Journal of Petrology

Journal of Petrology Advance Access originally published online on September 25, 2006
Journal of Petrology 2007 48(1):3-42; doi:10.1093/petrology/egl052

This Article Abstract FREE Full Text (PDF) Supplementary data All Versions of this Article: 48/1/3    most recent egl052v1 Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Add to My Personal Archive Download to citation manager Request Permissions Google Scholar Articles by Toothill, J. Articles by Tindle, A. G. Search for Related Content GeoRef GeoRef Citation Social Bookmarking          What's this?

© The Author 2006. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org

A Complex Petrogenesis for an Arc Magmatic Suite, St Kitts, Lesser Antilles

J. Toothill¹, C. A. Williams², R. MacDonald^1,*, S. P. Turner³, N. W. Rogers², C. J. Hawkesworth⁴, D. A. Jerram⁵, C. J. Ottley⁵ and A. G. Tindle²

¹Environment Centre, Lancaster University, Lancaster LA1 4YQ, UK
²Department of Earth Sciences, The Open University, Milton Keynes MK7 6AA, UK
³Department of Earth and Planetary Sciences, Macquarie University, Sydney, NSW 2109, Australia
⁴Department of Earth Sciences, University of Bristol, Bristol BS8 1RJ, UK
⁵Department Of Earth Sciences, University of Durham, Durham DH1 3LE, UK

RECEIVED FEBRUARY 3, 2005; ACCEPTED AUGUST 24, 2006

	ABSTRACT

TOP ABSTRACT INTRODUCTION GENERAL GEOLOGY ANALYTICAL METHODS PETROLOGY AND MINERAL CHEMISTRY PETROGENESIS CONTRIBUTIONS TO THE MANTLE... CONCLUSIONS SUPPLEMENTARY DATA REFERENCES

 
St Kitts forms one of the northern group of volcanic islands in the Lesser Antilles arc. Eruptive products from the Mt Liamuiga centre are predominantly olivine + hypersthene-normative, low-K basalts through basaltic andesites to quartz-normative, low-K andesites. Higher-Al and lower-Al groups can be distinguished in the suite. Mineral assemblages include olivine, clinopyroxene, orthopyroxene, plagioclase and titanomagnetite with rarer amphibole, ilmenite and apatite. Eruptive temperatures of the andesites are estimated as 963–950°C at fO₂ NNO + 1 (where NNO is the nickel–nickel oxide buffer). Field and mineral chemical data provide evidence for magma mixing. Glass (melt) inclusions in the phenocrysts range in composition from andesite to high-silica rhyolite. Compositional variations are broadly consistent with the evolution of more evolved magmas by crystal fractionation of basaltic parental magmas. The absence of any covariation between ⁸⁷Sr/⁸⁶Sr or ¹⁴³Nd/¹⁴⁴Nd and SiO₂ rules out assimilation of older silicic crust. However, positive correlations between Ba/La, La/Sm and ²⁰⁸Pb/²⁰⁴Pb and between ²⁰⁸Pb/²⁰⁴Pb and SiO₂ are consistent with assimilation of small amounts (<10%) of biogenic sediments. Trace element and Sr–Nd–Pb isotope data suggest derivation from a normal mid-ocean ridge basalt (N-MORB)-type mantle source metasomatized by subducted sediment or sediment melt and fluid. The eruptive rocks are characterized by ²³⁸U excesses that indicate that fluid addition of U occurred <350 kyr ago; U–Th isotope data for mineral separates are dominated by melt inclusions but would allow crystallization ages of 13–68 ka. However, plagioclase is consistently displaced above these ‘isochrons’, with apparent ages of 39–236 ka, and plagioclase crystal size distributions are concave-upwards. These observations suggest that mixing processes are important. The presence of ²²⁶Ra excesses in two samples indicates some fluid addition <8 kyr ago and that the magma residence times must also have been less than 8 kyr.

KEY WORDS: Sr–Nd–Pb isotopes; U-series isotopes; crystal size distribution; petrogenesis

	INTRODUCTION

TOP ABSTRACT INTRODUCTION GENERAL GEOLOGY ANALYTICAL METHODS PETROLOGY AND MINERAL CHEMISTRY PETROGENESIS CONTRIBUTIONS TO THE MANTLE... CONCLUSIONS SUPPLEMENTARY DATA REFERENCES

 
The Lesser Antilles volcanic arc is the result of the westward subduction of the Atlantic plate beneath the Caribbean plate (Fig. 1). The rate of subduction has been estimated at between 2 and 4 cm/a (e.g. Jarrard, 1986; Rosencrantz & Sclater, 1986), which places it at the low end of the arc spectrum. The slow subduction rates may be responsible for relatively low magma production rates. Over the three time scales 300, 10 000 and 100 000 years, the estimated magma production rates have been 5, 3 and 4 km³/Ma per km (Wadge, 1984). These rates are an order of magnitude lower than those over the same periods in Central America, where convergence rates have been around four times higher (Wadge, 1984).

View larger version (19K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 1. Map of the Lesser Antilles volcanic arc (after Heath et al., 1998a).

 
The Lesser Antilles arc is segmented in terms of the configuration of the Benioff zone (Wadge & Shepherd, 1984). To the north of Martinique, the arc segment trends about 330° and the Benioff zone dips at 50–60°. The southern segment, trending 20°, has a dip of 50–45° in the north but is nearly vertical in the south. The depth to the Benioff zone beneath St Kitts, the focus of this study, is 160 km.

Brown et al. (1977) recognized that there is an unusually wide range of magma types formed contemporaneously along strike in the arc, from, in their terminology, tholeiitic suites in the northern islands (Saba to Montserrat), through calc-alkaline in the central islands (Guadeloupe to St Vincent) to silica-undersaturated, alkaline varieties in the southern islands (Grenada and the Grenadines). In the classification scheme recommended by Tatsumi & Eggins (1995), the rocks of the northern islands are low-K suites, those of the central islands plot in the low-K and lower part of the medium-K fields, whereas the rocks of Grenada and the Grenadines are of medium-K type.

Issues of current interest in island arc magmatism include: interpretation of information available from the compositions of melt inclusions; reconciliation of the time-scale information for magmatic processes from crystal size distributions (CSD) and U-series isotopes, respectively; understanding the controls on oxygen fugacity; and the investigation of the role of different sediment suites in shallow-level contamination processes. Given the role of the Lesser Antilles in the development of ideas about subduction-related magmatism (Baker, 1968; Sigurdsson et al., 1973; Arculus, 1976; Brown et al., 1977; Arculus & Wills, 1980; Hawkesworth & Powell, 1980; Rea & Baker, 1980; Devine & Sigurdsson, 1983; Thirlwall & Graham, 1984; White & Dupré, 1986; Davidson, 1987), it is surprising how few detailed petrographic and geochemical studies have been published. In this paper, we describe aspects of the petrology and geochemistry of the eruptive rocks of the Mt Liamuiga (Mt Misery) centre on St Kitts, one of the northern group of islands. Specific aims are to:

1. assess how phenocryst assemblages and compositions relate to whole-rock compositions in a low-K suite;
   
2. model the differentiation history within the crust of the Mt Liamuiga magmas;
   
3. provide the first compositional data for melt inclusions trapped in phenocrysts of the Mt Liamuiga suite;
   
4. use U–Th isotopes from both whole-rocks and mineral separates to place constraints on the timing of fluid addition and magma crystallization beneath St Kitts and to compare these data with previous results for the Soufrière volcano of St Vincent;
   
5. use crystal size distribution (CSD) data to assess the role of plagioclase entrainment.
   

	GENERAL GEOLOGY

TOP ABSTRACT INTRODUCTION GENERAL GEOLOGY ANALYTICAL METHODS PETROLOGY AND MINERAL CHEMISTRY PETROGENESIS CONTRIBUTIONS TO THE MANTLE... CONCLUSIONS SUPPLEMENTARY DATA REFERENCES

 
St Kitts measures 39 km x 9 km and rises to a height of 1157 m above sea level (a.s.l.) at Mt Liamuiga in the north of the island (Fig. 2). Four volcanic centres are aligned along the length of the island on a NW–SE trend. These rest on a basement of older volcanic deposits exposed in the centre of the island. The relative degrees of dissection of the centres indicate that activity has migrated NW with time (Martin-Kaye, 1959).

View larger version (37K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 2. Geological sketch map of St Kitts (after Baker, 1984).

 
The oldest post-basement volcanic rocks on St Kitts are found on the Salt Pond Peninsula. They consist largely of pyroxene andesite lavas, domes and agglomerates, with rarer basaltic andesites and rhyolites (Martin-Kaye, 1959). The South East Range and Middle Range in the centre of the island are poorly exposed and inadequately studied. Baker (1984) recorded a K–Ar age of 1 Ma for an andesite of the South East Range. The Middle Range has been dated at 1–2 Ma (Maury & Westercamp, 1990), although this centre is generally believed to be somewhat younger than the South East Range (Baker, 1984). Pyroxene andesite pyroclastic deposits dominate the Middle Range volcanic rocks, but there are small deposits of dacite (Martin-Kaye, 1959).

