of sinistral shear zones, narrow vertical structures filled with felsic veins and ultramafic boudins. Metamorphism reached amphibolite facies conditions. The Barsaloian event, at around 580 My, saw the setting of major N-S shear zones, 20 km wide and several hundred of kilometers long, inside which older fabrics were transposed and rocks underwent partial melting. Most of these shear zones are dextral. Associated metamorphism was of amphibolite facies (6.8 kbar and 570-590 C- Suwa, 1981). A phase of exhumation and cooling then occured between 530 and 470 My (Kipsigian Event of Hetzel and Strecker, 1994), characterised by NE-SW folds and by sinistral SSE- NNW shear zones, with associated green schist facies metamorphism.
In the part of Kenya where the John Saul Mine is located, two major lithosratigraphic units have been defined by Saggerson (1962), and redefined by Pohl and Niedermayr (1979). The Kasigau Group, to the north east, is mostly constituted of monotonous quartzofeldspathic gneisses and orthoamphibolites. The Kurase Group, to the south west, is a serie of variegated metasediments including marbles, psammitic, semi-pelitic and pelitic gneisses. These rocks often contain appreciable amounts of graphite of organic origin (Arneth et al, 1983). These metasediments locally host mafic and ultramafic lenses, and are the direct host of most gemstone mineralisations in Southern Kenya, including the John Saul Mine. Mosley (1993) regards the Kasigau Group as a tectonic slice, which contact with the Kurase Group is marked in the Taita Hills by a string of ultramafic lenses (Horkel et al, 1979). However south of the Taita Hills the contact between the two units is not exposed. Granulites and charnockites are locally included in the Kurase Group, and have been considered by Pohl and Niedermayr (1979) as premozambiquian basement included into the metasediments during the mozambiquian collision.
Both units have undergone a amphibolite to granulite facies metamorphism and contain rocks with kyanite and sillimanite. Although the later is usually posterior to kyanite, no clear isograde exist, but rather a coexistence zone (Saggerson, 1962). Most rocks which chemical composition allows it are partially mobilised and show signs of anatexis or K-feldspar porphyroblasts. Pohl and Niedermayr (1979) consider that the main metamorphism phase, associated to paramigmatitic folds has reached P- T conditions of 7 kbar and 700C (sub facies almandine-sillimanite-muscovite and almandine-kyanite- muscovite). Key and Hill (1991) suggest that V-grossular ("tsavorite") bearing rocks of the Kurase Group have undergone granulite facies metamorphism. Mercier et al (1999a) have estimated to 5-6 kbar and 600-700 C the PT conditions undergone by rocks hosting ultramafic bodies of the Mangare area.
3. Local geology
In the Mangare area ruby deposits, including the John Saul Mine, are closely associated to small ultramafic lenses or bodies hosted by the Kurase Group. These ultramafites belong to type 3 ultramafites of Mercier et al (1999a), defined by these authors for carbonatised ultramafites located within the Kurase Group, and to the "ultramafic schists" group of Mosley (1993). The term of "schist" is somehow inadequate, in so far as some of these bodies are not foliated.The origin of these rocks is difficult to ascertain owing to the high degree of metamorphic and metasomatic alteration, and to their mode of occurrence, i.e. disrupted boudins and isolated bodies. Some of them are closely associated to mafic rocks (garnet-bearing amphibolites), as for example at Kalalani and Umba in nearby Tanzania (Seifert and Hyrsl, 1999). The length of these bodies varies from a few metres to more than one kilometre. The ultramafic body of the John Saul Mine to which the ruby mineralisation is linked has an ovoid shape. It is about 150 m long along the direction of the local strike, and 100 m across it, at surface level .
The ultramafic body of the John Saul Mine consists mostly of magnesian silicates and carbonates among which enstatite, talc and magnesite predominate. Dolomite, serpentineous minerals, tremolite, anthophyllite, calcedony, vermiculite and heamatite are also present in various proportions. This type of mineral composition is similar to that of sagvandites, ultramafic rocks described in Norway by Ohnmacht (1974). Enstatite (XMg = 0.9-0.92) is the main constituant of the rock and appears as centimetre size crystals. Its cleavages and cracks are filled up by talc lamellae. The rock is homogeneous and not foliated, expect in the eastern part of the ultramafic body, along its contact with its host rock. There a foliation marked by calcedony and serpentine veinlets can be observed.
