Journal of Petrology Advance Access originally published online on January 22, 2009
Journal of Petrology 2009 50(2):323-359; doi:10.1093/petrology/egp001
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Fractionation of Peralkaline Silicic Magmas: the Greater Olkaria Volcanic Complex, Kenya Rift Valley
1Environment Centre, Lancaster University, Lancaster LA1 4YQ, UK
2Institute of Geochemistry, Mineralogy and Petrology, University of Warsaw, 02-089 Warsaw, Poland
3Department of Earth Sciences, Cespar, Open University, Milton Keynes MK7 6AA, UK
4Grant Institute of Geosciences, University of Edinburgh, Edinburgh EH9 3JW, UK
RECEIVED MAY 21, 2008; ACCEPTED DECEMBER 31, 2008
 |
ABSTRACT |
|---|
The Greater Olkaria Volcanic Complex is a young (
20 ka) multi-centred system in the central Kenya Rift Valley, mainly represented at outcrop by peralkaline rhyolites. The rhyolites show significant compositional variation; peralkalinity [mol. (Na2O + K2O)/Al2O3] varies from 1·01 to 1·55, Zr contents from 442 to 3640 ppm and Rb contents from 262 to 1056 ppm. More peralkaline rhyolites were generated along multiple, but generally closely similar, liquid lines of descent by 
75% fractional crystallization of alkali feldspar–quartz-dominated assemblages from mildly peralkaline parental magmas, themselves probably derived by fractionation of trachytic magmas. Crystal fractionation took place at more than one upper crustal level. The rhyolite magmas were erupted from 13 centres, each having an eruptive history and geochemical evolution broadly similar to, but distinct from, those of the other centres. For most of the life span of the Olkaria system, almost the whole spectrum of peralkaline rhyolite compositions was erupted from vents in the complex at any one time. Apparent partition coefficients are presented for 34 trace elements in sanidine, fayalite, ferrohedenbergite, amphibole, biotite, ilmenite and chevkinite-(Ce). With the exception of certain values for Ba, Rb and Sr in sanidine, most coefficients vary systematically with whole-rock composition. KEY WORDS: Kenya Rift; peralkaline rhyolites; mineral/glass partition coefficients; plumbing system; eruptive centres
|
INTRODUCTION |
|---|
The Greater Olkaria Volcanic Complex in the south–central Kenya Rift Valley (Fig. 1), is a young (
20 ka), small-volume, frequently erupting, multi-centred system dominated at outcrop by a peralkaline rhyolite dome and lava field. Its peralkaline nature contrasts to the metaluminous rocks that more commonly form silicic dome fields, such as Long Valley Glass Mountain, California (Metz & Mahood, 1991
), the Coso volcanic field, California (Duffield et al., 1980; Bacon et al., 1981; Manley & Bacon, 2000) and the early (320 to 65 ka) history of the Taupo volcanic centre, New Zealand (Sutton et al., 1995).
|
Some earlier petrological studies (e.g. Davies & Macdonald, 1987
; Macdonald et al., 1987) referred to the Greater Olkaria Volcanic Complex (GOVC) as the Naivasha Complex. Clarke et al. (1990) introduced GOVC to stress the close association of the rocks with the well-known Olkaria geothermal field. Subsequent studies (Black et al., 1997; Scaillet & Macdonald, 2001, 2003: Heumann & Davies, 2002; Macdonald et al., 2008) have used the Clarke et al. term; we follow their example, simplifying it to Olkaria complex.
There have been several petrological and geochemical studies relevant to the petrogenesis of the Olkaria peralkaline rhyolites. On the basis of a comparison of the major element compositions with phase equilibria in the system Na2O–K2O–Al2O3–SiO2, Bailey & Macdonald (1970) suggested that the rhyolites represent a path of increasing partial melting within the continental crust in the presence of an alkali-bearing vapour. Macdonald et al. (1987) interpreted major and trace element data to show that closed-system fractional crystallization could not alone account for the compositional variation in the rhyolites. They invoked a model of partial melting of heterogeneous crustal source rocks, followed by variable amounts of crystal fractionation, with an important role for volatiles in promoting peralkalinity and in controlling trace element distribution patterns. Davies & Macdonald (1987) showed that whereas Sr–Nd isotope relationships are consistent with the rhyolites having been derived from the associated basalts by an assimilation–fractional crystallization (AFC) process, Pb isotopic systematics clearly show that the basalts and rhyolites are not part of a cogenetic suite. They proposed that the rhyolites represent crustal melts derived from 6 km depth.
Black et al. (1997) presented alpha spectrometric data to show that the Olkaria peralkaline rhyolites have initial (230Th/232Th) ratios (0·73–0·77) lower than the Olkaria basalts (0·8 to 1·2), confirming that the basalts and rhyolites were not part of a cogenetic suite. Heumann & Davies (2002) used Rb–Sr age determinations and U–Th disequilibrium to explore various aspects of the evolution of the Olkaria rhyolites, including magma fractionation rates, crustal residence times and phenocryst crystallization histories. They broadly accepted a crustal origin for the rhyolites but noted the need for an extended period of feldspar fractionation to produce the strong depletion in Ba and Sr in the rocks. Scaillet & Macdonald (2003) argued that the rhyolites could have formed by fluxing of a quartzo-feldspathic source by fluids having high F contents, provided that the oxygen fugacity (fO2) was lower than the quartz–fayalite–magnetite (QFM) buffer and temperatures were less than 800°C.
In the most recent study, Macdonald et al. (2008) showed that melts of mildly peralkaline rhyolitic composition formed by two mechanisms at Olkaria; namely, crystal fractionation of metaluminous trachyte and partial melting of syenitic rocks. They deemed that the latter process was too small-scale to have generated the Olkaria rhyolites and concluded that crystal fractionation has been the dominant mechanism in the generation of the least peralkaline rhyolites, the trachytes themselves having been derived by extensive fractionation of parental transitional basaltic magmas. The crystal fractionation model leaves unexplained the differences in Pb isotope compositions and Th–U disequilibria between the Olkaria basalts and peralkaline rhyolites. Resolution of the discrepancies will require a much fuller understanding of how the isotopes vary within and between the eruptive centres. This paper explores the further evolution of the least evolved rhyolites, whatever their ultimate origin.
As noted above, Macdonald et al. (2008) focused their petrogenetic discussion on the least peralkaline rhyolites at Olkaria; that is, those with a Peralkalinity Index [PI; mol. (Na2O + K2O)/Al2O3] of 1·0–1·1. However, the complex has erupted a wide range of peralkaline rhyolites, with PI up to 1·55, and offers an unusual opportunity to study in detail the further chemical evolution of peralkaline silicic magmas, which are themselves highly evolved. The pristine glassy nature of many of the rhyolites also makes it possible to determine apparent partition coefficients for a range of phenocryst phases and to assess how they vary with mineral and whole-rock compositions. Finally, the Olkaria rhyolites have been erupted from a number of discrete, geographically adjacent, centres that have experienced different evolutionary histories, allowing us to examine the subsurface structure of the complex. Specific aims of the study are:
- to provide a geochemical dataset that considerably expands, geographically and stratigraphically, the number of available analyses of Olkaria peralkaline rhyolites;
- to expand the available mineral-chemical dataset and to provide mineral/glass apparent partition coefficients for a range of phenocrysts, some for the first time;
- to explore the number and distribution of distinct magma batches emplaced during peralkaline rhyolitic activity in the Olkaria complex and the temporal relationships between them;
- to relate the geographical and temporal variations in rhyolite composition to the nature of the Olkaria plumbing system.
|
GEOLOGICAL OUTLINE |
|---|
The regional setting and stratigraphy of the Olkaria complex and surrounding areas have been described by Clarke et al. (1990
). The complex is bounded to the north by the Eburru complex, to the east and south by the Longonot and Suswa volcanoes, respectively, and to the west by the western rift margin (Fig. 1). At least 80 small volcanic centres have formed in the complex, most occurring either as steep-sided domes, formed of lavas and/or pyroclastic rocks, or as thick lava flows of restricted lateral extent. The earliest sequence exposed, however, is the Ol Njorowa Pantellerite Formation (unit O1; Fig. 2), represented by welded pyroclastic rocks, lava flows and plugs, and thought to be associated with the collapse of a caldera measuring 11 km by 7·5 km and now represented by the trace of a ring fracture (Clarke et al., 1990). Initial post-caldera activity was represented by the eruption of peralkaline rhyolitic lavas and pyroclastic rocks of the Lower Comendite Member of the Olkaria Comendite Formation (units O2/Op2; Fig. 2). Interbedded with the rhyolites in the SW of the complex are minor volumes of trachyte lavas (Olkaria Trachyte) and basalt lavas (Lolonito Basalt). The little studied Maiella Pumice comprises widespread trachytic and rhyolitic pyroclastic rocks to the NW, west and SW of the complex. They may have been erupted from vents within the complex (Clarke et al., 1990). The predominantly dome-building phase of the Middle Comendite Member (units O3/Op3) followed, leading on to the short, thick lava flows of the Upper Comendite Member (O4/Op4; Fig. 2). The most recent activity has been from the north–south-trending Ololbutot fissure (Ololbutot Comendite Member; O5/Op5).
|
An important feature of the Olkaria complex is that magmatic inclusions occur in rhyolites from all stages of the post-caldera activity. They range in composition from 50 to 63 wt% SiO2, filling a gap represented in the eruptive products, and thereby providing a continuum of compositions from basalt to peralkaline rhyolite at Olkaria. Mixing between the magmatic inclusions and the host rhyolites has been extremely limited (Macdonald et al., 2008
).
Eruption ages of the Olkaria rhyolites are rather poorly constrained. Clarke et al. (1990) used two new 14C dates from the neighbouring Longonot volcano to bracket the younger Olkaria events (Fig. 2). Ages from the literature (Fig. 2) were corrected for atmospheric changes with time. The Lower Comendite Member is older than 9150 ± 110 years BP; the Middle Member is younger than that date but older than 3280 ± 150 years BP. Carbonized wood from a pumice flow associated with the youngest, Olobutot, flow gave an age of 180 ± 50 years BP. Using regional and geomorphological relationships, Clarke et al. (1990) estimated the Lower Member to be 20 ka, the Middle 8 ka, the Upper 6 ka, and the Ololbutot Member <0·4 ka. The most poorly constrained age is the 20 ka for the oldest units, estimated from the fact that the eruptive rocks post-date a highstand of Lake Naivasha dated at 21 ka (Clarke et al., 1990). Peralkaline rhyolitic magmatism at Olkaria seems, therefore, to have been at least semi-continuous for the last 20 kyr.