The most recent activity on St Kitts has been at the Mt Liamuiga stratovolcano, with a maximum age, based on K–Ar dating, of 1 Ma (Baker, 1985). There have been no eruptions since the island was first settled by Europeans in 1624. The central cone consists of basalt and andesite lavas, whereas the lower ground around the cone is made up largely of pyroclastic deposits, termed by Baker (1969) the Mansion Series, with a few andesite domes and basalt lavas. Stratigraphic correlations (Baker, 1968, 1985; Roobol et al., 1981) are complicated by poor exposure and lack of continuity of deposits. The earliest dated deposits from Mt Liamuiga, 0·42 Ma [Units A and B of Roobol et al. (1981)], are basaltic cinders and pumices erupted at the base of the Mansion Series. They include the Lower Green Lapilli zone and the Cinder Zone. Unit C, the Upper Green Lapilli deposits, is of uncertain age but the similarity to Units A and B suggests that they are part of a related eruptive sequence (Harkness et al., 1994). Unit D, comprising pumice falls and flows, has been carbon dated at 4500–3500 years BP (Baker, 1985).

The pyroclastic flows of Unit E, found on the northern and western flanks of Mt Liamuiga, formed 1800 years BP (Harkness et al., 1994). Near their top, these deposits are interbedded with the ashes and surge deposits of Unit F, which include the lithic-rich ashes of the Steel Dust Series, the Upper Lamberts pyroclastic flow, and the Upper Pumice Unit, the youngest eruptive rocks on St Kitts.

A series of parasitic domes have been built on the flanks of Mt Liamuiga and the South East Range (Martin-Kaye, 1959). The most accessible, the andesitic Brimstone Hill, has been dated at <44 ± 1·2 ka by radiocarbon dating of a carbonate fossil (Westermann & Kiel, 1961) but may actually have been extruded during the time of the pre-Columbian Indians, 3000 BP (Baker, 1985). The Sandy Point dome has not been dated but the degree of weathering suggests that it is of similar age to Brimstone Hill.

The eruptive rocks of St Kitts range from basaltic to rhyolitic. Andesites and dacites predominate; basaltic andesites and basalts are relatively rare (7% by volume; Baker, 1984). In that we have particular interest in the parental magmas of the St Kitts suite and in the nature of the mantle sources, this study concentrates on rocks from Mt Liamuiga, which yields the only basalts on the island. Their young eruption ages have also allowed us to study U-series disequilibrium in the rocks. We selected for chemical analysis 24 rocks from the pyroclastic deposits of the Mansion Series, mainly blocks in pyroclastic flows and falls but also some bulk pumices, 14 rocks from lavas associated with the Mansion Series, and one from each of the Brimstone Hill and Sandy Point domes. Two samples from the Salt Pond Peninsula were also analysed, to test the contention (Baker, 1984) that these rocks are more potassic than the Mt Liamuiga rocks. Latitudes and longitudes, stratigraphic affinity, form and lithology of the analysed samples are given in the Appendix.

	ANALYTICAL METHODS

TOP ABSTRACT INTRODUCTION GENERAL GEOLOGY ANALYTICAL METHODS PETROLOGY AND MINERAL CHEMISTRY PETROGENESIS CONTRIBUTIONS TO THE MANTLE... CONCLUSIONS SUPPLEMENTARY DATA REFERENCES

 
Samples cleaned of weathered edges were broken into 5 cm diameter pieces using a hydraulic rock splitter. These were passed through a jaw crusher to obtain fragments <5 mm in size. Between 50 and 100 g was then ground to a fine powder (<60 µm) in an agate ball mill.

Bulk mineral separates for U–Th isotopic work were prepared by splitting, disc-milling and sieving followed by removal of magnetite using a hand magnet. Sodium polytungstate ( = 2·8 kg/m³) and Frantz magnetic separation were combined to obtain a clean plagioclase separate and a heavy magnetic fraction, from which olivine and pyroxene were purified. All fractions were visually inspected and hand picked to remove impurities. The final separates, estimated to be 95% pure, were placed in RO H₂O in an ultrasonic bath for 20 min prior to powdering in a dedicated agate mill that had been cleaned with 6M HCl and RO H₂O.

Electron microprobe analyses of phenocrysts, melt inclusions and matrix glasses were made at the Open and Edinburgh Universities, using Cameca SX100 and Camebax instruments, respectively. A total of 787 spot analyses were made of phenocrysts and matrix glasses in 19 rocks (seven basalts, two basaltic andesites and 10 andesites). A total of 45 melt inclusions, hosted by clinopyroxene, orthopyroxene or plagioclase, were analysed, of which 12 were also analysed for trace elements by ion probe (below). Mineral analyses were performed with an accelerating potential of 20 kV and a beam current of 25 or 20 nA in spot mode, whereas glass analyses were obtained using a 15 kV accelerating voltage and beam current of 10 nA. Where possible, for the glass analyses the beam was rastered over 10 µm, to minimize alkali migration. The concentration of each element was determined over a 30 s counting interval to overcome the scatter in individual measurements that is an inherent part of the technique.

The decay curve method of Nielsen & Sigurdsson (1981) is commonly used to reduce errors in Na measurement in glasses. In this study, SiO₂ and Na₂O were analysed first during each run, i.e. in the first 30 s of exposure of the sample to the electron beam. The decay curves obtained suggested that the error in the measurement of both oxides within that time interval was less than the error inherent in the technique and consequently the decay curve method was not applied to the analyses.

Ion probe analyses of melt inclusions were made on the Cameca IMS-4F facility at Edinburgh University. A –10·7 keV ¹⁶O primary beam was used and the sample was held at +4·5 keV, giving a net impact energy of the ions of 15·2 keV. The primary beam current was 5 nA and the beam width 25 µm. The number of elements analysed (35) made it necessary to run two routines on each point analysed, one for ‘light’ elements (atomic masses 1–85) and one for ‘heavy’ elements (88–238 plus ³⁰Si). The light elements were always analysed after the heavy elements, to minimize the contribution from adsorbed hydrogen on the sample surface. Element abundances were normalized to the ³⁰Si peak. Corrections were made manually, assuming a constant ion yield of each element relative to Si, for Na on V, Mg + Si on Cr, Cr and Al on Fe, FeSi on Rb, and BaO on Eu.

Whole-rock major element compositions were determined by X-ray fluorescence spectroscopy at the Open University (OU). Relative uncertainties (2) for those elements (as oxide wt %) are: SiO₂ 0·96, TiO₂ 1·50, Al₂O₃ 0·70, Fe₂O₃* 0·90, MnO 2·67, MgO 1·26, CaO 1·56, Na₂O 5·60, K₂O 1·35, P₂O₅ 7·82. Trace elements were determined by inductively coupled plasma–mass spectrometry (ICP–MS), using a Perkin–Sciex ELAN 6000 system at the University of Durham, following a standard nitric and hydrofluoric acid digestion (Ottley et al., 2003). Detection limits (in ppb) are: Co 19, Cr 532, Ni 897, Sc 212, V, 273, Cu 128, Rb 13.4, Cs 3.79, Ba 157, Sr 3.22, La 3.73, Ce 12.2, Pr 2.57, Nd 13.2, Sm 2.84, Eu 1.78, Gd 3.01, Tb 0.30, Dy 2.12, Ho 0.4, Er 0.23, Tm 0.40, Yb 0.44, Lu 0.12, Y 20.9, Zr 25.2, Hf 4.46, Nb 9.72, Ta 0.68, Pb 25.4, Th 2.73 and U 0.78.

Sr, Nd and Pb were separated using cation and anion exchange separation techniques following standard HF–HNO₃ dissolutions at the OU. Sr and Nd isotopes were analysed by thermal ionization mass spectrometry (TIMS) and corrected for within-run mass bias to ⁸⁶Sr/⁸⁸Sr = 0·1194 and ¹⁴⁴Nd/¹⁴⁶Nd = 0·7219. All Sr and Nd data are reported relative to the values 0·71025 for NBS 987 and 0·51184 for La Jolla. Uncertainties, as determined from the 2 reproducibility of the NBS 987 and J&M standards during the course of analysis, were 28 ppm for Sr and 21–25 ppm for Nd. Pb isotopes were analysed on a Nu-Instruments multi-collector ICP-MS system using Tl for internal mass bias correction (Belshaw et al., 1998). Reproducibility is estimated at 0·2% (2).Total procedural blanks were less than 1 ng, 500 pg and 500 pg for Sr, Nd and Pb, respectively.

U–Th separation was carried out at the OU on standard HF–HCl–HNO₃ dissolutions to which a mixed ²²⁹Th–²³⁶U tracer had been added. Samples were treated with HCl and H₃BO₄ to ensure sample-spike equilibration and to eliminate fluorides. U and Th were isolated using anionic exchange resin, with HNO₃, HCl and HBr as elutants, and then loaded onto degassed Re filaments along with colloidal graphite and a HNO₃–H₃PO₄ solution, respectively. Th and U concentrations and (²³⁴U/²³⁸U) ratios were determined to ±0·5% at the OU by TIMS fitted with an RPQ II energy filter for high abundance sensitivity (van Calsteren & Schwieters, 1995). The (²³⁰Th/²³²Th) measurements were made on the Nu-Instruments multi-collector ICP-MS system at the OU, using techniques reported by Turner et al. (2001a). Reproducibility was better than 1% 2. All samples had (²³⁴U/²³⁸U) ratios within error of unity, indicating that they had not suffered from seawater contamination. Decay constants used in the calculation of activity ratios (denoted by parentheses) were ²³⁰Th = 9·1952 x 10^–6, ²³²Th = 4·948 x 10^–11; ²³⁸U = 1·551 x 10^–10.