Until recently, the origin of Southern Kenya ultramafites has only be the subject of suppositions. Bridges (1982) describes these bodies as "pipes" similar to kimberlitic pipes. This comparison is unfortunate as these bodies are not rooted (Mercier et al, 1999a). The high Cr and Ni content of these rocks show that they are of igneous origin, and not Mg-rich metasediments (Pohl and Niedermayr, 1979). Key and Ochieng (1991) use the term of "ultramafic intrusions" whithout specifying the way of intrusion. Prochaska and Pohl (1983) consider them as tectonic intrusions of sub-continental mantellic material, included in the metasediments along tectonic accidents, and not as rocks of ophiolitic origin. Mercier et al (1999a) consider that the ultramafites of the Mangare area are allochtonous and were included as tectonic slices into the metasediments of the Kurase Group during the first major tectonic phase of the collision. Pohl et al (1980) suggest that the ultramafites may be ophiolitic relicts, and Frisch and Pohl (1986) liken them to serpentinites of the Kinyiki Hill ophiolitic complex, located within the Kasigau Group, thus suggesting that they may be of ophiolitic origin. Further more, these authors have identified an oceanic basalt geochemical signature in amphibolites of the Mwatate Formation, a sub-unit of the Kurase Group.
Ultramafic bodies of the Mangare area have a complex metamorphic and metasomatic history, including a serpentinisation (hydration episode), and a later silicification and carbonatation episode (Simonet, 2000). Most authors agree on the fact that they were emplaced early in the history of the Mozambique Belt. The highly disrupted mode of occurrence (lenses, boudins...) supports this idea, but prevents to delineate any suture. Geochemistry of major and trace elements do not yield much information on their origin, because of the high degree of metasomatic alteration. However the ultramafic body of the John Saul Mine shows a marked HREE depletion, suggesting that the peridotitic protolith had undergone partial melting (Simonet, 2000). Considering these observations, we suggest that the Mangare ultramafites may be scattered relicts of an old ophiolitic sequence, not necessarily a lateral equivalent of the Taita Hills ophiolites. This hypothesis still lack support, and more chemical analysis of metamafic and meta-ultramafic rocks of this area are required.
The geology of rocks immediately hosting the ultramafic body of the John Saul Mine has been described by Austromineral (1978), Pohl (1985), and Mercier et al (1999a). It is hosted to the west and to the north by a serie of graphite- and sillimanite-bearing gneisses showing small lenses of conformable anatectic pegmatites constituted of quartz, orthoclase and sodic plagioclase. On this side the contact with the metasediments is unconformable, but complicated by the presence of the Kimbo vein and of fault mirrors. To the south and to the east, the ultramafic body is conformably hosted by a two to four metre thick sequence of felsic gneisses (meta-arkoses?). These leucocratic rocks consist of orthoclase, sodic plagioclase, quartz, minor Cr-mica (fuchsite), and do not contain graphite. They are overlaid by a few metres of sulfide and graphite bearing gneisses, and by graphite-sillimanite gneisses similar to those of the western side. Kyanite is present in minor quartz-rich bands in these metasediments, and is partly replaced by sillimanite.
Several corundum-bearing veins are present around the ultramafic body. Each of them shows different petrographic and geometric characteristics, and each of them yields corundum crystals of different gemmological characteristics and value. This suggests that all these veins did not form under the same geological conditions, and justifies to study each mineralisation separately before any generalisation is attempted. We will now describe and study one of these mineralisations, the Kimbo vein, which is currently the source of the mine's ruby production.
4. Petrography
The Kimbo vein lines the western flank of the ultramafic body. It has a roughly WNW direction and is subvertical. As much as one can appreciate judging from available outcrops, it does not extend inside the host rock of the ultramafic body, neither northwards nor southwards. However to the south, the upper part of the vein and the ultrabasic body are cut by a fault that conceals a possible continuation of the vein inside the metasediments. To the north, observations are restricted due to the lack of outcrops. The thickness of the vein varies from a few centimetres to more than two metres, with an average if about one metre. The contact between the vein and the ultramafic rock is underlined by chlorite, serpentine and cryptocrystalline dolomite salbands. Chlorite crystals are oriented parallel to the contact, and locally show vertical lineations on slickensides. The contact between the vein and the host graphite gneisses is usually sharp and uncomformable, which means that it is either tectonic or intrusive. The first possiblity seems unlikely because no faulting can be observed between the vein and the gneisses.
Figure 3 - Picture of the Kimbo vein - 1. Ruby-tourmaline plagioclasite; 2. Tourmaline-rich ruby plagioclasite; 3. Apatite-phlogopite rock; 4. Vermiculite blackwall; 5. Tourmaline, ruby, +/- plagioclase; 6. Chlorite blackwall; 7. Serpentine; 8. Ultramafic rock; 9. Graphite gneiss; 10. Corundum-Tourmaline graphite gneiss.