Peralkaline rhyolites, associated with basalts and hawaiites, were erupted on the Ndabibi plains, an 11 km wide, low-lying area between the Olkaria and Eburru complexes (Fig. 1). Some of the rhyolites are strongly peralkaline and are related to the Eburru complex. Most, however, are more mildly peralkaline and are taken to be associated with the Olkaria complex (Macdonald et al., 1987). These constitute unit N of Clarke et al. (1990). We distinguish two groups. First is a set of domes (Group 1) located to the SW of Lake Naivasha and essentially contiguous with other eruptive rocks of the complex. Their precise age is unknown but they are overlain by the Maiella Pumice and are therefore 20–10 ka. The second group constitutes rocks of the Ndabibi centre. The majority of the rocks occur as domes ± lavas and pyroclastic cones whose well-preserved topography and relative lack of vegetation suggest that they are young, possibly equivalent in age to the Olkaria Comendite Member. Some Ndabibi flows are, however, cut by minor faults and must be rather older. Compositionally, they are identical to the younger Ndabibi peralkaline rhyolites.
Products of the Olkaria complex are easily distinguished from those of the neighbouring volcanic complexes (Fig. 1). Eburru comprises trachytes and rhyolites of strongly peralkaline affinity, contrasting with the more mildly peralkaline Olkaria eruptive rocks. Although fall deposits from Longonot can be intercalated with those from Olkaria, their brown and grey colours set them apart from the white Olkaria deposits. Magmatism at Suswa comprised trachytes and phonolites. Furthermore, the products of the complexes can be distinguished on the basis of their trace element characteristics; examples have been given by Clarke et al. (1990) and Scott & Skilling (1999).
The centres
Using Sr–Nd–Pb isotope and incompatible trace element (ITE) ratios, Davies & Macdonald (1987) and Macdonald et al. (1987) divided the Olkaria rhyolites into seven compositional groups. To stress that the rocks of each group tend to form discrete units geographically, Clarke et al. (1990) introduced the concept of discrete eruptive centres in the complex (e.g. the Gorge Farm centre). This term has been used in subsequent studies (Black et al., 1997; Heumann & Davies, 2002; Macdonald et al., 2008).
The Olkaria centres are distinguished by a combination of geographical proximity of the eruptive vents (Fig. 3), their occurrence as distinct topographical features and, as we shall show below, by various petrographic and/or geochemical features. Thirteen centres are recognized, their size ranging from only one flow (the North Portal and West Portal centres) through the line of small steep-sided domes of the Group 1 centre in the NW of the field to the 5 km wide amalgamation of lava flows found at Gorge Farm in the east–central area (Fig. 3). The term centre is not strictly applicable to the Arcuate Domes, which erupted in O2–O4 times along the southern trace of the ring fracture related to caldera collapse. An isolated dome of O4 age at the southern entrance to the Ol Njorowa Gorge (sample BB85.20) has not been assigned to a centre (Fig. 3).
|
The total volume of the eruptive products from each centre is poorly known, mainly because the pyroclastic units have yet to be fully assigned to the appropriate centres. Bone (1987
) estimated that the Broad Acres and the combined Kibikoni and Oserian centres each erupted 0·15 km3 of rhyolite magma, and Heumann & Davies (2002) estimated the volume of the eruptive rocks of the Gorge Farm centre to be 5 km3. The total volume of rhyolitic eruptive rocks is 13 km3 (C. M. Bliss, personal communication).
The range of stratigraphic units present in each centre is variable (Clarke et al., 1990) (Table 1). For example, the Kibikoni and Oserian centres erupted during O2 times only, whereas the Gorge Farm centre erupted from O2 to O4. Fuller details, including simplified maps, vent locations and sample localities may be found in Supplementary Data Figs 1–5, which are available for downloading at http://www.petrology.oxfordjournals.org. The main bulk of O2 activity is now exposed on the periphery of the complex (Oserian, Kibikoni, Olenguruoni and the southwesterly Arcuate Domes). However, O2 activity has also been identified at the Olkaria and Gorge Farm centres, and it is likely that O2 rocks underlie the Olobutot centre and possibly the eastern Portals–Broad Acres area. O3 activity occurred throughout the complex. O4 rhyolite lavas are concentrated in the east and centre of the complex, although they also occur at the Olenguruoni centre in the NW. Activity of O5 age has been relatively restricted; however, the Ololbutot lava flow (200 years) is the largest seen in the complex.
|
The syn- and post-caldera evolution of the Olkaria complex has many similarities to that of the Gedemsa volcano in the Central Ethiopian Rift (Peccerillo et al., 2003
). At Gedemsa, syncaldera pantelleritic ignimbrites were followed by the eruption of several small coalescing cones, mainly located within the caldera depression. They consist of obsidian domes and flows and fall deposits. The intra-caldera rhyolites host inclusions of basalt–trachyte composition (see Macdonald et al., 2008). |
SAMPLING STRATEGY AND ANALYTICAL METHODS |
|---|
 TOPABSTRACTINTRODUCTIONGEOLOGICAL OUTLINE SAMPLING STRATEGY AND ANALYTICAL... PHENOCRYST ASSEMBLAGES AND...GEOTHERMOBAROMETRYGEOCHEMISTRYPETROGENESISDISCUSSIONCONCLUSIONSSUPPLEMENTARY DATAAPPENDIX: SAMPLE DETAILSREFERENCES |
|---|
This study concerns the post-caldera peralkaline rhyolites that dominate the Olkaria complex, as did most previous petrological studies (Bailey & Macdonald, 1970
; Davies & Macdonald, 1987; Macdonald et al., 1987, 2002; Wilding et al., 1993; Black et al., 1997; Marshall et al., 1998; Scaillet & Macdonald, 2001, 2003; Heumann & Davies, 2002). These earlier studies tended to concentrate on non-hydrated glassy rocks, to avoid problems from the compositional modifications to which peralkaline rocks are prone during crystallization and secondary hydration (Noble, 1967, 1970; Baker & Henage, 1977). This resulted, at some centres, in a shortage, or absence, of data for some, especially older, stratigraphic units. For example, all the published trace element data from the Gorge Farm centre are for O3/Op3 rocks; none is from the O2 or O4 units. Our sampling strategy has been, therefore, to provide the fullest coverage possible of all units at each centre to address the question of how many compositionally discrete magma batches are present at Olkaria and how they are distributed in space and time.
Electron microprobe analysis (EMPA) was carried out in two laboratories. At Edinburgh University, the Camebax microprobe was used in wavelength-dispersive mode, with andradite as a general standard. An accelerating voltage of 20 kV and probe current of 10 nA were used in most analyses. Analyses of glass, sanidine, amphibole and biotite used a current of 10 nA rastered over 10 µm to minimize alkali migration. A 30 s counting time was used for peaks and 15 s for backgrounds. Mineral analyses were also obtained at the Open University, using a Cameca SX100 electron-microprobe operating in wavelength-dispersion mode. The following standards and X-ray lines were used: synthetic LiF (F K
), jadeite (Na K), forsterite (Mg K), feldspar (Al, Si and K K), synthetic KCl (Cl K), crocoite (Cr K), rutile (Ti K), bustamite (Mn and Ca K), hematite (Fe K), Ni metal (Ni K), willemite (Zn K), barite (Ba L) and SrTiO3 (Sr L). An operating voltage of 20 kV and probe current of 20 nA (measured on a Faraday cage) were used. A beam of 10 µm in diameter was used to minimize volatilization effects. Count times varied from 20 to 80 s per element, and data were corrected using a ‘PAP’ correction procedure (Pouchou & Pichoir, 1985).
A total of 105 rocks was analysed for major elements by X-ray fluorescence (XRF) at the University of Lancaster, using fusion discs. To ensure consistency with the dataset of Macdonald et al. (2008), the rocks were analysed for trace elements by XRF at the University of Edinburgh, using techniques outlined by Fitton et al. (1998). Also for reasons of consistency, we reanalysed for trace elements 34 rocks from Macdonald et al. (1987). We have also used nine analyses of Olkaria rhyolites from Black et al. (1997), analysed in the Edinburgh laboratory. The total dataset comprises 146 analyses.
Laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS) analyses of phenocryst phases and matrix glasses were carried out using a New Wave UV-213 laser ablation system in conjunction with an Agilent 7500a ICP-MS instrument. Samples were presented as thick (100 µm) electron microprobe slides and ablated under an atmosphere of helium. Ablation conditions were: 80 nm laser spot diameter operated at 10 Hz and a laser power of 7–10 J/cm2 and each sample was analysed in spot mode. Ablated material was transported to the plasma source using a gas flow control system at a flow rate of 0·5 l/min and the plasma operated at a power of 1400 W. These conditions produce a beam with an intensity of (1–3) x 104 c.p.s./ppm with a Th+/ThO+ ratio of <<0·3%. Signals from the ablated sample were recorded in time-resolved mode with a dwell time of 0·01 s on each mass over a period of 240 s, during the first 120 s of which the laser shutter remained closed to allow the measurement of the ‘gas blank’. The complete signal was subsequently interpreted using proprietary software (GLITTER) and Ca as the internal standard element. Analyses were calibrated against NIST SRM 612 glass reference material, doped with a nominal concentration of 40 ppm for most trace elements, and using values calibrated against the suite of MPI-DING glasses (Jochum et al., 2006) The values used are close to those reported by Pearce et al. (1997, 2004). Repeat analyses of USGS glasses BHVO-2g and BCR-2g (Supplementary Data Table 3) reveal good agreement with accepted values for the majority of trace elements (Pearce et al., 2004). Differences between determined and accepted values are invariably better than 10% and frequently significantly better than 5%, particularly for elements with masses >87 (Rb). Within-run precision is generally better than 5% (2
), although between-run precision may be slightly greater.
Ion microprobe analyses of feldspar and zircon phenocrysts and matrix glasses were made using the NERC Cameca IMS-4F facility at Edinburgh University. Compositions were determined with reference to the NIST SRM610 glass standard.
Earlier studies (Davies & Macdonald, 1987; Macdonald et al., 1987; Black et al., 1997; Heumann & Davies, 2002) commonly presented Olkaria sample numbers without a prefix. Here we prefix those rocks BL, to avoid confusion with the BB and SMN samples.
|
PHENOCRYST ASSEMBLAGES AND COMPOSITIONS |
|---|
Mineral assemblages (Appendix) and modal analyses of representative rocks (Table 2) have been established from thin sections. Additional modes have been given by Macdonald et al. (1987
) and Marshall et al. (1998). Although the use of thin sections may have resulted, in some cases, in the full phenocryst assemblage in phenocryst-poor rocks not being observed, we preferred thin sections over grain mounts to preserve the textural relationships between grains. Rocks for which thin sections were not available are noted in the Appendix.
|
About 25% of the Olkaria peralkaline rhyolites are aphyric, including all samples from the Kibikoni and Oserian centres and the majority of rocks from the Broad Acres centre and the O5 activity at the Ololbutot centre. Phenocryst abundances in the great majority of remaining specimens range up to 10% by volume, although a few samples from the Gorge Farm centre contain up to 17% (Table 2). There is apparently no systematic relationship between phenocryst abundance and the peralkalinity of the whole-rocks.