The results used in all subsequent figures here have been age-corrected to 2·3 ka on the basis of the ages given by Baker (1984). For most rocks, the initial ratios are very similar to the present-day values, as the rocks are young (ages <3 ka). The andesite from the Brimstone Hill dome (Kit7) and the basalt from the Sandy Point dome (Kit40) have been corrected to 44 ka, on the basis of a ¹⁴C date, as noted above, but both domes may be much younger.

Crystal size distribution (CSD) analysis was performed on thin sections of samples Kit42, Kit47 and Kit56. Textures were digitized and outlines of crystals identified for image analysis (e.g. Jerram et al., 2003). Plagioclase crystals were chosen for CSD analysis, with population numbers of 1830, 1385 and 1071, respectively, giving very high precision. Crystal shapes were determined using CSDslice (Morgan & Jerram, 2006), and CSD analysis was performed using CSDcorrections software v.1.35 (after Higgins, 2000). Data are presented as population density vs crystal size.

	PETROLOGY AND MINERAL CHEMISTRY

TOP ABSTRACT INTRODUCTION GENERAL GEOLOGY ANALYTICAL METHODS PETROLOGY AND MINERAL CHEMISTRY PETROGENESIS CONTRIBUTIONS TO THE MANTLE... CONCLUSIONS SUPPLEMENTARY DATA REFERENCES

 
Eruptive rocks from Mt Liamuiga range from basalt to andesite, with basalts and basaltic andesites making up 30% of the total volume (Baker, 1969). The dominant rock type is a two-pyroxene andesite; dacites have not been observed, although dacite and rhyolitic compositions occur as matrix glass and as glass (melt) inclusions in phenocrysts.

Rocks with clear evidence for magma mingling, such as mixed pumices, are found at Mansion on the eastern side of Mt Liamuiga and occasionally in the pyroclastic rocks of the NE of the volcano. Two mixed rocks have been specifically used in this study; Kit35, which has basalt (Kit35d) and hornblende andesite (Kit35l) as end-members, and Kit56, which is the mafic component of a basalt–andesite mixed rock.

Modal abundances of 17 representative samples are given in Table 1 and phenocryst compositions in Tables 2–5. Typical phenocryst assemblages are: basalts: olivine (ol)–clinopyroxene (cpx)–plagioclase (plag), with minor titanomagnetite (mt); basaltic andesites: plag–cpx–ol ± othopyroxene (opx) ± mt; andesites: plag–cpx–opx–mt–amphibole. Thus, clinopyroxene, plagioclase and an Fe–Ti oxide phase are present throughout the compositional spectrum, whereas olivine is absent in rocks with less than 3·5 wt % MgO. Olivine shows some overlap with orthopyroxene, but not amphibole, crystallization.

View this table: [in this window] [in a new window]   Table 1: Modal proportions of phenocrysts in St Kitts eruptive rocks

 

View this table: [in this window] [in a new window]   Table 2: Analyses of olivine phenocrysts in St Kitts rocks

 
Orthopyroxene is first found in rocks with MgO contents between 5 and 6 wt %, slightly more magnesian than the value (4 wt %) suggested for St Kitts by Macdonald et al. (2000). Amphibole is found only in rocks with MgO <3 wt %. The exception is the pargasitic amphibole in the basaltic component (Kit35d) of the mixed magma rock Kit35. Baker et al. (1980) and Defant et al. (2001) have recorded similar amphiboles forming phenocrysts in basalts on Saba, and Kit35d may represent an amphibole-phyric basalt magma type rare on St Kitts. Apatite forms microphenocrysts in Kit35d and is enclosed in larger plagioclases in the andesites. Baker (1968) recorded rare (2 vol.%) quartz phenocrysts in St Kitts andesites. These may be xenocrystic or may have crystallized from the host rock when the residual melt had reached rhyolitic composition.

Olivine
Olivine is a common phenocryst phase in the basalts and less common in the basaltic andesites. Crystals range up to 0·5 mm in diameter and are generally subhedral, although euhedral and rounded crystals also occur. Olivine occasionally includes plagioclase, clinopyroxene and titanomagnetite, and is itself often included in plagioclase and clinopyroxene. Melt (glass) inclusions in the olivines are partially to totally devitrified. Representative analytical data for olivines are presented in Table 2. The complete dataset is available at http://www.petrology.oupjournals.org. In the basalts and basaltic andesites, the compositional range in crystal cores is Fo_81–62. The more magnesian values in each rock are generally compatible with equilibrium crystallization from magmas equivalent in composition to the host rocks, with a K_D 0·3, assuming that olivine crystallized early in the sequence and that the host rocks approximate melt compositions (Fig. 3). The more Fe-rich values in the cores in the basalts (Fo₆₃; Table 2) are typical of more evolved magmas and may result either from crystallization from the host when the residual liquid was andesitic, or from magma mixing. The latter process is compatible with the presence in some rocks (e.g. Kit56) of two olivine populations (Fo₇₀ and Fo₆₃; Table 2) and strong reverse zoning (e.g. Kit11: Fo_72–82; Table 2). The compositional range of olivines in the andesites is Fo_77–72; they are almost certainly relicts of a higher temperature phase of crystallization and/or a result of magma mixing of andesite with basalt.

View larger version (17K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 3. Fe2+/Mg ratios of cores of olivine phenocrysts plotted against Fe2+/Mg ratios in whole-rocks. Vertical lines show range of compositions in individual samples. Dashed line is for an equilibrium KD value of 0·3 (Roeder & Emslie, 1970).

 
Clinopyroxene
Clinopyroxene occurs as large (<2 mm), euhedral crystals, pale brown to green in colour, as larger, strained crystals and as small rounded grains. It is also a common phase in mineral clusters comprising various combinations of plagioclase, clinopyroxene, orthopyroxene, Fe–Ti oxide and, rarely, amphibole, where it forms euhedra up to 4 mm across. Clinopyroxene includes olivine, plagioclase and titanomagnetite. Representative analyses are presented in Table 3. The full dataset is available at http://www.petrology.oupjournals.org. With the exception of sub-calcic augite in Kit45a, the clinopyroxenes are augitic. The compositional ranges (cores and rims) are: basalts Ca_35–48Mg_38–52Fe_10–20; basaltic andesites Ca_39–45Mg_41–46Fe_14–16; and andesites Ca_41–43Mg_38–43Fe_16–20; there is a tendency for more evolved rocks to carry slightly more Fe-rich pyroxenes. Zoning is negligible in most crystals but both normal and reverse examples are seen (Table 3). Normal zoning predominates in the more mafic rocks, with rim enrichments up to Fs_2·4, e.g. in Kit35d. Reverse zoning is less common but rimwards decreases in Fs of up to 2% have been recorded in Kit35d and in the andesite Kit28. Cr₂O₃ and TiO₂ contents are low, generally <0·04 and <1·0 wt %, respectively.

View this table: [in this window] [in a new window]   Table 3: Analyses of clinopyroxene phenocrysts in St Kitts rocks

 
Clinopyroxenes in the Mt Liamuiga basalts are rather Al-rich, with maximum atoms per formula unit (a.p.f.u.) values 0·30 (Al₂O₃ 7 wt %; Table 3). This is a feature that they share with the low-K suite of Soufrière, St Vincent (Heath et al., 1998a). Fe³/Fe² ratios have been calculated assuming stoichiometry (Droop, 1987). The Mt Liamuiga clinopyroxenes, especially those with higher mg-numbers, are distinctly oxidized, with Fe³⁺/(Fe³⁺ + Fe²⁺) values up to 0·56 (Fig. 4). This is in line with the high-pressure (7·5–20 kbar) experimental results of Pichavant et al. (2002a) on a high-MgO basalt from St Vincent, where the clinopyroxenes had Fe³⁺/(Fe³⁺ + Fe²⁺) values up to 0·33, the value increasing with melt water content. Predictably, there is at St Kitts a positive correlation between Fe³⁺/(Fe³⁺ + Fe²⁺) and Al p.f.u. (Fig. 4).

View larger version (12K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 4. (a) Fe3+/Fe ratios plotted against (a) mg-number [Mg/(Mg + Fe*), where Fe* is total Fe as Fe2+] and (b) Al p.f.u. for clinopyroxene phenocrysts in basalts, basaltic andesites and andesites from St Kitts.

 
Orthopyroxene
Orthopyroxene is found in rocks with MgO <5·5 wt % but is ubiquitous only in rocks with <3·5 wt % MgO. It forms euhedral, green–brown prisms up to 2 mm in diameter, occurring as discrete phenocrysts and in phenocryst clusters. Representative analyses are presented in Table 4. The full dataset is available at http://www.petrology.oupjournals.org. Compositions fall in the range Ca_1·5–4·2Mg_50–71Fe_26–48, and there is no systematic variation with whole-rock composition (Table 4). Both normal and reverse zoning are seen but the core–rim range never exceeds En₃. Al₂O₃ and TiO₂ concentrations are low, <2 and <0·5 wt %, respectively. Fe³⁺/(Fe³⁺ + Fe²⁺) ratios are up to 0·14 and are lower than those in clinopyroxenes in the same rock.