Figure 4 - Cross section of the Kimbo vein - 1. Ultramafic rock; 2. Cryptocrystalline dolomite vein; 3. Ruby plagioclasite (note the presence of this rock type within the gneissose host rock); 4. Chlorite; 5. Cryptocrystalline dolomite; 6. Serpentine; 7. Phlogopite veins with chalcedony; 8. Graphite gneiss; 9. Quartz-feldspar pegmatite lumps; 10. Graphite gneiss with corundum and tourmaline.
Figure 4 - Cross section of the Kimbo vein - 1. Ultramafic rock; 2. Cryptocrystalline dolomite vein; 3. Ruby plagioclasite (note the presence of this rock type within the gneissose host rock); 4. Chlorite; 5. Cryptocrystalline dolomite; 6. Serpentine; 7. Phlogopite veins with chalcedony; 8. Graphite gneiss; 9. Quartz-feldspar pegmatite lumps; 10. Graphite gneiss with corundum and tourmaline.
On at least two points, marrow diversions of the vein have been observed, directed towards the ultramafic rock. They have a thickness of a few centimetres to twenty centimetres. The distance to which they extend from the main vein does not exceed two metres. They are vertical, and connected to the chlorite veins network that cuts across the ultrabasic rock. Their petrography is otherwise similar to that of the main vein. This geometry is that of a pegmatite intrusion.
The vein has a pegmatoid structure, with a feldspathic matrix in which are scattered centimetre-size crystals of corundum and tourmaline. In the upper parts of the vein, crystals are regularly scattered. Locally the vein appears richer in corundum. At deeper levels, millimetre to centimetre size crystals are associated in nests with fuchsite, and in lenticular groups of small crystals. Such concentrations are vertical, and are not a metamorphic foliation.
Although the limit between the vein and its host rock is sharp, the gneiss immediately at the contact of the vein are also sometimes mineralised. In this case either scattered corundum and/or tourmaline crystals, or small pegmatoid lenses with corundum, are present in the gneiss. The thickness of this mineralised zone in the host rock can reach up to 4 to 6 metres, excluding the thickness of the actual vein. The boundaries of such mineralised zones cuts the foliation of the gneisses.
5. Mineralogy
Minerals were identified in thin sections, except fluorite that was observed in macroscopic samples. Microprobe analysis were performed at the Electron Microprobe Service of IFREMER (Brest, France). Calibration was done with appropriate standards. Analysis were done under a 20 nA current. Counting times for corundum analysis were 15 s for Ga, 10 s for Ti, V, Co, and 6 s for Al, Si, Mn, Fe, Cr. Counting times for other mineral analysis were 10 s for P, Ti, V, F, and 6 s for Si, Al, Cr, Fe, Mg, Mn, Ga, Na, K.
5.1. Plagioclase
The rock consists mostly of more or less altered plagioclase feldspar matrix. In the upper part of the vein, down to a depth of 5 to 10 metres at most, this matrix shows a granoblastic structure and feldspar (An41) is sometimes antiperthitic. In deeper parts of the vein, plagioclase porphyroclasts are surrounded by smaller neoblasts, the whole having a granuloblastic structure. Their composition varies but remains in the field of andesine (An36-An50).
5.2. Tourmaline
Tourmaline is poikiloblastic and euhedral. Macroscopic samples show a deep green colour, and occasionally a zonation with a dark green core surrounded by a yellowish green rim. The striated prisms have an elongation ratio of 1:3, and a subcircular section. Under the microscope it is colorless. Boron having not been analysed, the structural formula was calculated on the basis of 49 oxygen atoms, assuming the presence of 6 boron atoms, following the method of Michailidis et al (1996). This tourmaline is a magnesian (XMg=1) dravite containing 20 to 30 % of uvite. On a diagram of Henry and Guidotti (1985) it plots in the field of tourmalines associated to ultramafic rocks. Its green colour is owed to its chromium content (up to 0.415 wt % Cr2O3 - see Schmetzer et al, 1979), which is also linked to its association with ultramafites. Its composition is similar to that of other gem quality dravite-uvite tourmalines from East Africa (Simonet, 2000b). This mineral, present in large amount at surface level, tends to become scarce in the deeper parts of the vein. It is locally intergrown with corundum.
5.3. Muscovite
This mica is present in some parts of the vein. It is silvery grey or, more often, green (fuchsite) in macroscopic samples, and colourless in thin sections. It is either isolated in the plagioclase matrix or associated to corundum. Its contours cut those of plagioclase crystals, which shows that it is posterior to it. Its structural formula is:
(K0.81-0.89Na0.06-0.09)(Al1.83-1.97Mg0.14-0.17)(Al0.90-0.97Si3.03-3.09)(OH2)
This formula is that of a muscovite containing a small percentage of aluminoceladonite. In theory, it belongs to the phengite serie (Rieder et al, 1998). However since Si<3.1 it is justified to use the term muscovite (Rieder et al, 1998). The green colour of some samples is owed to the presence of Cr (0.023- 0.073 wt % Cr2O3).