The phenocryst phases include sodic sanidine, quartz, fayalite, ferrohedenbergite, titanomagnetite, ilmenite, biotite, amphibole, aenigmatite, zircon, apatite, chevkinite-(Ce) and fluorite. The generalized assemblages present at each centre are given in Table 1 and the assemblage in each sample in the Appendix. Phenocryst assemblages vary with bulk-rock composition and by centre. The ranges of rock compositions in which each phase has been recorded, as measured by the Peralkalinity Index, are shown in Fig. 4. Sodic sanidine, quartz, olivine, FeTi-oxide and chevkinite-(Ce) seem to have been stable over the whole compositional range. Clinopyroxene is, with one exception, found only in rocks with PI < 1·3, whereas amphibole and biotite do not occur in the least peralkaline types. Thus, clinopyroxene–amphibole and clinopyroxene–biotite are incompatible pairs in the Olkaria rhyolites. Aenigmatite has been found in only two samples, from the Gorge Farm centre (Macdonald et al., 1987; Black et al., 1997). Zircon and apatite are restricted to rocks with PI < 1·25. With one exception, fluorite is restricted to the most peralkaline rocks.
|
We noted above that rocks from different centres may show different modal phenocryst abundances. There are also some differences in phenocryst assemblages. Thus, biotite is a common phenocryst in Gorge Farm rocks but, with the exception of one lava from the Arcuate Domes centre (SMN89), is absent from all other centres, even in rocks of comparable bulk composition to the Gorge Farm rocks. No mafic phenocrysts have been found in rocks of the Ololbutot and Olkaria centres. There are also more subtle, intra-centre, differences. Thus rocks of O3 age at the Olkaria centre contain chevkinite-(Ce) but not fluorite, those of O4 age fluorite but not chevkinite-(Ce).
Although there is a generally positive correlation between the number of phenocryst phases and their total modal abundance, rocks with <3% phenocrysts can have seven or eight separate phases (Table 2). Because, as we show below, all phases seem to have been in, or close to, equilibrium with coexisting melt, these rocks represent multiply saturated melts very close to their liquidus. This is consistent with evidence from mineral inclusions; for example, sanidine phenocrysts commonly include clinopyroxene, FeTi-oxides, amphibole, zircon, chevkinite-(Ce) or fluorite. Wilding et al. (1993) showed that glass (melt) inclusions in quartz phenocrysts have the same composition as the matrix glass, indicating that the phenocrysts and residual melt were in equilibrium. In that the rhyolites are both highly fractionated and crystal-poor, the melts must have separated from a more crystal-rich environment (compare the Coso rhyolites, California; Manley & Bacon, 2000).
Alkali feldspar
Alkali feldspar occurs mainly as euhedral or subhedral prisms 0·5–3 mm long, although partially resorbed crystals are not uncommon. The feldspar is sodic sanidine in the range Or36–48 (Table 3; Supplementary Data Table 1), which slightly expands the range reported by Macdonald et al. (1987). The within-sample range of phenocryst core compositions varies from 0–2% Or in nine of 19 samples for which we have data [including the analyses reported by Macdonald et al. (1987)] to 4–8% Or in the other 10. The range of rim compositions is usually lower than that in the cores in the same rock, consistent with the crystals attempting to equilibrate with melt. Crystal rims may be either enriched or depleted in Or relative to cores, the variation usually being <Or3. An exceptional sample is a grain in sample BL570 from the Gorge Farm centre, which is zoned from Or37 (core) to Or56 (rim) (Macdonald et al., 1987). Heumann & Davies (2002) reported that some sanidine separates and a single grain analysis from the Gorge Farm centre are not in Pb isotope equilibrium with the coexisting glasses and suggested that those feldspars had been incorporated from within the volcanic pile during eruption.
|
The CaO contents of the alkali feldspars are extremely low, <0·2 wt% (An < 2). The higher Ca contents tend to occur in the least peralkaline rocks (e.g. BL002, SMN57). Low mineral/glass ratios (0·05–0·12) indicate that Ca is strongly partitioned into the melt during sanidine crystallization. Abundances of Fe2O3* (total Fe as Fe3+) are modest (<1 wt%) and are highest in the most peralkaline, and therefore most Fe-rich, host rocks [compare Mahood & Stimac (1990
), for Pantellerian trachytes and pantellerites].
Alkali distribution between feldspar phenocrysts and glass for seven rocks, representing six centres, is summarized in Fig. 5. Compositions of crystal rims are used rather than cores, as these are more likely to have been in equilibrium with melt; however, the generally small amount of zonation means that any disequilibrium must have been minor. With the exception of BL002, the feldspars are more potassic than the coexisting melt (glass), the normal situation in peralkaline rhyolites (Bailey & Schairer, 1964).
|
Trace element data for alkali feldspar phenocrysts and matrix glasses, determined by LA-ICP-MS and ion microprobe, are given in Table 4 and the full datasets in Supplementary Data Tables 2–4. The compositions used are normally averages because, particularly in the glasses, there is some variation between analyses probably related, inter alia, to proximity to the various phenocryst phases and analytical imprecision at low abundances, and possibly also to minor heterogeneous distribution of trace elements in the magmas. The chondrite-normalized rare earth element (REE) pattern for a representative phenocryst in BL210b is shown in Fig. 6. There is a steady decrease from La to Sm, a small negative Eu anomaly, and a flat pattern from Gd to Lu. The grey field includes all the feldspar data; there is very little change in REE behaviour with increasing host peralkalinity.
|
|
Table 5 presents phenocryst/glass trace element ratios, calculated from the LA-ICP-MS data in Table 4. Because the phenocrysts show relatively little zoning and the degree of crystallization is low (normally <10%), the ratios can reasonably be referred to as apparent partition coefficients. Supplementary Data Table 4 gives feldspar and glass data determined by ion microprobe. Rb and Sr abundances and partition coefficients for feldspar–glass pairs from the Gorge Farm and Ololbutot centres were determined by isotope dilution by Heumann & Davies (2002
) and these data are included here.
|
Apparent partition coefficients for Ba, Rb and Sr are plotted as a function of PI in Fig. 7, along with the generalized trends established by Mahood & Stimac (1990
) for the peralkaline trachyte–pantellerite suite of Pantelleria. Ba coefficients in the Olkaria rocks generally decrease with increasing host peralkalinity, in a similar way to those for the Pantelleria rocks and for peralkaline silicic rocks in general (White et al., 2003). Exceptionally high values for Ba (49, 59), determined by LA-ICP-MS, are found in two specimens from the Gorge Farm centre. K
values also decrease with PI; the majority are lower than in Pantelleria rocks of equivalent PI. This may be partly explained by the strong partitioning of Sr into chevkinite (K
27; Macdonald et al., 2002) and fluorite (K
50–78; Marshall et al., 1998). However, both phases occur in very low modal abundance (Table 2). The two rocks with unusually high apparent partition coefficients for Ba also have high values for Sr (19, 20). The reason for these high values does not seem to be analytical, as the Ba and Sr coefficients for BL210b and SMN57, also determined by LA-ICP-MS, are ‘normal’ at 7 and 4, and 3 and 2, respectively. However, White et al. (2003) have noted that the coefficients for Ba and Sr in salic rocks sometimes do not vary systematically with respect to either crystal or melt chemistry, and they offered as possible complicating factors kinetic effects and elevated Sr contents in accumulated feldspars. Given the evidence, noted above, that Gorge Farm feldspars can be out of Pb isotopic equilibrium with coexisting melt and also show greater zonation than feldspars from other centres, a more complete study of the evolution of the centre may elucidate the reason for the high Ba and Sr coefficients.
|
The majority of apparent partition coefficients for Rb (0·2–0·5) are generally similar to those at Pantelleria when plotted against PI. There is an initial increase as PI increases and then values flatten out at between 0·2 and 0·3 (see White et al., 2003
). Coefficients for the two rocks with high coefficients for Ba and Sr, obtained from LA-ICP-MS, are an order of magnitude lower than those from ion microprobe and isotope dilution, reflecting low Rb values. Our Rb values for the USGS glasses BHVO-2g and BCR2-g are 9·1 ppm and 47·9 ppm, respectively, similar to the published values of the glass from a variety of analytical techniques (Jochum et al., 2005) and close to the accepted values of the corresponding powdered reference materials. We have no reason, therefore, to doubt our LA-ICP-MS data for Rb. As for Ba and Sr (above), the differing coefficients may be a reflection of variations in KD not simply related to whole-rock or mineral composition.
Caesium is essentially excluded from alkali feldspar; apparent partition coefficients in three rocks of 0·02–0·05 are rather higher than those for the pantelleritic trachyte (0·0075) from Mahood & Stimac (1990). Lithium is also strongly incompatible (0·06).
With the exception of Eu, abundances of the REE in sanidine are low, with extremely low partition coefficients (0·04; Table 5). Mahood & Stimac (1990) reported similarly low coefficients for pantelleritic trachytes and pantellerites from Pantelleria, and White et al. (2003) presented KDLasan/gl for six peralkaline vitric samples in the range 0·01–0·03. KDEusan/gl for BL210b, SMN35 and SMN39 (LA-ICP-MS) are 0·37, 0·28 and 0·26, respectively. These values are similar to those for rocks of equivalent peralkalinity on Pantelleria (Mahood & Stimac, 1990) and reinforce the point that Eu is incompatible in alkali feldspar in rocks with sufficiently high PI.
Quartz
Quartz phenocrysts are present in the majority of Olkaria rhyolites, in the size range 0·8–2·0 mm and varying from euhedral to resorbed. The more anhedral/resorbed phenocrysts tend to occur in the less peralkaline rocks (e.g. from the Group 1 and Olenguruoni centres), perhaps a result of decompression crystallization (Blundy & Cashman, 2001), whereas in more peralkaline rocks (e.g. the Gorge Farm and Olkaria centres), quartz tends to be euhedral. Quartz phenocrysts always occur in the same assemblages as sanidine, with which it sometimes forms granophyric intergrowths, rarely up to 5 mm long, especially in rocks of the Gorge Farm centre (Appendix). Lowenstern et al. (1997) and Bachmann et al. (2002) have ascribed the formation of granophyric texture in eruptive rocks to near-instantaneous isothermal undercooling caused by the rapid decompression and devolatilization of magma remaining after eruption of the upper part of the chamber.