View this table: [in this window] [in a new window]   Table 4: Analyses of orthopyroxene and amphibole phenocrysts in St Kitts rocks

 
Plagioclase
Plagioclase is the dominant phenocryst phase in all rocks, up to 48% modally (Table 1), and ranges in size from large (up to 1 cm) crystals, which exhibit complex zoning, to smaller laths transitional in size to groundmass crystals. Resorption textures, such as embayed margins and zones of inclusions, are ubiquitous in the larger phenocrysts of some samples (e.g. Kit19) and essentially absent in others (Kit15). There is no correlation between phenocryst texture and bulk-rock composition.

The plagioclases show very complex core–rim zonation patterns, as exemplified by two phenocrysts from basalt Kit40 and andesite Kit66 (Fig. 5). The variations are summarized in Fig. 5, which presents core to rim data for the two phenocrysts. The full dataset, including those for a further 10 traverses in crystals from six rocks, is available at http://www.petrology.oupjournals.org. All the plagioclases are marked by very low K₂O (<0·1 wt %) and BaO (0·03 wt %) and modest Fe₂O₃* (1 wt %) abundances. The main features are as follows.

1. Strongly calcic cores (An_95–88) occur in all rock types (e.g. Kit40; Fig. 5). More rarely, relatively sodic cores are found in basalts (e.g. An₅₂ in Kit35d).
   
2. There are several different styles of zonation, even in phenocrysts from the same rock. Thus, some crystals show normal zoning (Kit40; Fig. 5; also Kit45a, Kit47, Kit56), some show reverse zoning (Kit66; Fig. 5; also Kit28), and yet others show (variably frequent) changes from more calcic to more sodic compositions (Kit28, Kit45a). Reverse zoning is most strongly developed in the andesites.
   
3. Overall, the range of phenocryst core compositions (An_95–58) is larger than that of rim compositions (An_80–60).
   

View larger version (76K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 5. Back-scattered electron images of, and microprobe compositional data for traverses across, complexly zoned plagioclase phenocrysts in basalt Kit40 and andesite Kit66. The position of the traverses is noted on the BSE images.

 
These features are consistent with a combination of crystal fractionation and magma mixing, where crystal growth in a relatively shallow magma chamber has periodically been interrupted by the injection of mafic magma.

Fe–Ti oxides
Fe–Ti oxides (Table 5) are present as a phenocryst phase in the majority of rocks. They form small, rounded, discrete crystals (<0·5 mm) but occur more frequently as inclusions within, or associated with, other phenocryst phases. Ilmenite occurs in only two samples; as a single subhedral phenocryst in the mafic portion (Kit35d) of mixed magma rock Kit35 and as exsolution lamellae in titanomagnetite in andesite Kit28. Representative analyses are presented in Table 5. The full dataset is available at http://www.petrology.oupjournals.org. Spinel phase compositions (cores and rims) range between Mt₄₉Usp₅₁ and Mt₈₆Usp₁₄. The usp/mt ratio slightly decreases from basalts to andesites, the average X_usp being 37 and 32, respectively. There is significant overlap between rock types and some rocks contain two populations, e.g. basalt Kit45a contains spinels clustering around X_usp 40 and 30. Some samples show weak normal zoning, with rimward decrease in Ti (<3 a.p.f.u.). Reverse zoning (X_usp 25–34) was found in phenocrysts in andesite Kit28 (Table 5).

View this table: [in this window] [in a new window]   Table 5: Analyses of Fe–Ti oxide phenocrysts in St Kitts rocks

 
With one exception (basalt Kit48, 1·08 wt %), Cr₂O₃ concentrations are low (<0·5 wt %). Al₂O₃ and MgO abundances are more variable, up to 9 wt % (0·38 a.p.f.u.) and 5·5 wt % (0·014 a.p.f.u.), respectively. Al and Mg are positively correlated and both show weak negative correlations with X_usp. Cr/(Cr + Al) ratios are in the range 0–0·09, Mg/(Mg + Fe²⁺) 0·03–0·25, and Fe³⁺/(Fe³⁺ + Al + Cr) 0·81–0·94. These fall within the ranges for titanomagnetites from the basalt–andesite suite of Soufrière, St Vincent (Heath et al., 1998a) and are similar to those from Saba (Defant et al., 2001). The only specimen to contain coexisting oxides is the mafic portion (Kit35d) of the mixed rock Kit35.

Amphibole and apatite
Brown–green, prismatic, occasionally resorbed, amphibole phenocrysts occur mainly in andesites with MgO < 2·5 wt % and more rarely in basalts. Crystals are commonly rimmed by oxides. The amphibole in the mafic portion (Kit35d) of mixed magma Kit35 is pargasitic (mg-number 60), whereas that in the andesitic fraction (Kit35l) is edenitic (mg-number 71; Table 4).

Apatite occurs mainly as inclusions in plagioclase phenocrysts. Partial analysis of a euhedral phenocryst in Kit35d shows moderate F and Cl abundances (2·44 and 1·39 wt %, respectively).

Melt inclusions and matrix glasses
Glass (melt) inclusions are fairly common in plagioclase, olivine, clinopyroxene, orthopyroxene and amphibole phenocrysts in lavas, pumices and lava blocks from block and ash flow deposits. Inclusions range in size from <10 to 60 µm and vary in colour from clear to pale brown. All inclusions in olivine were partially devitrified and we have not analysed any from that phase. Inclusions in clinopyroxene are circular to oval in shape, whereas in plagioclase many are square or rectangular and aligned parallel to the compositional zoning. Orthopyroxene crystals tend to host smaller, rarer inclusions. Many inclusions contain gas bubbles with volumes ranging from 20–40% relative to inclusion volume in some larger inclusions to 0–5% (more commonly).

Brown, isotropic glass is found as small patches in the matrix of a few basaltic to andesitic samples.

Geothermometry
The two-pyroxene geothermometer (5 kbar projection) of Lindsley (1983) was used to estimate the temperatures of magma crystallization of two andesites whose pyroxenes are thought to be close to equilibrium compositions and for which the estimated temperatures are within 50°C of each other. Kit66 and Kit64 gave 963 and 950°C, respectively. These values are in line with those found in andesites and dacites elsewhere in the Lesser Antilles [960–740°C; compilation by Macdonald et al. (2000)]. However, they are higher than those found for water-rich andesites of the Soufrière Hills, Montserrat (820–840°C; melt water 4·27 ± 0·54 wt %; Barclay et al., 1998) and 20th-century eruptions of Mt Pelée, Martinique (875–900°C; melt water 5·3–6·3 wt %; Pichavant et al., 2002b). This is consistent with the St Kitts andesites being relatively water-poor or with heating by mafic magma.

The coexisting oxide pair in the mafic portion (Kit35d) of the mixed rock Kit35 meets the Mn–Mg partitioning test of Bacon & Hirschmann (1988). Temperature estimates were made using the ILMAT program of LePage (2003).The solution model of Spencer & Lindsley (1981) yields temperatures of 865 and 849°C and fO₂ of –11·22 and –11·67 for cores and rims, respectively. These temperatures are low for a basalt, which we ascribe to cooling during mingling with the andesitic component of the rock prior to eruption. Given that the original crystallization temperatures were probably higher, the oxygen fugacity ( NNO + 1, where NNO is the nickel–nickel oxide buffer) is in line with previous estimates from the Lesser Antilles arc [compiled by Macdonald et al. (2000)] and indicates that the magmas were relatively oxidized, which is consistent with the high Fe³⁺ contents of the clinopyroxenes noted above (Fig. 4).

Summary
In summary, therefore, Mt Liamuiga rocks show phenocryst distributions and compositional ranges that are comparable with those found in other Lesser Antilles suites. This suggests that the evolution of the magmas has largely been controlled by crystal–liquid equilibria (Macdonald et al., 2000). However, the large range of phenocryst core compositions in individual rocks, the common lack of correlation between mineral and host-rock compositions and the common occurrence of reverse zoning in phenocrysts are consistent with the idea that magma mixing and entrainment of xenocrysts have played roles in magma differentiation.

GEOCHEMISTRY
Major and trace element and Sr–Nd–Pb isotopic data for St Kitts rocks have been published by Baker (1968, 1980, 1984, 1985), Baker & Holland (1973), Brown et al. (1977), Davidson (1985, 1987), White & Dupré (1986), Thirlwall et al. (1994), Turner et al. (1996) and Chabaux et al. (1999). We complement this dataset with new major element analyses of a further 40 samples, 30 of which have also been analysed for trace elements and Sr–Nd–Pb isotopes (Tables 6 and 7). In this section, we assess the geochemical affinity of, and compositional variation within, the Mt Liamuiga suite.

View this table: [in this window] [in a new window]   Table 6: Major element analyses of St Kitts rocks

 

View this table: [in this window] [in a new window]   Table 7: Trace element and Sr–Nd–Pb isotopic compositions of St Kitts rocks

 
With the exception of rocks from Salt Pond, the St Kitts rocks define, on the K₂O–SiO₂ diagram of Tatsumi & Eggins (1995), a fairly continuous series from ol + hy-normative, low-K basalt to q-normative, low-K andesite (Fig. 6). The Salt Pond rocks, which are 2·3 Myr older than those of Mt Liamuiga and geographically separated from them (Fig. 2), form a medium-K trend, extending to rhyolitic compositions (see Baker, 1984). On a plot of SiO₂ against FeO*/MgO ratio (Fig. 7), the Mt Liamuiga rocks show neither a clear tholeiitic or calc-alkaline trend in Miyashiro's (1974) terms, which is typical of the low-K suites of the Lesser Antilles (Macdonald et al., 2000).