5.4. Kyanite
Kyanite is rare in the Kimbo vein, although it is sometimes present in large quantities in other corundum deposits of the Mangare area. It is always in association with muscovite and corundum, usually between these two minerals.
5.5. Corundum
Corundum is pink to purplish red in macroscopic sample, and colourless in thin sections. It is subeuhedral to euhedral, in which case it appears as z (221) or occasionally n (223) bipyramids. The pinacoid face (001) is absent. Corundum is observed at the contact of the plagioclase, or within groups of Mg-muscovite crystals. In this case it may be, although rarely, separated from the mica by a kyanite blade. When corundum is in contact with plagioclase, no paragonite or margarite alteration fringe can be seen. When it is in contact with Mg-muscovite, its contour is often denticulate, and small corundum granules, partly coalescent with the main corundum crystal, are visible in the mica. This suggest that corundum developpes inside Mg-muscovite as small crystallites, that are eventually incorporated by recrystallisation in larger corundum crystals. The external part of the corundum crystals is usually rich in various inclusions, while the inner part is free of mineral inclusions, that is of gem quality. The corundum-tourmaline intergrowth mentionned above show that these minerals have crystallised together.
The red colour of corundum is owed to the presence of chromium in substitution for aluminium in the crystalline lattice (Fritsch and Rossman, 1987). Corundum crystals of this vein typically contain 0.1 to 0.35 wt % Cr2O3. Other trace elements present include iron (FeO < 0.028 wt %) and gallium (Ga2O3 < 0.063 wt %). The presence of measurable amounts of Ga in some Kenyan rubies has also be noted by Muhlmeister et al (1998). Co, V and Mn could not be detected in the conditions of the microprobe analysis. Ti is present as epitaxic rutile needles ("silk"), 20 to 100 micrometre long for a thickness of a few micrometre. This prevents to measure the actual Ti content of the corundum as it can not be differentiated from the Ti of the rutile inclusions. Global TiO2 content (corundum + rutile inclusions) is typically less than 0.16 wt %. The red colour of corundum intensifies from the top towards the bottom of the vein, due to a slight increase of the chromium content.
Crystal sections perpendicular to the ternary axis show complex colour and silk zonations with hexagonal boundaries. The detail of these zonations (frequency and width of silk bands) is complicate and varies very much from one crystal to the other. Such oscillatory zoning is considered by Yardley (1991) as typical of metasomatic crystallisation. Two main zonation patterns, however, can be identified. Most corundum crystals (95-100 %) show a silk rich core surrounded by a thin silk free rim of deep red colour. Chemically, this translates into a Ti-rich and Cr-poor core, and a Ti-poor and Cr-rich rim. However, other crystals (less than 5 %) show a "reverse" zonation with a red silk free core and a silk rich rim of pink colour. Both type of zonation pattern can be present in the same part of the vein, although the "reverse" zonation tends to be more common in the deeper parts of the vein, and is absent at surface level. This fact, together with the general increase of chromium content with depth, translates into an actual improvement of the rubies' quality with depth.
These zonation patterns show the existence of several generations of corundum crystals, especially in the deeper parts of the vein. The fact that these generations are not part of different paragenesis shows that they have crystallised in similar thermobarometric conditions. This supports the hypothesis of a metasomatic origin for this mineralisation, and the different generations can be considered as the result of several mineralising fluids input.
5.6. Accessory minerals
Graphite is present in notable amounts in the lower part of the vein, but is seldom observed at surface level. It occurs as isolated flakes or as radiating aggregates of flakes. Monazite is a common accessory mineral. It is systematically included in Mg-muscovite, at the contact of or close to corundum crystals. However, it has not been observed as an inclusion in this mineral. Iron sulfide globular crystallites are often observed, especially as inclusions in corundum. Zircon (partly as inclusions in corundum), apatite and more rarely fluorite are also present.
6. Bulk geochemistry
Four analysis were performed on the corundum bearing plagioclasite collected at different depths. Representative samples of rocks were collected at different levels of the vein, and crushed in a jaw crusher and then in a tungstene ring crusher. Evidence of abrasion by corundum grains on the later led us to discard W and Nb results due to possible contamination. Analysis were done at the CRPG of Nancy (France). Samples were melted in LiBO2 and dissolved in HNO3. Major elements were analysed by ICP-AES, and trace elements by ICP-MS. REE were normalised using CI chondrite values of Sun and McDonough (1989).