Melt and fluid inclusions in quartz phenocrysts are common. The melt inclusions are green and up to 100 µm across. Their composition is similar to the host matrix glass (Wilding et al., 1993). We have no data for the fluid inclusions.
Olivine
Pale amber fayalite phenocrysts occur over almost all the whole-rock compositional range in the Olkaria peralkaline rhyolites. Crystals are up to 2 mm in size and vary from euhedral (SMN29) to embayed (SMN57); inclusions of FeTi-oxide and/or chevkinite-(Ce) are common. Olivine in SMN89 contains melt inclusions and is rimmed by opaque oxide. The compositional range (Supplementary Data Table 6; Macdonald et al., 1987) is very restricted, Fo1–2Fa95–96Tp0·3. CaO abundances are low (<0·25 wt%; Ca < 0·01 a.p.f.u.). Crystals are essentially unzoned. Forsterite and Ca contents decrease with increasing whole-rock peralkalinity.
Trace element data for two analyses of a fayalite phenocryst (Fa97) in SMN39 are given in Table 6 and chondrite-normalized REE patterns shown in Fig. 8a. The analyses differ slightly at the light REE (LREE) end, probably as a result of their abundances being close to detection limits, but both show very strong heavy REE (HREE) enrichment ([La/Yb]CN 0·002, 0·001), strong negative Eu anomalies (Eu/Eu* 0·046, 0·057), and a positive Ce anomaly (Ce/Ce* 1·63, 1·47). There is a steady increase in the apparent partition coefficients towards the HREE and that for Lu approaches unity (Table 5; Fig. 9). The transition elements show increasing compatibility in the sequence V < Sc < Zn < Cr < Co. Li is moderately incompatible (Table 5); all other analysed elements are strongly incompatible.
|
|
Clinopyroxene
Ferrohedenbergite occurs in rocks in the peralkalinity range 1·00–1·27 (Fig. 4). It is usually euhedral, and the colour varies from pale green to green or mottled green/yellow with increasing peralkalinity of the hosts. It is very commonly associated with, or contains as inclusions, (micro)phenocrysts of FeTi-oxides and chevkinite-(Ce). The compositional range is from Ca41·7Mg6·6Fe45·4 to Ca50·3 Mg1·5 Fe53·3 (Table 7; Supplementary Data Table 7), the Fe/Mg ratio increasing with host-rock peralkalinity. The clinopyroxenes are Na- and Ti-poor (Na2O < 1·5 wt%, TiO2 < 0·35 wt%), abundances being slightly higher in more peralkaline rocks. However, all are peralkaline (PI > 1).
|
The chondrite-normalized REE patterns for clinopyroxene phenocrysts in BL210b, with mg-number [100Mg/(Mg + Fe2+)] of 6·4, and SMN57 (mg-number 5·3) show a fairly continuous drop to Er (except for a strong negative Eu anomaly (Eu/Eu* 0·02), and an upward HREE tail (Fig. 8b). The REE mineral–glass partition coefficients are shown in Fig. 9. The slightly higher values for SMN57 may reflect the more peralkaline, and thus depolymerized, nature of the host-rock (PI 1·16 and 1·05, respectively). Maximum partitioning is shown by Sm, Yb and Lu. Strontium is incompatible in BL210b but compatible in SMN57. The potential of clinopyroxene to slightly fractionate Zr (and Hf) from Nb (and Ta) should be noted.
|
Amphibole
Amphibole forms euhedral (Gorge Farm centre) or subhedral to embayed (Arcuate Domes centre) phenocrysts, varying in form from euhedral laths to rod-shaped and up to 0·5 mm long, and also occurs as inclusions in sanidine and quartz. Amphibole composition (Table 8; Supplementary Data Table 8) generally changes with increasing whole-rock peralkalinity (PI 1·26–1·44) from kataphorite to richterite to silica-poor riebeckite. The colour correspondingly changes from brown–green to blue–green. There is, in some samples, variation in core compositions; SMN1 (Ololbutot centre) cores include potassian ferroricherite, ferrikatophorite and ferrobarroisite, and BL565 (Gorge Farm) has cores of magnesio-richterite, ferririchterite and magnesio-arfvedsonite composition. Fluorine levels do not exceed 2 wt%. The partitioning of F between amphibole and melt in two rocks is 2·0 (SMN87) and 1·2 (SMN70), the decrease with increasing whole-rock peralkalinity mirroring that in the amphiboles crystallized experimentally from Olkaria rocks (Scaillet & Macdonald, 2003
). Chlorine contents are remarkably low (0·08 wt%), even though melt Cl contents reach 0·52 wt% (Supplementary Data Table 14), a point noted for the natural rocks by Macdonald et al. (1987) and for the experimental amphiboles by Scaillet & Macdonald (2003).
|
Apparent partition coefficients and element abundances for amphiboles in two rocks are presented in Tables 5 and 6 and chondrite-normalized REE patterns, selected to show the range in [La]CN, in Fig. 8b. There is more element scatter within and between phenocrysts than shown by the other mafic phases, consistent with the range in major element compositions. Overall, the REE patterns are rather similar to those for clinopyroxene at lower REE abundances, but the upward HREE tail is more pronounced ([Ho/Lu]CN 0·14–0·40, as opposed to 0·51–0·55 in clinopyroxene). All analyses show a small, but persistent, positive Ce anomaly (Ce/Ce* 1·02–1·09). The REE partitioning pattern (Fig. 9) shows a gentle increase to Nd, a slight decrease to Ho and an upward HREE tail. Only Lu has an apparent partition coefficient exceeding unity. Unlike the other mafic phases, the amphiboles do not fractionate Eu relative to adjacent elements. The transition elements and Li, Sr and Ge are compatible. The amphiboles could potentially fractionate Ta from Nb, but not Hf from Zr or U from Th.
Biotite
Biotite forms 0·5–1·5 mm long, subhedral to euhedral phenocrysts in rocks from the Gorge Farm and Arcuate Domes centres (Appendix). Our new data (Table 8; Supplementary Data Table 9) confirm the observation of Macdonald et al. (1987) that there is little compositional variation. Fe/(Fe + Mn + Mg) ratios range from 0·92 to 0·97 in biotites from the Gorge Farm rocks and are slightly lower in those from the Arcuate Domes (0·87–0·89), reflecting the more Fe-rich nature of the Gorge Farm host-rocks. Titanium varies from 0·4 to 0·5 a.p.f.u. and F from 0·8 to 1·2. The biotites are Al-poor, having insufficient Al to fill, with Si, the tetrahedral site. Compositionally, they are broadly similar to the biotites synthesized from BL002 by Scaillet & Macdonald (2001, 2003) but we have not found in the natural rocks the tetrasilicic mica montdorite, recorded by them in the experimental products of SMN49.
Chondrite-normalized REE patterns for phenocrysts in SMN39 (representative analysis in Fig. 8a) show modest LREE enrichment ([La/Yb]CN 3), strong Eu anomalies (Eu/Eu* 0·1) and modest negative Ce anomalies (Ce/Ce* 0·5–0·9). Partition coefficients for the REE are very low but the biotite concentrates Co, Cr, V, Zn, Li, Ba, Rb, Ga and Ge relative to the coexisting glass (Table 5). Biotite is the only phenocryst phase into which Cs enters in significant amounts, with an apparent partition coefficient of 0·58.
Fe–Ti oxides
Coexisting spinel and rhombohedral phases have not been observed in the Olkaria rhyolites (see Macdonald et al., 1987), which is the normal situation in peralkaline rhyolites (Nicholls & Carmichael, 1969). Both phases generally form euhedral grains up to 0·3 mm long. Ilmenite also occurs as inclusions in clinopyroxene (SMN 56, SMN57) and as elongate, embayed needles up to 0·8 mm long in some rocks from the Olenguruoni and Arcuate Domes centres. The ulvöspinel component in the cores of titanomagnetite phenocrysts ranges from 37 to 56 mol% (Table 9; Supplementary Data Table 10), generally decreasing with increasing PI, and thus Ti/Fe ratio, of the host-rocks. The Mn, Mg and Al contents are low, 0·03 a.p.f.u. Zoning is slight, with rimward decreases in ulvöspinel (<3%). Ilmenite core compositions vary little (Xilm 97–95) and phenocrysts are essentially unzoned (Table 9; Supplementary Data Table 11).
|
We have insufficient data to predict which oxide phase will crystallize from a given melt composition. Ilmenite occurs in rocks with PI ranging from 1·02 to 1·34, titanomagnetite in rocks with PI 1·05–1·39 (Fig. 4). Both phases coexist with varying combinations of olivine, clinopyroxene, amphibole and biotite. Whereas BL002 contains titanomagnetite phenocrysts, Scaillet & Macdonald (2001
, 2003) synthesized ilmenite from the same rock at low fO2. Whatever the control over which phase crystallizes, it is clear that in the complex P–T–X–fO2 space occupied by the Olkaria rhyolite magmas, ilmenite and titanomagnetite have separate stability fields. Trace element abundances for ilmenite are given in Table 6 and Supplementary Data Table 5 and chondrite-normalized REE patterns in Fig. 8a. Abundances are very low, the LREE and middle REE (MREE) being below detection in SMN39. The pattern for SMN57 shows a decrease from La to Gd, with a strong Eu anomaly (Eu/Eu* 0·08), and then marked enrichment in HREE. Apparent partition coefficients exceed 0·01 only for Tb–Lu in SMN57 (maximum 0·24). Co, Cr, V and Zn are strongly compatible. Ilmenite strongly fractionates Ta from Nb.
Apatite
Apatite-phyric rocks have PI in the range 1·03–1·24 (Fig. 4). The mineral is present as rods up to 60 µm long and 25 µm across, either as discrete crystals or associated with chevkinite-(Ce) and zircon. Marshall (1999) presented incomplete analyses of apatite in BL002 and SMN23; analytical totals are low (90–94 wt%) and high F values (exceeding the maximum 2 a.p.f.u.) may indicate that matrix glass was irradiated during analysis. The data indicate, however, that the apatites contain 15% of the britholite component and are thus closely similar to apatite microphenocrysts in the most evolved Olkaria trachytes (Macdonald et al., 2008).
Zircon
Zircon occurs as microphenocrysts, exceptionally up to 0·6 mm long and only infrequently euhedral, in rocks with low PI (Fig. 4); it is found, for example, in 73% of the Group 1 and Ndabibi rocks collected, commonly associated with apatite and titanomagnetite. It forms inclusions in sanidine in BL143b.