View larger version (26K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 6. Variation of wt % K2O vs wt % SiO2 (Tatsumi & Eggins, 1995) for St Kitts rocks. Data sources: this study (Table 6), Baker (1980, 1984), Davidson (1987) and Turner et al. (1996).

 

View larger version (14K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 7. Variation of wt % SiO2 vs FeO*/MgO, showing Miyashiro's (1974) calc-alkaline and tholeiitic fields and rocks from the higher-Al and lower-Al groups of Mt Liamuiga (see text for details).

 
The range of major and trace element compositional variation in the eruptive rocks as a function of wt % MgO is illustrated in Figs 8 and 9. There is a sufficient range at a given MgO content to suggest that the suite comprises several magmatic lineages. However, there is a primary distinction, made on the basis of Al₂O₃, between two groups of samples, which we term subsequently the higher-Al and lower-Al groups. At a given MgO, the higher-Al group has lower SiO₂ and higher CaO and Al₂O₃, abundances (Fig. 8). It also has, on average, lower Ba, Cs, Rb and Zr, and higher Sc, Sr and V, concentrations than the lower-Al group (Fig. 9). The higher-Al group has a more restricted compositional range (MgO 4·6–2·2 wt %) than the lower-Al group (7·0–1·5 wt %). There appear to be no stratigraphic or geographical differences between the groups, suggesting that both magma types were available for eruption throughout much of the time period represented in our collection.

View larger version (15K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 8. Whole-rock major element abundances in Mt Liamuiga eruptive rocks, matrix glasses and melt inclusions as a function of MgO content (wt %).

 

View larger version (12K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 9. Whole-rock trace element abundances (ppm) in Mt Liamuiga eruptive rocks as a function of MgO content (wt %). One melt inclusion with Zr 198 ppm and MgO 0·28 wt % has been omitted to preserve the scale.

 
It has been shown experimentally (Sisson & Grove, 1993; Pichavant & Macdonald, 2003) that ol–pl–cpx–water-saturated basaltic melts at 2–4 kbar can contain up to 22 wt % Al₂O₃. The Mt Liamuiga eruptive rocks contain up to 21·5 wt % Al₂O₃ (Fig. 8). The higher-Al rocks are displaced, relative to the lower-Al rocks, towards the range of plagioclase phenocryst compositions (Al₂O₃ 28–36 wt % at 0 wt % MgO), which would be consistent either with plagioclase accumulation or with these rocks representing a genuinely more aluminous magma type.

Petrological and geochemical evidence does not support a model of plagioclase accumulation. (1) Modal data (Table 1) indicate that the higher-Al group contains 25–48% plagioclase phenocrysts, and the lower-Al group 25–39%, excluding the exceptional Kit7 and Kit35d (1% and 5%, respectively). (2) On a SiO₂–Al₂O₃ plot (Fig. 10), both groups form trends parallel to the compositional ranges of plagioclase (and clinopyroxene) phenocrysts in the eruptive rocks; there is no indication that the higher-Al rocks represent lower-Al magmas enriched in plagioclase phenocrysts. (3) On a SiO₂–FeO*/MgO plot (Fig. 7), the groups occupy separate, but roughly parallel and slightly overlapping, fields, which would not be expected if the relationship was one of simple plagioclase accumulation. (4) Sr/Nd ratios, which would be expected to be higher in plagioclase-accumulitic rocks are similar in both groups (higher-Al group, n = 11, range 20–38, mean 28·3 ± 4; lower-Al group, n = 19, range 27–39, mean 31·5 ± 4). (5) Plagioclase-accumulitic rocks might be expected to show higher Eu/Eu* ratios. However, the Eu anomalies in both groups are very similar: higher-Al group, n = 11, range 0·96–1·04, mean 0·98 ± 0·04; lower-Al group, n = 19, range 0·89–1·08, mean 0·99 ± 0·05. We conclude, therefore, that the higher-Al group generally represents a slightly more aluminous magma type than the lower-Al group.

View larger version (13K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 10. Variation of wt % SiO2 vs wt % Al2O3 to show that the relationship between the higher-Al and lower-Al groups is not one of simple plagioclase accumulation.

 
Melt inclusions
We selected for analysis melt inclusions in five andesites (Kit15, Kit16 from blocks in pyroclastic flows; Kit14, Kit29 from pumice falls; and Kit4 from a lava) and the basaltic portion (Kit35d) of a mixed magma rock (Kit35) from blocks reworked from a pyroclastic flow. None of the inclusions showed signs of post-entrapment leakage, such as the presence of bisecting cracks, and there was no indication of post-entrapment devitrification, e.g. the presence of microlites within the inclusion. Ideally, there was no sign of post-entrapment crystallization of the host; where this was detected in some plagioclase-hosted inclusions, we have not included the data here. Representative analyses are given in Table 8. The range is from silicic andesite to rhyolite, the majority being dacitic or rhyolitic. There are no systematic differences between the compositions of glasses hosted by the different mineral phases. Basalt Kit35d contains rhyolitic melt inclusions within plagioclase phenocrysts; andesite Kit15 contains the widest range of melt inclusion compositions, from 61 to 73 wt % SiO₂ (100% anhydrous). Overall, the melt inclusions show compositions commensurate with their being fractionation products of andesitic magmas, e.g. higher SiO₂ and K₂O abundances and lower Al₂O₃, TiO₂, FeO* and CaO abundances, and P₂O₅ peaks at 1 wt % MgO. There is, however, considerable scatter in the MgO–element plots (Fig. 8), which may relate to combinations of analytical imprecision at low concentration levels (Ti, P), Na loss during microprobe analysis, contamination by the host mineral during analysis, and undetected post-entrapment crystallization. The important point is that the inclusions preserve melt compositions not represented among the whole-rock compositions of the eruptive products of Mt Liamuiga and they provide information on the high-silica end of the St Kitts liquid line of descent.

View this table: [in this window] [in a new window]   Table 8: Analyses of representative glass (melt) inclusions and matrix glasses

 
Brown, isotropic glass is found as small patches in the matrix of a few samples. Large variations in Na₂O/K₂O ratio, 0·9–5·3 (Table 8), suggest that the electron microprobe beam has sometimes impinged on microlites embedded in the glass and/or the glasses have been secondarily hydrated; however, the compositional range is clearly from andesitic to rhyolitic. The matrix glass compositions (Table 8) fall within the major element fields defined by the melt inclusion data in Fig. 8.

Minor and trace elements
Figure 11 presents primitive mantle-normalized minor and trace element abundance patterns for three relatively magnesian basalts [mg-number >55; Kit59 and Kit35 from this study and KB64 from Turner et al. (1996)]. The patterns are very similar and characteristic of subduction-related magmas; specifically they have high large ion lithophile element/high field strength element (LILE/HFSE) and light rare earth element (LREE)/HFSE ratios, Nb and Ta depletion relative to La, and relatively high Ba/La and Sr/Nd ratios. Chondrite-normalized REE patterns for the range of compositions (Fig. 12) show enrichments relative to chondrite of between 11 and 21. Absolute abundances of the REE increase from basalts to andesites. The patterns are relatively flat; [La/Yb]_CN ranges only from 1·06 to 1·65 and there are no significant differences between the rock types (mean values in the basalts, basaltic andesites and andesites are 1·22 ± 0·11, 1·21 ± 0·10 and 1·23 ± 0·18, respectively). Europium anomalies are minor; Eu/Eu* ratios are in the range 0·89–1·09 and, as for [La/Yb]_CN, do not differ between rock types (mean values in basalts, basaltic andesites and andesites are 0·99 ± 0·04, 0·97 ± 0·04 and 1·00 ± 0·05, respectively). The melt inclusions show increasing LREE/heavy REE (HREE) ratios with increasing silica content; [La/Yb]_CN reaches 2·7 in the high-silica rhyolite inclusions in Kit35d. LREE enrichment is accompanied overall by larger, negative Eu anomalies, with Eu/Eu* as low as 0·40. The drop might be ascribed to an increasing tendency for Eu to enter plagioclase during melt differentiation.

View larger version (25K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 11. Primitive mantle-normalized trace element patterns for three relatively magnesian basalts (mg-number >55) from Mt Liamuiga. Samples used: Kit35d and Kit59, this study (Table 7; low-Al group), and KB64 (Turner et al., 1996; high-Al group). Normalizing factors from Sun & McDonough (1989).

 

View larger version (27K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 12. Chondrite-normalized REE patterns for a basalt (Kit59), two basaltic andesites (Kit33 and Kit45b), an andesite (Kit7) (Table 5) and three melt inclusion (Table 6). Normalizing factors from Sun & McDonough (1989).

 
Radiogenic isotopes
Sr and Nd isotope data for whole-rocks are given in Table 7 and shown in comparison to the range of data from other Lesser Antilles islands in Fig. 13. The ranges are small (⁸⁷Sr/⁸⁶Sr 0·70354–0·70369; ¹⁴³Nd/¹⁴⁴Nd 0·51295–0·51307) and are within the ranges previously reported for St Kitts rocks (Davidson, 1985, 1987; White & Dupré, 1986). The St Kitts samples have Sr isotopic ratios slightly higher ratios than, and Nd isotopic ratios similar to, Atlantic normal mid-ocean ridge basalt (N-MORB) samples (Saunders et al., 1988). There is no systematic variation of Sr–Nd isotopic composition with whole-rock composition and no difference between rocks of the higher-Al and lower-Al groups. The range of Pb isotope compositions (Table 7) is also small (²⁰⁶Pb/²⁰⁴Pb 18·961–19·046; ²⁰⁷Pb/²⁰⁴Pb 15·630–15·679; ²⁰⁸Pb/²⁰⁴Pb 38·649–38·837) but there are positive correlations with SiO₂ content (e.g. Fig. 14).