The Kimbo corundum bearing plagioclasite is an alumina-rich and silica-poor rock (32 to 42 wt % Al2O3; 43 to 52 wt % SiO2). It contains varying but low amounts of Mg, Ca, Na, and K, depending on the proportions of tourmaline, plagioclase and mica in the rock. Na and Ca content in the rock show a positive correlation. Co, Ni and Cr contents (respectively 26 to 95 ppm, 15 to 100 ppm, and 150 to 290 ppm) are exceptionnal for a felsic rock (see Krauskopf, 1967). Cu and Pb content are relatively low. Sr content is very high (360-1500 ppm), correlated to Ca, and seems to increase with depth. B content varies depending on the mode of tourmaline. F, Cl and Li content are respectively 240 to 600 ppm, less than 20 to 60 ppm, and 40 to 80 ppm. The plagioclasite is REE enriched, with La and Lu values of respectively 218-684 and 24-102 times chondritic values. Two types of REE spectrum are present, the difference being in the slope of the HREE spectrum (La/Lu ratio of 2-3 in samples RT005 and RT008, and 14 in samples RT006 and RT007). Both types show a more or less well marked negative Eu anomaly, which is surprising for a plagioclase-rich rock.
7. Mass balance calculations
Field observation and petrography allow to formulate a starting hypothesis concerning the origin of the Kimbo corundum-bearing plagioclasite. For instance, its geometry is that of a magmatic intrusion, such as a pegmatite dyke, and excludes the possibility of an anorthosite layer in a mafic-ultramafic complex, like those of some other East African ruby deposits such as Longido or Lossongonoi in Tanzania (Game, 1955; Solesbury, 1965; Simonet, 2000), and Kitui in Kenya (Barot and Harding, 1994). However it is clear that corundum is not of magmatic origin, because of the absence of cumulate structures, and of the presence of corundum outside the vein itself, inside the host gneisses. A purely metamorphic origin is also ruled out due to the small size of the mineralisation, the absence of foliation, and the chemistry of the bulk rock. These observations, as well as petrographic and mineralogic data, thus points towards a metasomatic origin, that is the chemical alteration of a protolith.
The study and characterisation of the metasomatic phenomena that led to the formation of the plagioclasite are difficult, owing to the heterogeneity of the rock, and to the absence of a non- metasomatised protolith. Several authors proposed that this protolith is a pegmatite (Austromineral GmbH, 1978; Horkel et al, 1979; Pohl and Niedermayr, 1979; Pohl, 1985; Key and Ochieng, 1991), and this is in accordance with our observations. In an attempt to characterise, at least in a qualitative way, the transformations that led from a pegmatite to a corundum bearing plagioclasite, we had to select a plausible protolith.
The protolith we used is a conformable pegmatite sill of a plurimetric thickness, that occurs within the graphite gneisses directly hosting the ultramafic body on its eastern side, a few tens of metres away from the Kimbo vein. This pegmatite consists of quartz, orthoclase, sodic plagioclase, with minor amounts of dravitic tourmaline, muscovite and graphite. The grain size does not exceed a few centimetres and the rock is quite homogeneous, with no zoning nor quartz cores. The characteristics of this pegmatite are similar to that of most anatectic pegmatites of the Mangare area.
The pegmatite is Si rich (about 74 wt % SiO2) and Al poor (about 15 wt % Al2O3) and has a typical granitic composition. The two samples yielded different REE spectra. The REE spectrum of sample RT009 is U shaped, with La/Lu=1.17, and the REE spectrum of sample RT010 has a La/Lu ratio of 3.69 and a positive Eu anomaly.
In order to assess the chemical modification that affected the pegmatite, it is reasonable to assume that Al remained immobile during metasomatism andto take this element as a reference for mass- balance calculations. The immobility of Al has been noted in many corundum deposits where corundum crystallises as a result of selective leaching of Si (desilication process - Kerrich et al, 1987; Grapes et al, 1996), and in other metamorphic contexts (Kovácic, 1996). This assumption is supported by the fact that other elements known to be immobile such as Zr and Hf show the same behaviour that Al. Calculations using average compositions for both type of rocks show that the transformation of a pegmatite into a corundum-bearing plagioclasite involves a 50 % weight loss and 60 % volume loss .
This weight loss is mainly due to a 74 % loss of Si. Na, K, Rb, Ba and Pb also show a loss of respectively 33, 90, 92, 31 and 77 %. Ca is enriched by 38 % and Sr by 81 %, and Mg and Cr by respectively 157 and 3462 %. Co shows a slight loss (32 %) and Ni remains unchanged when calculating with the average plagioclasite composition, although when using individual compositions some samples show an gain of 30 % and some a loss of 30 %. LREE show a strong gain of 320 to 580 %, while HREE and Y show a gain of 40 to 130 %. Eu shows a gain of 58 %, and P is increased by 93 % . Elements that are not significantly modified include V, Ti, Hf, and Zr. Th shows a 143 % enrichment. U shows a different behaviour in individual samples, with in some cases a loss of up to 68% and in some cases a gain of up to 59 %. B also shows a strong gain, although this is difficult to calculate reliably due to the uneven distribution of tourmaline crystals in the plagioclasite.