Trace element abundances and apparent partition coefficients for zircon from two rocks are given in Table 10. The very strong HREE enrichment relative to LREE, of Zr relative to Hf, and of U relative to Th (K
/ K
D z/gl 6.5) should be noted. Chondrite-normalized REE patterns (Fig. 10) show a strong enrichment from LREE to HREE, with marked positive Ce and negative Eu anomalies. This pattern has been recorded in other igneous zircons (compilation by Hanchar & van Westrenen, 2007). The Ce anomalies are not found in the coexisting glasses. REE partition coefficients show a five orders-of-magnitude increase between La and Lu (Fig. 10).
|
|
Fluorite, chevkinite and aenigmatite
Full descriptions of the Olkaria fluorite phenocrysts, including detailed textural evidence regarding their phyric nature, were given by Marshall et al. (1998
). In their experimental study of Olkaria rhyolites, Scaillet & Macdonald (2001) found that fluorite crystallizes at temperatures in excess of 800°C at moderate melt water contents and that the fluorite-in curve is, in places, less than 20°C below that of the liquidus phase. Fluorite can be, therefore, a stable high-temperature phase in peralkaline rhyolites (Scaillet & Macdonald, 2004). Macdonald et al. (2002) described chevkinite-group minerals occurring as phenocrysts in the Olkaria rhyolites; according to the classification scheme of Macdonald & Belkin (2002), these phases are chevkinite-(Ce). We have a new analysis of chevkinite-(Ce) in BL210b (Supplementary Data Table 12), which is very similar compositionally to that in BL002, also from the Ndabibi centre (Macdonald et al., 2002). The new analysis allows us to comment on apparent partition coefficients for a wider range of elements. The transition metals are strongly compatible, especially Sc and V. The partition coefficients for Li and Sr are 0·2 and 2·6, respectively. Those for Nb and Ta are 79 and 42, respectively, the Nb value being higher than in BL002 (56). Macdonald et al. (2002) recorded partition coefficients of six and five for Zr and Hf, respectively, in BL002. In BL210b, the values are 7 and 12. The REE partition coefficients in BL210b are generally rather lower than those for BL002, except for Tm–Lu which are slightly higher. Apparent partition coefficients for Th and U are higher in BL210b but the ratio of the two values is the same.
Macdonald et al. (1987) and Black et al. (1997) reported the occurrence of aenigmatite phenocrysts in two peralkaline rhyolites from the Gorge Farm centre (BL570, BL575). These remain the only Olkaria rhyolites in which the phase has been found, although it is common in the pantellerites erupted earlier in the complex's history. Scaillet & Macdonald (2001) were unable to synthesize aenigmatite in BL575 and suggested that it is stable at oxygen fugacities lower (below FMQ) than those imposed in their experiments. From a review of experimental data, Kunzmann (1999) also suggested that aenigmatite stability is also restricted to low oxygen fugacities.
Xenocrysts
Considering the common occurrence of mixed magma rocks at Olkaria (Macdonald et al., 2008), the rhyolites are remarkably devoid of xenocrysts. This is consistent with the effective homogeneity of the major element compositions of the matrix glasses, within-specimen variability normally being within analytical error (Supplementary Data Table 14). Such homogeneity indicates crystal-limited equilibration scales (Pichavant et al., 2007). The scarcity of xenocrysts contrasts, for example, with the situation in the Pleistocene Coso volcanic field in California, where several rhyolite domes carry up to seven xenocryst phases (Manley & Bacon, 2000). Generally, however, it seems that viscosity and/or temperature differences between the inclusions and the host rhyolites were sufficiently large to limit effective mixing between them. A few Olkaria rocks show a range of up to Or8 in sanidine core compositions, some of which may be xenocrystic. We cannot preclude some limited rhyolite–rhyolite mixing within rocks from each centre. An unusual, 0·6 mm long, crystal in BL210b is reversely zoned from Or41 to Or–15 (Table 3). Heumann & Davies (2002) noted that some bulk sanidine separates from the Gorge Farm centre have Pb isotopic compositions that are not in equilibrium with the coexisting glasses and also reported fayalite and biotite phenocrysts falling off U–Th internal isochrons in BL570, again consistent with a xenocrystic origin.
|
GEOTHERMOBAROMETRY |
|---|
Scaillet & Macdonald (2001
) determined the phase relationships in three Olkaria obsidians (BL002, BL575 and SMN49; Appendix) at 150 and 50 MPa, at oxygen fugacities of NNO –1·6 and NNO + 3·6 (where NNO is the nickel–nickel oxide buffer), between 800 and 660°C, and at melt water contents ranging from nominally anhydrous to saturation. They found the closest matches between the natural and experimental mineral assemblages when fO2 was at NNO –1·6 (i.e. at or below FMQ), temperatures were between 740 and 660°C, and melt evolution occurred under near water saturation conditions. Ion microprobe analyses of glass inclusions in quartz phenocrysts showed pre-eruptive water contents up to 3·4 wt% (Wilding et al., 1993). However, by restoring the pre-eruptive degassing melt water contents using measured 
D and matrix H2O values of glassy rocks, Wilding et al. (1993) estimated pre-eruptive values up to 5·7 wt%. Equilibration pressures were estimated by Scaillet & Macdonald (2001) to be little higher, and probably lower, than 150 MPa. The inferred depth range (5 km) is comparable with that (5–6 km) estimated from seismic data for the interface between the Pan African basement and the Miocene–Holocene volcanoclastic rift infill beneath Olkaria (Mooney & Christensen, 1994; Mechie et al., 1997).
Equilibration temperatures calculated from the rim compositions of several olivine–clinopyroxene phenocryst pairs in SMN57 using the QUILF program (Andersen et al., 1993) yielded values of 680–677°C at 1–2 kbar, consistent with the experimental results cited above. Zircon saturation thermometry (Hanchar & Watson, 2003), using ion microprobe data (Table 10), gives temperatures of 881°C and 882°C for BL002 and SMN59, respectively. This estimate is considerably higher than the QUILF and experimental results, and it may be that the geothermometer is not well calibrated for near water saturated, halogen-rich, peralkaline compositions.
|
GEOCHEMISTRY |
|---|
General features
Representative whole-rock major and trace element analyses are given in Table 11, and the full data set in Supplementary Data Table 13. Peralkaline rocks are prone to compositional modification during crystallization and/or secondary hydration, especially loss of Na (Noble, 1967
, 1970; Baker & Henage, 1977). We have data for four glass–crystalline pairs from the same flows (SMN174a–SMN174b; SMN29–SMN204; SMN122–SMN123; SMN205–SMN206). They show relative Na2O losses from the crystalline facies of between 3 and 13 wt%. High loss on ignition (LOI) values (up to 3·73 wt%) in pumices of Op3 age from the Gorge Farm centre (e.g. SMN70) are associated with low Na2O abundances, strongly indicative of Na loss during secondary hydration. Loss of Na2O lowers the PI, which we use as an important measure of the degree of melt evolution. To counter this general problem, White et al. (2003) introduced an alternative measure of peralkalinity, FK/A [mol. (Fe + K)/Al, with all Fe calculated as Fe2+], where FeO*, K2O and Al2O3 are considered to be relatively immobile. There is an excellent correlation between PI and FK/A in the glassy Olkaria rocks (r2 = 0·93). We have accordingly adjusted the PI of the crystalline and secondarily hydrated rocks on the basis of their FK/A value.
|
The Olkaria peralkaline rhyolites are comendites in the classification scheme of Macdonald (1974
). The PI ranges from 1·00 to 1·44, with the exception of a pumice from the Gorge Farm centre (SMN70; PI = 1·55). A notable feature is the extreme depletion or enrichment of certain trace elements (Davies & Macdonald, 1987; Macdonald et al., 1987; Clarke et al., 1990; Black et al., 1997; Heumann & Davies, 2002; this paper). For example, Ba concentrations range from 23 to <1 ppm and Sr from 17 to <1 ppm. Among enriched elements, Zr varies in the range 442–3640 ppm, Nb 178–1022 ppm, Rb 262–1056 ppm, and Y 36–507 ppm. The rhyolites are also halogen-rich (F 0·95 wt% and Cl 0·47 wt%; Macdonald et al., 1987). New LA-ICP-MS analyses of matrix glasses in four rocks (Table 4 and Supplementary Data Table 3) show comparable trace element abundances to the whole-rocks.
It would be useful at this stage to comment on what we refer to as the ‘degree of evolution’ in the Olkaria rhyolites, and indeed in any rocks of similar composition. For many petrologists, the term more evolved would be synonymous with higher silica contents. For others, it would be related to increased ITE abundances and to still others it would mean the product of greater degrees of fractionation from some putative parental magma. The experimental work of Scaillet & Macdonald (2001, 2003) confirmed the modelling results of Macdonald et al. (1987) on the natural samples that crystallization in the Olkaria rhyolites generated residual melts with lower SiO2 contents and higher values of PI. Furthermore, although there is a good overall correlation between PI and ITE, there are inter-centre differences. Thus, rhyolites of the Gorge Farm centre generally have higher ITE contents than rocks of the same PI from other centres. In the following, we use the term more evolved specifically to mean more peralkaline (and thus more Fe-rich; Fig. 11a), in that PI most effectively measures the degree to which major element compositions have changed from those of the inferred parental magmas.
|
Internal compositional variations
With increasing PI, there are overall increases in Na2O, FeO* and TiO2 (scattered), and decreases in SiO2 and Al2O3 (Fig. 11a), whereas K2O is essentially constant. The ITE all show positive correlations with PI. Maximum enrichment factors, over rocks of all ages from all centres, are: Zr, Pb
9; Nb, Th, Zn, Eu, Y 5/6; and La, Ce, Rb 4. Thus ITE ratios vary with changing major element composition. Figure 12 shows how the compositional range varies with stratigraphic age. Whereas the earliest-erupted rocks (unit N) show a relatively restricted range, rocks of O2 age cover more than half of the total range. Maximum levels of peralkalinity and Zr were achieved during O3/Op3 and O4 times. O5 rocks are on average somewhat less evolved (less peralkaline). Rocks with PI < 1·2 have been erupted until O4 times and it seems likely that such compositions are still present in the system.
|
The broad geochemical trends mask considerable complexity related to inter-centre and intra-centre differences. A critical feature of the Olkaria rocks is that single centres are geochemically, as well as petrographically (Table 1), distinct, as shown by trace and minor element abundances and ratios (Macdonald et al., 1987
), isotopic ratios (Davies & Macdonald, 1987; Heumann & Davies, 2002), U–Th isotope systematics (Black et al., 1997; Heumann & Davies, 2002), and pre-eruptive water contents and inferred degassing histories (Wilding et al., 1993). For example, the clear separation of rocks from the Gorge Farm and Ololbutot centres on the plot of peralkalinity index against Rb (Fig. 11b) should be noted. The differences can be very subtle. Rocks from the Oserian and Kibikoni centres are all aphyric and of O2 age. They cannot be distinguished on most plots (e.g. Fig. 11a and b). However, they occupy slightly different fields on, for example, FeO*–Zr and Zr–Nb plots (not shown) and, although data are scarce, they seem to have different Sr and Nd isotopic ratios (Davies & Macdonald, 1987).