View larger version (22K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 13. Variation of 143Nd/144Nd vs 87Sr/86Sr for the St Kitts eruptive rocks. (a) Data from this study. (b) Comparison of the St Kitts data with data from other Lesser Antilles islands (compilation by Williams, 2000) and Atlantic MORB (Saunders et al., 1988). The arrow points to the average Northern Antilles sediment, from Plank & Langmuir (1998).

 

View larger version (8K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 14. 208Pb/204Pb vs wt % SiO2 for the Mt Liamuiga extrusive rocks.

 
U–Th isotope data are listed in Table 9 and plotted on the U–Th equiline diagram in Fig. 15. Th concentration data can be difficult to reproduce at low levels and there are some differences between the ICP-MS (Table 7) and TIMS (Table 9) whole-rock data. The TIMS data should be the more accurate and the use of a mixed U–Th spike provides additional robustness for the U/Th ratios used in the discussion below.

View larger version (25K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 15. U–Th equiline diagram for the data from St. Kitts (this study) compared with literature data for other Lesser Antilles islands and the composition of the average sediment being subducted beneath the southern Lesser Antilles (from Heath et al., 1998b).

 

View this table: [in this window] [in a new window]   Table 9: U–Th analyses of St Kitts rocks and separated minerals

 
The whole-rock (²³⁰Th/²³²Th) ratios range from 1·16 to 1·52 and (²³⁸U/²³²Th) ratios range from 1·19 to 1·89; the variation in (²³⁸U/²³⁰Th) disequilibria is from 0·98 to 1·50. These new data are similar to, but expand the range of, published data for St. Kitts rocks (Gill & Williams, 1990; Turner et al., 1996; Chabaux et al., 1999) and for rocks from other islands in the Northern Antilles. Figure 15 shows that many of the St Kitts samples lie close to the equiline and that when the full dataset for the Lesser Antilles is considered, there is a clear separation between lavas from the northern islands (Saba, St. Kitts, Statia and Redonda) and those further south (Montserrat, Guadeloupe, Martinique, St. Lucia, Grenada and the Grenadines). The outstanding exception is Soufrière, St. Vincent, which plots with the northern group (Turner et al., 1996). However, the compositions of the low-K basalts and andesites from St. Vincent are also more akin to those in the northern Antilles (Heath et al., 1998b) and St. Vincent is an exception to the general north to south compositional progression along the arc.

	PETROGENESIS

TOP ABSTRACT INTRODUCTION GENERAL GEOLOGY ANALYTICAL METHODS PETROLOGY AND MINERAL CHEMISTRY PETROGENESIS CONTRIBUTIONS TO THE MANTLE... CONCLUSIONS SUPPLEMENTARY DATA REFERENCES

 
Primary and parental magmas
The most magnesian basalt from St Kitts so far analysed has a mg-number of 0·64 and a Ni content of 36 ppm (Turner et al., 1996), which is unlikely to represent a primary mantle-derived magma. Rocks with near-primary geochemical characteristics (mg-number >0·65 and Ni >235 ppm) have been found on Martinique, St Vincent, and Grenada and the Grenadines, and Macdonald et al. (2000) have argued that all the Lesser Antilles magma series have been derived from high-Mg (MgO 12 wt %) basalts.

The rocks of the Soufrière volcano on St Vincent are transitional between tholeiitic and calc-alkaline compositions (Thirlwall et al., 1994; Heath et al., 1998a) and the Soufrière suite is compositionally similar to the Mt Liamuiga suite on St Kitts (Macdonald et al., 2000). Pichavant et al. (2002a) conducted experiments on a high-MgO basalt from St Vincent, which may also be, therefore, a close analogue for the primary magmas of the Mt Liamuiga suite. They found that the composition studied was multiply saturated on its liquidus with a spinel lherzolite phase assemblage at 1235°C, 11·5 kbar for 1·5 wt % H₂O in the melt, and at 1185°C, 16 kbar for 4·5 wt % H₂O in the melt. Various phase constraints suggested to Pichavant et al. (2002a) that the basalt magma at St. Vincent was extracted relatively dry (2 wt % H₂O) from its mantle source. However, it is plausible that the basalt last equilibrated with mantle lherzolite at the relatively shallow depths indicated by the experiments (<50 km). We believe that, overall, relatively dry melting of mantle lherzolite is a credible model for the primary magmas of Mt Liamuiga.

Macdonald et al. (2000) divided the eruptive suites of the Lesser Antilles into two groups—high-Ca and low-Ca. The high-Ca group comprises rocks from St Eustatius, St Kitts, Redonda, Guadeloupe, Dominica, Bequia and the C-series of Grenada. The low-Ca group comprises rocks from Saba, Montserrat, Martinique, St Vincent and the M-series of Grenada. Distinctions between the groups are most easily made at MgO contents between 6 and 8 wt %. Generally, at a given MgO, high-Ca suites tend to have lower SiO₂ and higher Al₂O₃, TiO₂ and FeO* than rocks of the low-Ca group. Macdonald et al. (2000) suggested that the difference between the two groups was derived from the early stages of polybaric fractionation and, in particular, the temperature difference between the appearance of olivine and clinopyroxene crystallization. Delayed clinopyroxene crystallization, perhaps as a result of magma ascent history and/or melt water content, resulted in relative Ca enrichment of residual liquids. Because the high-pressure pyroxenes were aluminous, the residual liquids were also relatively Al-rich.

We can adopt this model for the Mt Liamuiga rocks. Although both the higher-Al and lower-Al groups are high-Ca in the Macdonald et al. (2000) sense, the compositional differences between them are similar to those between the high-Ca and low-Ca groups. We suggest that the higher-Al magmas of St Kitts were derived from parental magmas in which clinopyroxene crystallization was slightly delayed compared with its appearance in the lower-Al group. Possible evidence for this idea comes from Sc and V data, as both elements are partitioned into clinopyroxene (Rollinson, 1993). At a given MgO, the higher-Al rocks generally have higher Sc contents than those of the lower-Al group (Fig. 9), consistent with derivation from parental magmas that had experienced less clinopyroxene fractionation. The higher-pressure clinopyroxenes were almost certainly Al-rich; the most aluminous clinopyroxene reported in this study has 7·05 wt % Al₂O₃ (Table 3), whereas Pichavant et al. (2002a) found Al₂O₃ values up to 11 wt % in clinopyroxenes crystallized at 17·5 kbar and 1240°C from a high-magnesia basalt from Soufrière, St Vincent. Earlier clinopyroxene crystallization from the parental magmas of the lower-Al group would have tended to lower Al abundances in residual liquids relative to the higher-Al group. The lower-Al character would have been exacerbated if plagioclase had also separated earlier, perhaps as a result of somewhat higher melt water contents.

Modelling of major element variations
Major element data for St Kitts rocks are plotted against MgO in Fig. 8. With decreasing MgO, there are overall increases in SiO₂, Na₂O, K₂O and P₂O₅ concentrations and decreases in Al₂O₃ and CaO. FeO* and TiO₂ show slightly peaked trends. More evolved compositions are generally more silica oversaturated, as is normal for Lesser Antilles suites (Macdonald et al., 2000). The overall coherence of the trends is qualitatively compatible with a fractional crystallization model. We have accordingly modelled the compositional variations using the Ghiorso & Sack (1995) MELTS computer code. Basalt Kit59 (MgO 6·89 wt %) from the lower-Al group was used as the starting composition, as this rock shows no or minimal evidence of mixing. For simplicity, pressure was fixed, at 1 kbar, and temperatures were modelled in decreasing increments of 5°C. Where fO₂ was specified as unconstrained throughout the run, FeO/Fe₂O₃ ratios of the starting composition were recalculated at QFM + 1 (where QFM is the quartz–fayalite–magnetite buffer). Where fO₂ was specified, FeO/Fe₂O₃ was recalculated at that value. The initial, ‘starting’ temperature was set at 1000°C and the liquidus temperature was then calculated. The models were allowed to function either until the model finished or the compositions became too unrealistic (SiO₂ <30 wt %). Details of the 14 runs made have been given by Williams (2000). The results of the run (2 wt % H₂O, no fO₂ restrictions) that most closely reproduced the natural compositional variations are given in Table 10. The crystallizing assemblage is comparable with the natural phenocryst assemblage in the St Kitts suite (namely, olivine–clinopyroxene–plagioclase–oxides) likely to have been crystallizing over the temperature range 1139–939°C, with orthopyroxene appearing at the lower end of that range.