Two groups of incoming elements can be considered, depending on their affinities. The first group includes Mg and Cr. These elements have ultramafic affinities and could originate from the adjacent ultramafic rocks. The Si depletion of the rock is also an ultramafic characteristic. The second group includes P, B, REE, Y, Th, and in some samples U. These elements are not originating from ultramafic rocks but rather from felsic rocks, and that are present in pelitic metasediments hosting the mineralisation. Except B, they are concentrated in apatite and monazite. A high mobility of REE was also observed in other metasomatic corundum deposits, such as "verdite" deposits (Kerrich et al, 1987), however the behaviour of REE in metamorphic and metasomatic processes is known to be unclear and irregular (Graunch, 1989). The Ca enrichment can not be related easily to any of these two groups. The fact that the different metasomatic minerals in the vein (corundum, tourmaline, Mg-muscovite) contain elements from both groups show that they were incorporated to the vein by the same process.
8. Geochronology
A representative sample of the Kimbo plagioclasite weighting about 8 kilos was collected at a depth of about 10 metres.
Zircon- Zircon crystals separated from the rock are euhedral, and opaque owing to their high U concentration (about 3000 ppm). They are rich in Pb (300 ppm) and high 206Pb/ 208Pb show a low Th content. An intercept was obtained at 612 +/- 0.6 My. Zircon has been observed as inclusions in corundum and is therefore anterior to this mineral. The rather immobile behaviour of Zr and Hf during the transformation of the pegmatite into a plagioclasite suggest that zircon remained unaffected by this transformation, and that zircon thus dates the setting of the pegmatite protolith.
Monazite- Monazite crystals yielded two different ages. Some monazites crystallised at 609 +/- 0.4 My. These contain 3000 ppm Pb, 1.4 % U, and about 1% Th. The second group crystallised around 590 My and contains 2000 ppm Pb, 1 % U, and 1 % Th. Monazite occurs in muscovites crystals that are a result of the metasomatic phenomena and that are, at least partly, the immediate precursor of corundum. They are thus dating the metasomatic event. The fact that they are two generations of monazite crystals suggest that the metasomatic crystallisation of corundum took place during two separate stages. This could be supported by the observation of different generations of corundum crystals, described above.
9. Origin of the Kimbo mineralisation
9.1. Metasomatic phenomena
We suggest that the mineralising fluid originated from metasediments such as those hosting the ultramafic body, and containing tourmaline, hence boron. These fluids were re-equilibrated with ultramafic rocks, leading to silica undersaturation of the fluid and silicification of the ultramafites. At this stage the fluids became Cr- and Mg-enriched. Cr is known to be mobile in the presence of a fluid phase in regional metamorphism conditions (Jan et al, 1972; Tracy, 1991; Peretti et al, 1996). Last, the fluids reacted with a pegmatite located along the ultramafic body. Their silicodeficient character led to the selective leaching of Si from the protolith, and the relative Al enrichment. Na, K, Rb, Ba and Pb were also leached out. The same fluid, reacting by the same process, led to the formation of the tourmaline +/- corundum mineralisation in the host gneisses.
9.2. P-T conditions
P-T conditions of formation of the vein are difficult to estimate because of the lack of adequate intermineral reactions. The rock obviously formed in the stability field of kyanite and on the right side of the usually retromorphic reaction:
margarite <-> corundum + anorthite + water
This reaction occurs at 598oC for a 5 kb pressure with XH2O=1 (Grapes and Palmer, 1996). However the presence of graphite shows that XH2O was actuallt lower than 1, which shifts the reaction towards lower temperatures. The presence of Na also shifts the reaction towards lower temperatures (Dymek, 1983). This gives us a rough estimate of a temperature above 500oC and pressure above 5 kb. The phengite barometre of Massone and Schreyer (1987) can not be used here in the absence of biotite, K feldspar and quartz. The presence of tourmaline suggest that the mineralisation did not form in conditions much higher than the low grade granulite facies, this mineral usually disapearing in these conditions and gives way to kornerupine (Grew, 1988). According to this author, in rare cases where tourmaline subsists in granulite facies (in Mg- and Al-rich rocks), it is systematically associated to grandidierite or kornerupine. This fact was also observed by Haapala et al (1977) on the granulite facies corundum deposit of Kittilä in Finnish Lapland, where corundum is associated to sapphirine, kornerupine, and to a lesser extent tourmaline. Sapphirine, although present in other neigbouring corundum deposits of the Mangare area such as Aquamines (Austromineral GmbH, 1978; Mercier et al, 1999a), was never observed in the Kimbo corundum-bearing plagioclasite.