Compositional differences can also be distinguished in the products of single stratigraphic units within centres. For example, O4 eruptive rocks at the Olenguruoni centre form a main dome and two smaller domes. One of the smaller domes has K/Rb ratios (<130) comparable with other Olkaria rocks of similar PI and FeO* contents. The main dome and the other smaller dome have ratios (150) that are unique in the complex.
In Table 1 we list FeO* and Zr (as proxies for degree of evolution; Figs 11a, b and 12) for the rocks of each centre, relating them to specific stratigraphic units. The most significant features, here shown through selected examples, are as follows.
- Relatively unevolved rocks (e.g. PI < 1·15; FeO* < 2·5 wt%) form the only eruptive products at the Ndabibi centre and in Group 1. They are also present at some other centres, sometimes having been erupted early in the history of each centre (O2; Olenguruoni, Olkaria, Arcuate Domes) but sometimes as late as O3 (Olenguruoni, Olkaria, Olobutot) or O4 (Olenguruoni). Such magmas may still underlie the whole, or parts, of the complex.
- Some centres show reversals in the degree of magma evolution with time; for example, the Olkaria centre has the least evolved rocks in the middle of the sequence (O3). In such cases, the evolved magmas must have been completely erupted, allowing less evolved melts to ascend in a later phase of activity.
- Rocks of one stratigraphic unit can be compositionally restricted (e.g. Ololbutot O5), or can span (almost) the full major element compositional range (e.g. O2 at Olenguruoni).
- O4 and O5 rocks mainly show evolved compositions, with the exception of the O4 rocks of the Olenguruoni centre (e.g. Zr 437–1031 ppm).
- Successive activity at some centres involved a subtle change in chemistry; for example, magmas of O4 age at Olenguruoni evolved along a trend of higher Fe/Zr ratio than older magmas (O2–O3). Op3 and O4 rocks at Gorge Farm cover the same FeO* range as O3 rocks but have higher ITE abundances.
- At any given time, different centres were erupting magmas of different degrees of evolution; for example, the Olobutot centre erupted relatively unevolved magmas during O3 times, whereas O3 rocks at the Gorge Farm centre are among the most evolved in the Olkaria complex.
It is clear, therefore, that not only was each centre evolving separately from, and at different rates from, the other centres but that successive magma batches at each centre frequently evolved along slightly different trends. The fact that each centre is petrographically and/or compositionally distinct from the others makes it essentially impossible that the centres were tapping the same rhyolitic reservoirs, even at different times in the development of those reservoirs.
|
PETROGENESIS |
|---|
Role of crystal fractionation
Macdonald et al. (2008
) showed that the least evolved peralkaline rhyolites at Olkaria were probably formed by fractional crystallization of trachytic magmas, themselves derived by fractionation of basaltic magmas via mugearites and benmoreites. In this section, we examine the role of crystal fractionation in generating the more evolved rhyolites.
On the basis of least-squares modelling, Macdonald et al. (1987) showed that the major element compositions of the most peralkaline rocks could have been formed by 83% crystallization of an alkali feldspar (56%), quartz (24%), clinopyroxene (2%), titanomagnetite (0·4%) assemblage from the mildly peralkaline rhyolites of Group 1 type. Scaillet & Macdonald (2003) determined experimentally that melts with PI > 1·3 (e.g. BL575) can be produced by 75% crystallization of the least evolved rhyolites, exemplified by BL002. There are evidently no major element constraints to the Olkaria suite having been formed by closed-system fractional crystallization dominated by alkali feldspar and quartz.
On a plot of total Fe as FeO vs Al2O3 (Fig. 13), the glass compositions from the Scaillet & Macdonald (2003) experiments are matched very closely by the whole-rock compositions. However, tie-lines between whole-rocks and the matrix glasses of natural samples are generally at a high angle to the overall whole-rock trend and some matrix glasses are lower in FeO*, for a given Al2O3 value, than any rock or experimental glass (Fig. 13). The potential fractionation paths indicated by the tie-lines cannot be those followed by the rhyolitic magmas. This strongly suggests that melt evolution occurred at two or more levels; a deeper level where crystal fractionation generated the range of rhyolitic magmas, and shallower levels where phenocrysts formed but generally were not separated from the melt. Thus, the compositional range was generally established prior to crystallization of the observed phenocrysts, as was noted for the Bishop Tuff, California, by Hildreth (1979) and substantiated there by more recent work (Hildreth & Wilson, 2007).
|
We shall return to the question of the episodes of crystallization below. Here we examine the role of fractional crystallization using trace element abundances and ratios. Major element models (Macdonald et al., 1987
; Scaillet & Macdonald, 2001, 2003) assumed that the rocks of Group 1 and the Ndabibi centre are good proxies for the parental, mildly peralkaline, magmas. In trace element terms, they are less suitable as parents, having different Zr/Nb, Zr/Rb and La/Yb ratios from the rest of the suite (Macdonald et al., 1987). Critically, Zr/Hf ratios in Group 1 and Ndabibi rocks and in the matrix glass from BL210b are in the range 29·5–32, whereas those in rocks and matrix glasses from all the other centres are between 33 and 39 (Macdonald et al., 1987; Supplementary Data Table 3). Because the Group 1 and Ndabibi rocks contain phenocrysts of zircon, fractional crystallization would lower, not raise, Zr/Hf ratios in residual melts. In the trace element modelling, therefore, we have used a parental composition that is representative of the less-evolved compositions from the other centres.
We use simple mass balance to model trace element distribution between end-members that are about median values for the data spread at PI 1·05 and 1·36 (Fig. 11b). The following assemblage and modal proportions were employed, based on data in Table 2: alkali feldspar (0·6)–quartz (0·3)–fayalite (0·02)–clinopyroxene (0·06)–titanomagnetite (0·035)–chevkinite-(Ce) (0·0001). Mineral trace element abundances were taken from Tables 4 and 6. We did not include zircon in the assemblage because it has been recorded in only two post-Ndabibi or Group 1 rocks. The degree of crystallization required to produce the more evolved rock was taken to be the 75% determined experimentally by Scaillet & Macdonald (2003). Element abundances predicted by the calculations are given in Table 12. The major features are as follows.
- The increase in Zr is consistent with the idea that the more evolved rocks were derived from parental rhyolites not saturated in zircon and that Zr was totally incompatible. Of the other phases, only clinopyroxene and chevkinite-(Ce) contain significant amounts of Zr but both are in low modal abundance.
- The chosen assemblage satisfactorily explains the overall increases in Rb, Th, U and Y and the decrease in Sr.
- That La (and the other LREE) have been less incompatible than Zr can be explained by the fact that any fractionation of chevkinite-(Ce) in amounts approaching its modal proportions (up to 0·15%; Macdonald et al., 2002) depletes La in residual melts. This is clearly shown on a Zr–La plot (Fig. 14), where the more evolved (Zr-rich) rocks of the Olenguruoni centre show relative La depletion. It can be no coincidence that these rocks include the highest proportion of chevkinite-phyric rocks in the Olkaria complex (Appendix). The modelled proportion of chevkinite-(Ce) is lower than its modal amounts, by an order of magnitude, perhaps because chevkinite abundances were variable in the fractionating assemblages.
- The reason for the rather high calculated value for Nb is that we omitted ilmenite from the fractionating assemblage. Although there is an overall positive correlation between Nb and Zr, in detail there are compositional ranges where Nb stays about constant at increasing Zr; for example, at Nb 300 (Fig. 15). It may be that ilmenite was present in the phenocryst assemblage.
- The calculations result in a negative value for Ba. To obtain a residual melt value of 3 ppm, the feldspar Ba content would have to be 54 ppm. Either the LA-ICP-MS feldspar values (Table 4) are too high or they are not representative of feldspars in the parental magmas. Macdonald et al. (1987) reported a Ba concentration of 35·8 ppm in sanidine phenocrysts from BL333 from the Olenguruoni centre.
|
|
|
Macdonald et al. (2008
) calculated that the more evolved Olkaria trachytes formed by 92% crystallization of parental basaltic magmas. A further 26·5% (by volume) crystallization of trachyte generated a peralkaline rhyolitic melt comparable with the least evolved rhyolitic eruptive rocks. We have shown here that trace element abundances are broadly compatible with the 75% crystallization value determined by Scaillet & Macdonald (2003) for formation of the most from the least evolved rhyolites. Thus, 1·5% of the parental basaltic melts remained. This represents an extremely long liquid line of descent, which we assume was made possible by the very high mass flux through the system, the heating of more evolved magmas by mixing with mafic melts, and the low viscosity of the rhyolites resulting from high water and halogen contents and their peralkaline nature. At least during the Op3 stage at the Gorge Farm centre, the fractionation path was even more extended. Pumice SMN70 contains 3640 ppm Zr, which implies a further 42% of closed-system fractionation from a magma with 2100 ppm (Table 12).
|
DISCUSSION |
|---|
Evolution of the Olkaria peralkaline rhyolite field
Macdonald et al. (2008
) cited the presence in the complex of a continuous sequence of rocks from basalt to rhyolite, the unusual degree of Sr and Ba depletion even in the least peralkaline rhyolites, and geochemical modelling as evidence for derivation of the rhyolites by extreme crystal fractionation of basaltic magma via mugearitic, benmoreitic and trachytic compositions. The ubiquity in the post-caldera rhyolites of magmatic inclusions ranging from mugearite to trachyte, the occasional eruption of trachyte within the complex [e.g. an O3 flow in the Arcuate Domes (Clarke et al., 1990) and the outer skin of the Central Tower plug] and recent eruption of basalts and mixed basalt–rhyolite flows on the Ndabibi plains (Fig. 3) strongly suggest that the full range of magmatic compositions still underlies the complex.