View this table: [in this window] [in a new window]   Table 10: Results of MELTS modelling

 
The modelling reproduces the whole-rock compositional trends reasonably well, especially as the rocks almost certainly represent a number of different liquid lines of descent. The least satisfactory matches are for FeO* and TiO₂. However, MELTS is very sensitive to the fO₂ imposed, with consequent effects on when the Fe–Ti oxides begin to crystallize. The effect of Fe–Ti oxide fractionation in the calculated proportions is to lower the Fe³⁺/Fe²⁺ ratios in the residual magmas, as is reflected in the compositions of the clinopyroxene phenocrysts. We conclude, therefore, that the major element variations in the lower-Al Mt Liamuiga eruptive rocks are consistent with fractional crystallization of the observed phenocryst phases from a basaltic parental magma with MgO 7 wt %. Given the overlap or parallelism of the compositional trends of the higher-Al and lower-Al groups (Fig. 8), we apply the same conclusion to the former group.

However, there is mineral chemical evidence, complementing field evidence, that magma mixing has also played a role in magmatic evolution, such as the presence of highly calcic cores to some plagioclase phenocrysts (Fig. 5), the common occurrence of reverse zoning in olivine, plagioclase and pyroxene phenocrysts, and the presence of two olivine phenocryst populations in some rocks. Many Mt Liamuiga rocks are hybrids from mixing of basalt, basaltic andesite and andesite magmas, and it is highly likely that open-system processes have overprinted magmatic processes dominated by fractional crystallization.

We now check whether trace element variations are consistent with closed-system fractionation ± magma mixing. In Fig. 16, Ba, La and Th are plotted against Zr; all these elements would be expected to display near-incompatible behaviour during fractional crystallization. Maximum enrichment factors shown in the whole-rocks are Ba and Th 3, La and Zr 2. The value for Ba and Th is close to that predicted from the major element modelling (Table 10), i.e. the proportion of melt remaining at the andesitic stage is 0·24. The lower value for Zr suggests that it has partitioned into a crystallizing phase or phases, such as clinopyroxene. However, La has a similar D value to Th and the lower enrichment factor remains unexplained. We conclude, however, that the trace element variations are broadly consistent with closed-system crystal fractionation.

View larger version (14K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 16. Zr vs Ba, La and Th (ppm) to show that incompatible trace element variations in the Mt Liamuiga rocks are broadly explicable by Rayleigh fractionation, assuming KD 0 for all elements (continuous lines). Numbered ticks refer to the amount of melt remaining at each stage.

 
A role for crustal contamination?
Several studies of Lesser Antilles eruptive rocks have concluded that fractional crystallization within the crust was accompanied by assimilation of crustal rocks, the so-called assimilation–fractional-crystallization (AFC) processes (DePaolo, 1981; Thirlwall & Graham, 1984; Davidson, 1985, 1986, 1987; Davidson & Harmon, 1989; Thirlwall et al., 1994; Smith et al., 1996). According to Macdonald et al. (2000), AFC has been most pronounced in the central islands from Dominica to Bequia (Fig. 1). Recently, however, Defant et al. (2001) have suggested that certain trace element characteristics of the eruptive rocks of Saba, a neighbouring island of St Kitts (Fig. 1), are best explained by intra-crustal fractional crystallization combined with assimilation of biogenic sediments.

On St Kitts, Sr and Nd isotopic compositions are essentially independent of SiO₂ content, indicating either that no significant crustal contamination has occurred or that, if it has, the contaminant was isotopically similar to the magmas. However, Pb isotopic ratios covary positively with SiO₂ (Fig. 15), perhaps indicative of AFC. We test this, with specific comparison with the Saba rocks (Defant et al., 2001), in Fig. 17, a plot of [Ba/La]_CN against [La/Sm]_CN. Whereas the Mt Liamuiga basalts form a flat trend, consistent with fractional crystallization of the mineral assemblage determined by modelling (Table 10), the andesites and the majority of basaltic andesites are displaced to higher Ba/La ratios. They form a weak trend, within the field of Saba data, towards the field of biogenic sediments and away from that of terrigenous sediments from the Antilles. This suggestion that the St Kitts rocks have been contaminated by sediments within the crust is supported by the positive correlations between [Ba/La]_CN and ²⁰⁸Pb/²⁰⁴Pb ratios (Fig. 18) and Pb isotopes and SiO₂ contents (Fig. 15). Mass-balance calculations indicate that the amount of sediment required is low, <10% for even the most contaminated rock (Kit62a).

View larger version (23K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 17. [Ba/La]CN vs [La/Sm]CN for Mt Liamuiga rocks. Also shown are the fields of 90% island arc basalts (IAB), biogenic and terrigenous sediments from Deep See Drilling Project hole 543 to the east of Saba, eruptive rocks from Saba, literature data for Lesser Antilles eruptive rocks, and sediment analyses from the eastern side of the northern (N LAT) and southern (S ANT) Lesser Antilles, all taken from Defant et al. (2001, fig. 15). Normalizing factors from Kay (1980). CN, chondrite-normalized. Two fractional crystallization trends from basaltic parental magmas are shown, one involving orthopyroxene (opx), the other amphibole (am). Fractionation of amphibole results in decrease in Ba/LaCN with increasing La/SmCN. Fractionation of orthopyroxene gives a much flatter trend; the St Kitts basalts follow a similar, flat trend, consistent with the presence of orthopyroxene in the MELTS major element modelling. The third trend shows the trend of melt compositions formed by increasing degrees of MORB-type mantle; Ba/LaCN is little affected by degree of partial melting. All three vectors from Defant et al. (2001).

 

View larger version (13K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 18. [Ba/La]CN vs 208Pb/204Pb. The slope of the regression line is given by y = 6·98x – 267·7 and the correlation is significant at the 95% level. The size of the symbols is about equal to the typical error bar on the 208Pb/204Pb values. Normalizing factors from Kay (1980).

 
Mineral ages and the entrainment of old plagioclase crystals?
Mineral separate data for samples Kit42, Kit47 and Kit56 (Table 9) are plotted on U–Th equiline diagrams and are used along with plagioclase crystal size distribution (CSD) plots in Fig. 19 to evaluate the ages of different crystal suites and textural evidence for the accumulation or entrainment of plagioclase.

View larger version (31K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 19. U–Th equiline diagrams for mineral separates and CSD plots for plagioclase from Kit42, Kit47 and Kit56. In each case the CSD plots indicate separate slopes for the small and large size fractions, and growth times were calculated using an assumed growth rate of 1 x 10–11 cm/s (Jerram et al., 2003; Turner et al., 2003a). Only the minerals from Kit47 define a reasonable U–Th isochron (the plagioclase point was excluded from the age calculation). For Kit42 and Kit56 isochrons were tied to the initial ratio defined by the array from Kit47. It should be noted that in all three rocks the plagioclase separate lies above the isochrons, implying a greater age as indicated. The concave-upward CSD profiles and older plagioclase ages are interpreted in terms of entrainment of cumulate plagioclase crystals (see text for discussion).

 
The data for the mineral separates span a range of U–Th isotope compositions similar to the whole-rocks and it is notable that the plagioclase separate has the highest (²³⁰Th/²³²Th) ratio in each of the three rocks (Table 9). However, it should be noted that some of the mineral separates have the lowest Th and U concentrations measured and this demands some caution in interpreting the results. Furthermore, calculated average partition coefficients for both U and Th into the minerals are 0·02, 0·09 and 0·7 for plagioclase, pyroxene and magnetite, respectively. These values far exceed those predicted theoretically (Blundy & Wood, 2003) and, combined with the general lack of U/Th fractionation, suggest that melt inclusions dominate the U and Th budgets of the mineral separates. Additionally, the data for these three samples do not form true isochrons and, in all cases, plagioclase lies above the other data, which cluster for both Kit42 and Kit56. Only the data (excepting plagioclase) from Kit47 form a linear array. For this sample, the whole-rock lies at the high (²³⁸U/²³²Th) end of the array. A further complexity for Kit47 is that the minerals do not mass balance to produce the whole-rock value. Heath et al. (1998b) obtained a similar result from mineral separates on St. Vincent and observed that there must be a high-U/Th phase that had not been analysed. Because the rocks analysed in both studies are basalts or basaltic andesites, this is unlikely to be zircon and the most likely candidate is early formed spinel, which is predicted to have a high U/Th ratio (Blundy & Wood, 2003).

With these caveats, the data from basaltic andesite Kit47 form a plausible isochron (Fig. 19), yielding an age of 30 ka if the plagioclase separate is omitted, as discussed below. One approach for Kit42 (basalt) and Kit56 (basalt fraction of mixed magma rock) is to adopt the same initial (²³⁰Th/²³²Th) ratio as that implied by the Kit47 isochron (1·25). This assumption yields maximum age estimates, because the majority of samples from St. Kitts have (²³⁰Th/²³²Th) ratios 1·25, and the ages obtained from Kit42 and Kit56 by tying their whole-rock, groundmass, pyroxene and magnetite separates to this initial ratio are 13 and 68 ka, respectively. If the assumption is correct, the implication is that the pyroxenes and titanomagnetites in these lavas are 13–68 kyr old (Fig. 19a–c). Similarly old mineral ages were obtained by Heath et al. (1998b) for lavas from Soufrière, St. Vincent, and because the groundmass from each lava also lies close to these isochrons, it might be inferred that these ages represent magma residence times. However, it is important to remember that the U–Th data from the bulk mineral separates are dominated by melt inclusions and the age constraints apply only if these included melts were trapped with an initial (²³⁰Th/²³²Th) ratio the same as that inferred for the ‘isochrons’ (see below).