9.3. Proposed model
We propose the following model for the origin of the Kimbo corundum mineralisation:
1) During an early phase of the Mozambiquian Orogeny ultramafic rocks, most likely of ophiolitic origin, were set in the metasedimentary series of the Kurase Group (Horkel et al, 1979; Pohl et al, 1980; Mercier et al, 1999a). This phase is probably the Samburuan event of Mosley (1993).
2) At 612 +/- 0.6 My a pegmatite dyke intruded the contact between the ultramafic body and its host graphite gneisses, and faults perpendicular to this contact. The geometry of the intrusion was determined by local rheologic discontinuities between different rocks and by faults.
3) At 609 +/ 0.4 and 590 My, the pegmatite underwent two phases of metasomatic alteration due to fluids of metapelitic origin which had been reequilibrated with ultramafic rocks. These fluids were Si depleted, and rich in B, Mg, Cr, Th, REE and Y. Reaction of these fluids with the pegmatite and locally with the host gneisses led to their desilication, that translated into the dissappearance of quartz. Orthoclase and sodic plagioclase were transformed in andesine, and Mg-bearing muscovite crystallised. Corundum appeared at the expense of muscovite, and probably also directly of feldspar. These transformations all betray a loss of Si and an relative gain of Al, as shown by the ratio of these elements in involved minerals (Figure). Crystallisation of tourmaline was synchronous with that of corundum. Chromium gave the red colour to corundum and the green colour to tourmaline. Exiting fluids were enriched in Si, Na, K, Ba and Rb.
4) Tectonisation led to the formation of sliken side between the vein and the ultramafic rocks in chlorite salbands and to local transversal faulting of the vein.
10. Discussion
10.1. Origin of the mineralisation
Most authors agree on the fact that the formation of the ruby deposits of the Mangare area is the result of the desilicification of pegmatites or felsic metamorphic rocks in the vicinity of small ultramafic bodies (Bridges, 1982; Pohl, 1985; Key and Ochieng, 1991; Levitsky and Sims, 1997; Mercier et al, 1999a). Austromineral GmbH(1978) proposed a model in which pegmatites are directly desilicated by ultramafic rocks. According to this model the pegmatite provides CO2, H2O, and SiO2 to the ultramafites, and receive Mg and Cr from them. This is accompanied by the carbonisation, hydratation and silicification of the ultramafites. However this model does not explain all petrographic and geochemical characteristics of the mineralisation, especially the enrichment in "felsic" elements such as boron or REE, as such elements can not originate from the ultramafites.
P-T conditions for the formation of the Kimbo vein are not be well constrained and only minimum conditions were determined. P-T conditions for other veins at the John Saul Mine are better known. Key and Ochieng (1991) have estimated to 630-670oC and more than 7 kbars the conditions of formation of a corundum bearing rock from the Main Pit area, using 2 feldspars thermometry and the stability field of kyanite. Using the same method, Simonet (2000a) determined that the near by corundum mineralisation of the Cow Boy Pit formed under quite similar P-T conditions (620-630oC and more than 6 kbar). Mercier et al (1999a) determined that the Mangare area corundum-bearing rocks recorded temperatures around 700-750oC and pressures in the range 8-10.5 kbar. These data were obtained using the reaction sapphirine + water -> corundum + chlorite + spinel, experimentaly determined by Seifert (1974) and Ackermand et al (1975). However this reaction occurs in only one of the corundum- bearing rocks of the Mangare area (at the Hard Rock Mine). We believe that it is impossible to generalise and to assume that the P-T conditions of one of the Mangare area corundum veins is that of all the veins. In deed it is clear that the Hard Rock Mine corundum mineralisation did not form in the same conditions that the Cow Boy Pit and Main Pit veins. We also note that the P-T conditions that are recorded in the Cow Boy Pit and Main Pit corundum-bearing veins are consistant with those determined by Mercier et al (1999a) for the host metasediments.
Mercier et al (1999a) suggest that corundum bearing veins of the Mangare area have an exotic origin and got emplaced in the metasedimentary series together with the ultramafites. These authors base this conclusion on the fact that they determined different P-T conditions for corundum-bearing rocks and their host metasediments, and that corundum mineralisation are usually bound to the immediate surroundings of the ultramafic bodies. We have seen that their conclusions regarding the P-T conditions recorded in corundum-bearing rocks are not valid for all the deposits, and that some of the Mangare deposits show the same P-T conditions that their host rock. The observation, by Mercier et al (1999a), that "the ultrabasic and the ruby-bearing rocks together seem to form an entity without any observable genetic association with the surrounding gneisses" does not fit with our observations (figure).