The complex displays a shadow zone, where propagation of mafic magma is prevented by a low-density layer resulting from the crust becoming ductile and/or partially molten. The question arises as to the depth of this layer. The experimental results of Scaillet & Macdonald (2001) suggest that for the more peralkaline magmas, the pressure of the storage zone is probably >50 MPa, as indicated by the occurrence in the rocks of near-liquidus biotite. However, the pressure is probably little more than 150 MPa, as further increase in pressure would make it difficult to preserve the liquidus phase relationships observed in the rocks. We noted earlier that the inferred storage depth (5 km) is comparable with that (5–6 km) estimated from seismic data for the interface between the Pan African basement and the Miocene–Holocene volcanoclastic rift infill beneath the complex, and magmas may have stalled at this density interface (Fig. 16). Macdonald et al. (2008) raised the possibility that the magma chamber that fed the caldera-forming (O1) eruptions was located at about this crustal level. The O1 pantellerites are too poorly exposed to be able to reconstruct the nature of the chamber and the caldera-forming events. We may, however, speculate that caldera collapse followed eruptions from many, small, partially interconnected reservoirs rather than one major chamber and that the location of these reservoirs guided the location of post-caldera trachytic chambers.
|
We assume, therefore, that the storage zone contained the trachytic reservoirs from which the rhyolites were erupted. It is noteworthy that although their alkali feldspar phenocrysts are usually strongly resorbed and some rocks are hybrids of trachyte and benmoreite, the trachyte lavas are much less mixed than the mugearites and benmoreites, suggesting that they were heated by underlying mafic–intermediate magma but penetrated by it only during eruptive phases. The trachytes thus have acted as a buffer zone between the intense mixing at greater depths and the less intense mixing shown in the rhyolites. We have assumed in Fig. 16 that mafic and/or intermediate magmas underlie trachyte in the storage areas. There is a considerable amount of stratigraphic, petrographic and geochemical evidence that each centre has evolved separately from its neighbours. This requires that each centre has its own conduit, of dyke-like form at least close to the surface (Fig. 16). There must be as many trachytic reservoirs as centres, which is consistent with the range of compositions shown by the eruptive trachytes; for example, in the degree of silica and alumina undersaturation (Macdonald et al., 2008
).
Judging from the phenocryst-rich nature of the erupted trachytes (32%: Macdonald et al., 2008), the reservoirs are probably filled by a crystal mush. A metaluminous trachyte from Olkaria contains 25% by volume of a mildly peralkaline rhyolitic matrix glass (Macdonald et al., 2008) and this may be typical of the amount of melt in the reservoirs prior to eruption of the mildly peralkaline rhyolites. The estimate is consistent with the model of Bachmann & Bergantz (2004), who considered the volume of mush in such reservoirs to be as high as 70–80%. Segregation rates from such a crystal-rich source would be reduced by the high crystallinity but enhanced by the ultra-low viscosity of the melts and, in the later stages, by volatile saturation and exsolution leading to overpressure of the gas phase driving melt out of the porous matrix (Sisson & Bacon, 1999; Bachmann & Bergantz, 2004; Simon & Reid, 2005). Extraction might also have been aided by melt channelization through fractures in the country rocks related to caldera formation.
It is unclear whether the more peralkaline rhyolites were formed by continued crystallization in the inferred trachytic reservoirs. The most evolved rocks would have required the reservoirs to have been 95% crystallized if the trachytes were metaluminous and it seems unlikely that expulsion of melt from a crystal mush would have been possible. However, if the parental trachytes were peralkaline, as are some erupted trachytes in the complex (Macdonald et al., 2008), the rhyolites would have achieved higher PI values at lower levels of trachyte crystallization. Also, trachyte crystallization may have resulted in the formation of wall and/or floor cumulates rather than a mush, facilitating melt removal, or a further period (or periods) of crystallization occurred en route to the surface, either against conduit walls or in shallow reservoirs.
Some 25% of the Olkaria rhyolites are aphyric and the majority of the remainder are phenocryst-poor (<10% modally; Appendix). The compositions of the aphyric rocks completely overlap those of the porphyritic varieties (Table 1). We have suggested above, on the basis of whole-rock–matrix glass relationships, that the observed phenocryst assemblages were not always the fractionating assemblages. These features perhaps indicate that the rhyolitic melts were aphyric when they were expelled from the trachytic, or shallower, reservoirs and that they experienced different post-extraction histories. Some, such as the aphyric rocks from Group 1 and the Ndabibi, Oserian and Kibikoni centres may have been erupted without further storage (Fig. 16). Others, such as the rocks of the Ololbutot and Olkaria centres, were erupted after crystallization of alkali feldspar and quartz but before mafic phases formed. Still others, close to saturation with up to eight or nine phases (Appendix), must have crystallized in a storage area or areas at lower pressure than the low-density trap envisaged for the trachytes. In the case of the O3/Op3 units at the Gorge Farm centre, higher-level storage sometimes lasted sufficiently long for compositional zonation to develop by continued crystal fractionation, with volatile- and ITE-enriched, more strongly peralkaline, upper zones, which tended to be erupted as pyroclastic rocks (Op3), overlying less evolved magmas, which formed lavas and domes (O3). The compositional zonation must have developed rapidly, in a few thousand years at most. The Gorge Farm rocks also show disequilibrium features in the phenocrysts, perhaps indicating mixing of magma batches and/or mush remobilization in the inferred higher-level reservoir.
There is some information available on the longevity of the Olkaria silicic magma reservoirs. A major fractionation event affecting the rhyolites from the Gorge Farm centre is defined by a Rb–Sr glass isochron of 22 ± 4 ka, a value substantiated by an internal 230Th–238U isochron of 24 ± 1 ka (Heumann & Davies, 2002). An earlier phase of magma evolution (47 ± 0·2 ka) is suggested by fayalite–glass ages from Gorge Farm samples but requires further study (Heumann & Davies, 2002). Thus eruption ages at the Gorge Farm centre post-dated fractionation events by between 39 and 16 kyr; thus we can infer that the Olkaria rhyolitic system is at least 40 kyr old. Our interpretation of the crystallization histories of the rhyolites implies that the fractionation events refer at least partly to processes in the trachytic reservoirs.
Olkaria is a hot system; temperatures >300°C have been measured in exploration wells drilled to depths of 1000–2600 m (Omenda, 1998). It seems very likely that the magmatic system has been maintained by the heat and mass flux provided by the underlying basaltic–intermediate magmas. Hildreth & Wilson (2007) introduced the idea that the bottom of large silicic magma chambers, where heat and/or magma replenishment is concentrated, resembles a hot plate. Given the dyke-like form of the Olkaria reservoirs, the relevant comparison may be to a hot poker.
Finally, as reviewed by Hildreth (2004) and Hildreth & Wilson (2007), high-silica rhyolites are widely accepted as liquids expelled from voluminous crystal mushes that eventually crystallize to granitoid plutons. Olkaria may be providing evidence of expulsion of peralkaline rhyolites from small mush zones, which are likely to solidify as sills or bosses.
|
CONCLUSIONS |
|---|
- The Olkaria complex is a young (20 ka), small-volume (13 km3) system dominated by the eruption of peralkaline rhyolites from 13 petrographically and geochemically distinct centres.
- Phenocryst assemblages and compositions vary systematically with whole-rock composition, suggesting, in tandem with homogeneous matrix glass compositions, that the crystals and melts were generally close to being in equilibrium.
- Apparent partition coefficients vary fairly systematically with whole-rock composition, with the exception of some unusually high Ba and Sr, and low Rb, coefficients for sanidine from the Gorge Farm centre.
- The compositional range of the rhyolites was generated by fractional crystallization of alkali feldspar–quartz-dominated assemblages from mildly peralkaline rhyolites, themselves generated by fractionation of trachytic magmas, probably ranging in composition from metaluminous to peralkaline.
- Non-systematic whole-rock compositional variations were caused by inter-centre and intra-centre differences, temporal changes and changes in fractionating assemblages. The whole Olkaria rhyolite suite thus represents multiple liquid lines of descent and each centre records a different evolutionary history.
- The observed phenocryst assemblages were generally similar to, but not the same as, the fractionating assemblages.
|
SUPPLEMENTARY DATA |
|---|
Supplementary data for this paper are available at Journal of Petrology online.
|
APPENDIX: SAMPLE DETAILS |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
*Crystalline or secondarily hydrated.
obs, obsidian; cryst., crystalline; pum, pumiceous; devit., devitrified; hydroth., hydrothermally; alt., altered. ab, amphibole; ae, aenigmatite; af, alkali feldspar; bi, biotite; ch, chevkinite; cpx, clinopyroxene; fl, fluorite; ol, olivine; ox, FeTi-oxides; q, quartz; qf, quartz–feldspar intergrowths; z, zircon.
|
ACKNOWLEDGEMENTS |
|---|
A.S.M. acknowledges receipt of an NERC research studentship and R.M. tenure of a Visiting Research Professorship at the Open University. We thank Dr Nic Odling (Edinburgh) for XRF analytical assistance, and Kay Green and Michelle Higgins (Open University) for thin sections. Eric Christiansen, Gail Mahood, John White and Marjorie Wilson provided insightful and constructive reviews, for which we are deeply grateful. Fieldwork in Kenya was greatly assisted by Geoffrey Muchemi and Johnson Mungania at the Olkaria Geothermal Project. The Warden of Hell's Gate National Park, and John Mackay and Tim Trent of Oserian Development Co. Ltd made access to parts of the field area possible.
*Corresponding author. Present address: Environment Centre, Lancaster University, Lancaster LA1 4YQ, UK. E-mail: r.macdonald{at}lancaster.ac.uk
|
REFERENCES |
|---|
Andersen DJ, Lindsley DH, Davidson PM. QUILF: a Pascal program to assess equilibria among Fe–Mg–Ti oxides, pyroxenes, olivine and quartz. Computers and Geosciences (1993) 19:1333–1350.[CrossRef]
Bachmann O, Bergantz GW. On the origin of crystal-poor rhyolites: extracted from batholithic crystal mushes. Journal of Petrology (2004) 45:1565–1582.
Bachmann O, Dungan MA, Lipman PW. The Fish Canyon magma body, San Juan volcanic field, Colorado: rejuvenation and eruption of an upper-crustal batholith. Journal of Petrology (2002) 43:1469–1503.
Bacon CR, Macdonald R, Smith RL, Baedecker PA. Pleistocene high-silica rhyolites of the Coso volcanic field, Inyo County, California. Journal of Geophysical Research (1981) 86:10223–10241.
Bailey DK, Macdonald R. Petrochemical variations among mildly peralkaline (comendite) obsidians from the oceans and continents. Contributions to Mineralogy and Petrology (1970) 28:340–351.[CrossRef][Web of Science]
Bailey DK, Schairer JF. Feldspar–liquid equilibria in peralkaline liquids—the orthoclase effect. American Journal of Science (1964) 262:1198–1206.[Abstract]
Baker BH, Henage LF. Compositional changes during crystallization of some peralkaline silicic lavas of the Kenya Rift Valley. Journal of Volcanology and Geothermal Research (1977) 2:17–28.[CrossRef][Web of Science]
Black S, Macdonald R, Kelly MR. Crustal origin for peralkaline rhyolites from Kenya: evidence from U-series disequilibria and Th-isotopes. Journal of Petrology (1997) 38:277–297.