We can not agree to the model of exotic origin of Mercier et al (1999a) for the following reasons.
(1) The ruby bearing rocks are not bound to the ultrabasic bodies, as host gneisses are also affected by metasomatic phenomena and contain corundum. The close association of ultramafites and corundum- bearing rocks is due to the fact that the former are necessary so that the later can undergo desilication and receive a chromium supply, and not to a common origin.
(2) Datations and field evidence show that the John Saul Mine ultramafic body was emplaced long before the Kimbo mineralisation, during a different phase of the Mozambique Belt history.
(3) At least part of the corundum-bearing rocks of the Mangare area have recorded P-T conditions that are consistant with that recorded in the host metasediments.
Their about a dozen of known corundum-bearing mineralisations in the Mangare area, each of them showing different geological characteristics. This shows that they formed under different geological conditions, and this is supported by the fact that the Hard Rock Mine sapphirine-corundum rock and the Cow Boy Pit and Main Pit corundum-bearing rocks formed under different P-T conditions. It is thus necessary to study each vein as a separate entity before any general conclusion can be drawn.
10.2. Comparisons with similar deposits.
The corundum deposits of the John Saul Mine show many strong similarities with other corundum deposits in East Africa. The Umba and Kalalani sapphire and ruby deposits in Northern Tanzania (Solesbury, 1967; Seifert and Hyrsl, 1999), actually distant from the John Saul Mine by a hundred kilometres along the strike, are associated to ultramafic bodies similar to that of the John Saul Mine. There corundum mineralisation is the result of metasomatic interactions between the ultramafites and pegmatites. The variety of colours found in these deposits is due to variations in the relative concentrations of trace elements in the corundum crystals. Sapphire deposits in Madagascar (Kiefert et al, 1996; Schwarz et al, 1996) and Kashmir (Peretti et al, 1990) also originate from complex metasomatic alteration of felsic rocks in the presence of ultramafites.
However, we postulate that the ruby deposits of Longido and Lossongonoi in Tanzania (Game, 1955; Solesbury, 1965; Simonet, 2000), the pink sapphire deposit of Kitui in Kenya (Barot and Harding, 1994), and ruby-bearing anorthosites of Madagascar (Nicollet, 1986; Mercier et al, 1999b) have a different geological origin. There are significant differences between these deposits and the deposits of the Mangare area (Table...). Corundum-bearing amphibolites and anorthosites are of purely metamorphic origin and result from the hydratation of rocks of gabbroic compositions in granulite facies (Lasnier, 1977; Tenthorey et al, 1996; Simonet, 2000a). Such rocks are associated to ultramafic rocks because they have a common origin, as parts of layered intrusion or ophiolitic mafic-ultramafic assemblages.
10.3. Regional geology
Geochronological data obtained in the course of this study are particularly interresting because such data are scarce in this part of Kenya. Until now the only datations available were a 827 +/- 55 My whole rock Rb/Sr age from a gneiss sample from the Taita Hills (Shibata, 1975), and K/Ar cooling ages ranging from 519 My (hornblende) to 456 My (feldspar), obtained by Shibata (1975) and Frisch and Pohl (1986) on rocks of the Taita Hills area.
Our data fit well in the history of the Mozambique Belt in Kenya as it was outlined by Key et al (1989) and Mosley (1993). The age of the pegmatite crystallisation and of the first metasomatic phase (respectively 612 and 609 My) is close to that of the Baragoian event, and the age of the second metasomatic phase corresponds to that of the Barsaloian event. Both these events are characterised by shear zones and strike-slip movements, that is conditions favourable to fluid and heat transfers, hence anatexis and metasomatism. In the Mangare area, Baragoian structures are obscured by later tectonic events but the Barsaloian event is represented by the ubiquitous low angle north-plunging lineation and by boudinage structures.
It is thus clear that there is not one but several episodes of corundum deposits formation in this part of the Mozambique Belt, and that these episodes are strongly connected to its regional history. Many studies of individual deposits of corundum and other gemstones will be required to get a better idea of the timing of these episodes, and a general picture of processes leading to gemstone deposits formation.
11. Conclusion
As a conclusion, we wish to emphasize that we are far from being able to produce a general picture of corundum formation in terms of P-T-t-fluids conditions for the Mangare area. This will only be possible after more studies of different veins and deposits as separate entities. On a wider scale, this is also true for other types of gemstone deposits, and for the whole of the Mozambique Belt. As a matter of example, there are about 40 known occurences and deposits of corundum, and almost one hundred of other gemstones, in a 100 km radius circle around the Mangare area. Each of these occurences has its own characteristics and it is rare to find two similar occurences of one gemstone type. It is therefore clear that trying to draw general conclusions from the study of only a few occurences would only lead to wrong conclusions.