Blundy J, Cashman K. Ascent-driven crystallisation of dacite magmas at Mount St Helens, 1980–1986. Contributions to Mineralogy and Petrology (2001) 140:631–650.[Web of Science]
Bone BD. The geological evolution of the S.W. Naivasha volcanic complex, Kenya. In: Ph.D. thesis (1987) University of Lancaster.
Clarke MCG, Woodhall DG, Allen D, Darling G. Geological, volcanological, and hydrogeological controls on the occurrence of geothermal activity in the area surrounding Lake Naivasha, Kenya. Report (1990) Nairobi: Ministry of Energy. 138.
Davies GR, Macdonald R. Crustal influences in the petrogenesis of the Naivasha basalt–comendite complex: combined trace element and Sr–Nd–Pb isotope constraints. Journal of Petrology (1987) 28:1009–1031.
Duffield WA, Bacon CR, Dalrymple GB. Late Cenozoic volcanism, geochronology, and structure of the Coso Range, Inyo County, California. Journal of Geophysical Research (1980) 85:2381–2404.
Fitton JG, Saunders AD, Larsen LM, Hardarson BS, Norry MJ. Volcanic rocks from the southeast Greenland margin at 63°N: Composition, petrogenesis and mantle sources. In:. Saunders AD, Larsen HC, Wise SW Jr, eds. (1998) College Station, TX: Ocean Drilling Program. 331–350. Proceedings of the Ocean Drilling Program, Scientific Results, 152.
Hanchar JM, van Westrenen W. Rare earth element behaviour in zircon–melt systems. Elements (2007) 3:37–42.
Hanchar JM, Watson EB. Zircon saturation thermometry. In:. Zircon. Mineralogical Society of America, Reviews in Mineralogy—Hanchar JM, Hoskin PWO, eds. (2003) 53:89–112.
Heumann A, Davies GR. U–Th disequilibrium and Rb–Sr age constraints on the magmatic evolution of peralkaline rhyolites from Kenya. Journal of Petrology (2002) 43:557–577.
Hildreth W. The Bishop Tuff: evidence for the origin of compositional zonation in silicic magma chambers. In:. Ash-flow Tuffs. Geological Society of America, Special Papers—Chapin CE, Elston WE, eds. (1979) 180:43–75.
Hildreth W. Volcanological perspectives on Long Valley, Mammoth Mountain, and Mono Craters: several contiguous but discrete systems. Journal of Volcanonology and Geothermal Research (2004) 136:169–198.[CrossRef]
Hildreth W, Wilson CJN. Compositional zoning of the Bishop Tuff. Journal of Petrology (2007) 48:951–999.
Jochum KP, Willbold M, Raczek I, Stoll B, Herwig K. Chemical characterisation of the USGS reference glasses, GSA-1G, GSC-1G, GSD-1G, GSE- 1G, BCR-2g, BHVO-2G and BIR-1G using EPMA, ID-TIMS, ID-ICPMS and LA-ICP-MS. Geostandards and Geoanalytical Research (2005) 29:285–302.[CrossRef][Web of Science]
Jochum KP, Stoll B, Herwig K, et al. MPI-DING reference glasses for in situ microanalysis: New reference values for element concentrations and isotope ratios; technical brief. Geochemistry, Geophysics, Geosystems (2006) 7. doi:10.1029/2005GC001060.
Kunzmann T. The aenigmatite–rhönite mineral group. European Journal of Mineralogy (1999) 11:743–756.
Lowenstern JB, Clynne MA, Bullen TD. Comagmatic A-type granophyre and rhyolite from the Alid volcanic centre, Eritrea, Northeast Africa. Journal of Petrology (1997) 38:1707–1721.
Macdonald R. Nomenclature and petrochemistry of the peralkaline oversaturated extrusive rocks. Bulletin Volcanologique (1974) 38:498–516.[CrossRef]
Macdonald R, Belkin HE. Compositional variation in minerals of the chevkinite group. Mineralogical Magazine (2002) 66:1075–1098.
Macdonald R, Davies GR, Bliss CM, Leat PT, Bailey DK, Smith RL. Geochemistry of high-silica peralkaline rhyolites, Naivasha, Kenya rift valley. Journal of Petrology (1987) 28:979–1008.
Macdonald R, Marshall AS, Dawson JB, Hinton RW, Hill PG. Chevkinite-group minerals from salic volcanic rocks of the East African Rift. Mineralogical Magazine (2002) 66:287–299.
Macdonald R, Belkin HE, Fitton JG, Rogers NW, Nejbert K, Tindle AG, Marshall AS. The roles of fractional crystallization, magma mixing, crystal mush remobilization and volatile–melt interactions in the genesis of a young basalt–peralkaline rhyolite suite, the Greater Olkaria Volcanic Complex, Kenya rift valley. Journal of Petrology (2008) 49:1515–1547.
Mahood GA, Stimac JA. Trace-element partitioning in pantellerites and trachytes. Geochimica et Cosmochimica Acta (1990) 54:2257–2276.[CrossRef][Web of Science]
Manley CR, Bacon CR. Rhyolite thermobarometry and the shallowing of the magma reservoir, Coso volcanic field, California. Journal of Petrology (2000) 41:149–174.
Marshall AS. High-silica peralkaline magmatism of the Greater Olkaria Volcanic Complex, Kenya Rift Valley. In: PhD thesis (1999) Lancaster University.
Marshall AS, Hinton RW, Macdonald R. Phenocrystic fluorite in peralkaline rhyolites, Olkaria, Kenya rift valley. Mineralogical Magazine (1998) 62:477–486.[Abstract]
Mechie J, Keller GR, Prodehl C, Khan MA, Gaciri SJ. A model for the structure, composition and evolution of the Kenya rift. Tectonophysics (1997) 278:95–119.[CrossRef][Web of Science]
Metz JM, Mahood GA. Development of the Long Valley, California, magma chamber recorded in precaldera rhyolite lavas of Glass Mountain. Contributions to Mineralogy and Petrology (1991) 106:379–397.[CrossRef][Web of Science]
Mooney WD, Christensen NI. Composition of the crust beneath the Kenya Rift. Tectonophysics (1994) 236:391–408.[CrossRef][Web of Science]
Nicholls I, Carmichael ISE. Peralkaline acid liquids: a petrological study. Contributions to Mineralogy and Petrology (1969) 20:268–294.[CrossRef]
Noble DC. Sodium, potassium, and ferrous iron contents of some secondarily hydrated natural silicic glasses. American Mineralogist (1967) 52:280–286.[Web of Science]
Noble DC. Loss of sodium from crystallized comendite welded tuff of the Miocene Grouse Canyon Member of the Belted Range Tuff, Nevada. Geological Society of America Bulletin (1970) 31:2677–2687.
Omenda PA. The geology and structural controls of the Olkaria geothermal system, Kenya. Geothermics (1998) 27:55–74.[CrossRef][Web of Science]
Pearce NJG, Perkins WT, Westgate JA, Gorton MP, Jackson SE, Neal CR, Chenery SP. A compilation of new and published major and trace element data for NIST SRM 610 and NIST SRM 612 glass reference materials. Geostandards Newsletter (1997) 21:115–144.[CrossRef][Web of Science]
Pearce NJG, Westgate JA, Perkins WT, Preece SJ. The application of ICP-MS methods to tephrochronological problems. Applied Geochemistry (2004) 19:289–322.[CrossRef][Web of Science]
Peccerillo A, Barberio MR, Yirgu G, Ayalew D, Barbieri M, Wu TW. Relationships between mafic and peralkaline silicic magmatism in continental rift settings: a petrological, geochemical and isotopic study of the Gedemsa volcano, Central Ethiopian Rift. Journal of Petrology (2003) 44:2003–2032.
Pichavant M, Costa F, Burgisser A, Scaillet B, Martel C, Poussineau S. Equilibration scales in silicic to intermediate magmas—implications for experimental studies. Journal of Petrology (2007) 48:1955–1972.
Pouchou JL, Pichoir F. ‘PAP’ procedure for improved quantitative analysis. Microbeam Analysis (1985) 20:104–105.
Scaillet B, Macdonald R. Phase relations of peralkaline silicic magmas and petrogenetic implications. Journal of Petrology (2001) 42:825–845.
Scaillet B, Macdonald R. Experimental constraints on the relationships between peralkaline rhyolites of the Kenya Rift Valley. Journal of Petrology (2003) 44:1867–1894.
Scaillet B, Macdonald R. Fluorite stability in peralkaline magmas. Contributions to Mineralogy and Petrology (2004) 147:319–329.[CrossRef][Web of Science]
Scott SC, Skilling IP. The role of tephrachronology in recognising synchronous caldera-forming events at the Quaternary volcanoes Longonot and Suswa, south Kenya Rift. In:. Volcanoes in the Quaternary. Geological Society, London, Special Publications—Firth CR, McGuire WJ, eds. (1999) 161:47–67.[CrossRef]
Simon JI, Reid MR. The pace of rhyolite differentiation and storage in an ‘archetypical’ silicic magma system, Long Valley, California. Earth and Planetary Science Letters (2005) 235:123–140.[CrossRef][Web of Science]
Sisson TW, Bacon CR. Gas-driven filter pressing in magmas. Geology (1999) 27:613–616.
Sun S.-S, McDonough WF. Chemical and isotopic systematics of oceanic basalts: implications for mantle composition and processes. In:. Magmatism in the Ocean Basins. Geological Society, London, Special Publications—Saunders AD, Norry MJ, eds. (1989) 42:313–345.
Sutton AN, Blake S, Wilson CJN. An outline geochemistry of rhyolite eruptives from Taupo volcanic centre, New Zealand. Journal of Volcanology and Geothermal Research (1995) 68:153–175.[CrossRef][Web of Science]
White JC, Holt GS, Parker DF, Ren M. Trace-element partitioning between alkali feldspar and peralkaline quartz trachyte to rhyolite magma. Part I: Systematics of trace-element partitioning. American Mineralogist (2003) 88:316–329.
Wilding MC, Macdonald R, Davies JR, Fallick AE. Volatile characteristics of peralkaline rhyolites from Kenya: an ion microprobe, infrared spectroscopic and hydrogen isotope study. Contributions to Mineralogy and Petrology (1993) 144:264–275.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  R. Macdonald and B. Baginski The central Kenya peralkaline province: a unique assemblage of magmatic systems Mineralogical Magazine, June 15, 2009; 73(1): 1 - 16. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Macdonald, B. Baginski, H.E. Belkin, P. Dzierzanowski, and L. Jezak REE partitioning between apatite and melt in a peralkaline volcanic suite, Kenya Rift Valley Mineralogical Magazine, April 17, 2009; 72(6): 1147 - 1161. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||