A striking observation from Fig. 19d–f is that the plagioclase CSDs are kinked or curved concave-upwards. Such variation on a CSD plot is indicative of mixed populations and/or changes in the crystallization history of the population (Marsh, 1998; Jerram et al., 2003). As noted above, all the St. Kitts plagioclase separates lie above the U–Th isochrons and one interpretation is that the plagioclases are 26, 140 and 170 kyr older than the other minerals in the three lavas analysed (but see below). Plagioclase has the lowest density of all the minerals crystallized in these lavas and thus the greatest propensity for older crystals to be entrained. Similar inferences have been made for data from Mount St Helens (Cooper & Reid, 2003) and Tonga (Turner et al., 2003a), and Turner et al. (2003a) have argued that the old mineral U–Th isotope ages from St Vincent may be similarly explained.

If one considers the slopes of the kinked CSD plots using realistic growth rates for plagioclase (1 x 10^–11 cm/s; e.g. Jerram et al., 2003; Turner et al., 2003a), the growth time for the small crystal populations is several tens of years. This value is likely to represent the time spent in a high-level magma chamber directly prior to eruption. In each CSD it is possible to measure a growth time for the larger crystal population (lower slopes on CSD plots; Fig. 19). However, there may be a significant time gap between the growth time for the larger ‘older’ population and the smaller ‘younger’ population, which cannot be resolved. Even if we use a slower growth rate of 1 x 10^–12 cm/s, suggested for some systems (Marsh, 1998), the maximum growth time is only a few thousand years. These combined observations strongly suggest that the older ages reflect the entrainment of older crystals or crystal cores, which are mixed with crystals associated with the final stages of magma ascent and eruption, giving an averaged older age.

The above age interpretations rely heavily on the initial (²³⁰Th/²³²Th) ratio assumed for the different samples and that the crystals grew from, and trapped melt inclusions from, the same melts as their enclosing groundmass. Notwithstanding these substantial assumptions, one problem with taking the ages at face value is that it is also known that most young lavas from the Lesser Antilles (and other arcs), including Kit45b and Kit50, preserve whole-rock ²²⁶Ra–²³⁰Th disequilibria (Gill & Williams, 1990; Chabaux et al., 1999; Turner et al., 2001b). Assuming that the ²²⁶Ra excesses derive from fluid addition to the mantle source region (Turner et al., 2001b), these data restrict the crustal residence times of the St Kitts magmas to <8000 years, which clearly disagrees with the apparent U–Th ages. Furthermore, the U–Th data must be dominated by melt inclusions. These combined observations strongly suggest that the ages inferred above need to be treated with great caution; we suggest two possible interpretations of why the bulk mineral separates have higher (²³⁰Th/²³²Th) than their groundmass.

The important conclusion is that both interpretations require that the minerals did not crystallize from their present groundmass. In the first, the higher (²³⁰Th/²³²Th) ratios of the minerals, especially the plagioclase separates, reflect the presence of inclusions of melt that now have a higher (²³⁰Th/²³²Th) ratio than the groundmass as a result of ageing. Given the Ra constraints for the groundmass, this requires the entrainment of older crystals, or crystal cores, that had perhaps been stored in cumulates (compare St Vincent; Turner et al., 2003a). Assuming a scenario in which the minerals crystallized from and entrapped magma with the same initial (²³⁰Th/²³²Th) ratio as their groundmass, the age inferences above remain valid. However, if the minerals crystallized, and trapped melt inclusions, from magmas with different initial (²³⁰Th/²³²Th) ratios compared with their present groundmass, then age constraints cannot be obtained from Fig. 19. In the end-member case, where the initial (²³⁰Th/²³²Th) ratios of their parent and entrapped magmas were higher [the whole-rock (²³⁰Th/²³²Th) ratios for St Kitts lavas extend to >1·4], the minerals need not necessarily have differing ages from, or be significantly older than, their groundmass. Distinguishing between these competing models is not easy and, although in situ radiogenic isotope analysis (e.g. Davidson et al., 2000) of mineral–groundmass pairs might help, both interpretations demand that mixing was important in the Mt Liamuiga magmas.

	CONTRIBUTIONS TO THE MANTLE SOURCE OF ST KITTS MAGMAS

TOP ABSTRACT INTRODUCTION GENERAL GEOLOGY ANALYTICAL METHODS PETROLOGY AND MINERAL CHEMISTRY PETROGENESIS CONTRIBUTIONS TO THE MANTLE... CONCLUSIONS SUPPLEMENTARY DATA REFERENCES

 
Magnesian basalts from St Kitts are enriched in LILE and LREE, and depleted in middle REE (MREE) and HREE and HFSE, relative to N-MORB. LILE/LREE ratios tend to be high, especially Ba/La. On the basis of Th/La and Sm/La relationships, Plank (2005) showed that the primary magmas of the Lesser Antilles suites were generated in a mantle source (Sm/La 0·9) compositionally similar to that of N-MORB (Sm/La = 1·0). Although the element depletions relative to N-MORB may be related to slightly higher degrees of melting, the enrichments must be due to slab additions. Plank (2005) argued that Th–La–Sm relationships in mafic volcanic rocks of the northern islands were consistent with magma generation in mantle modified by addition of sediments identical to the actual composition of the bulk sedimentary column being subducted beneath the Northern Antilles arc.

Using mass-balance calculations for a Mt Liamuiga basalt with 7 wt % MgO, Turner et al. (1996) showed that the relative contribution of Th by subducted sediment to the whole-rock composition was 2%. Assuming around 10% partial melting and that Th is highly incompatible, this corresponds to 0·2% sediment addition to the mantle wedge. Using this mass balance, Turner et al. (1996) were able to show that the bulk of the Ba, Cs, K, Pb, Rb, Sr and U budgets of the primary magmas were contributed by a fluid component. Despite concerns about the potential for Cl- and F-rich fluids to transport Th as well as U [see discussion by Turner et al. (2003b)], the low (²³⁰Th/²³²Th) ratio of the Lesser Antilles sediments, relative to the erupted lavas, strongly argues that the fluids involved did not carry significant Th from the sediments because this would result in mixing arrays with a negative slope in Fig. 15. We now use our new U–Th isotopic data to comment on the timing of this fluid addition to the mantle sources.

The strong contrast between the northern and southern portions of the arc on the U–Th equiline diagram (Fig. 15) implies contrasting compositions in the mantle wedge beneath these areas. This is atypical, because it is commonly assumed that the Th isotope composition of the mantle wedge is normally dictated by that of the subducting sediment (Turner et al., 2003a). Nevertheless, it is well known that the composition of the sediments being subducted beneath the Lesser Antilles changes strongly along the arc (White & Dupré, 1986), although the composition of the sedimentary column beneath the southern part of the arc is poorly known (Plank, 2005). However, the inferred (²³⁰Th/²³²Th) ratio of the sediments varies in the wrong sense, changing from 0·34 in the north to 0·87 in the south, based on data of Plank & Langmuir (1998). Thus, the distinction on the equiline diagram cannot be due to this change in sediment composition in any simple way. For example, partial melting could potentially fractionate the U/Th ratio of the sediment and allow for the >30 kyr of ageing in order to return this modified composition to the equiline near the end of the lava array (e.g. Elliott et al., 1997; Turner et al., 1997). This would have to dramatically increase the U/Th ratio by sediment melting in the north (perhaps because of residual apatite or allanite) yet decrease it in the south (perhaps because to residual zircon). Clearly, only the significant, and difficult to constrain, involvement of accessory minerals would permit such models. A within-arc distinction on the equiline diagram unrelated to sediment composition was also found in Vanuatu and ascribed to differences between Indian and Pacific mantle in the wedge beneath that arc (Turner et al., 1999). However, we know of no evidence for large changes in the composition of the mantle wedge prior to sediment addition along the Lesser Antilles (see Plank, 2005).

The alternative explanation is that the amount of Th in the sediment added to the mantle wedge beneath the northern portion of the arc is insufficient (0·2%) to shift significantly its (²³⁰Th/²³²Th) ratio from a value similar to the MORB source (1·2), whereas the mantle wedge further south has a (²³⁰Th/²³²Th) ratio (0·87) more closely dictated by the larger amount of sediment addition (1–2%) that occurs there (Turner et al., 1996). The reason why the lavas from St Vincent plot with those of the northern islands can then be explained by a reduced amount of sediment addition beneath that island (Heath et al., 1998b), possibly as a result of locally efficient off-scraping into the accretionary wedge, which otherwise becomes increasingly prominent southwards along the arc (e.g. Clift & Vannucchi, 2004).

²³⁸U excesses are characteristic of island arc lavas, and the Lesser Antilles is no exception to this (Turner et al., 1996). Addition of U to the mantle wedge by fluids released from the downgoing slab is the most likely cause of these disequilibria and U–Th isotope data have been used to constrain the timing of this U addition in numerous studies [see Turner et al. (2003b) for the most recent review]. However, the whole-rock U–Th isotope data from Mt Liamuiga, and the Lesser Antilles more generally, do not form quasi-linear arrays like that found, for example, in the Marianas (Elliott et al., 1997). Where more scattered data are present, the best constraint may often be that U addition occurred <350 kyr ago, although modelling of Sr–Th isotope relationships has been used to infer a period of 90 kyr for some Lesser Antilles eruptive rocks (Turner et al., 1996). In some studies, a line through the base of the data array has often been used to constrain the time since fluid addition of U (e.g. Turner et al., 1997). However, Fig. 15 shows that the lavas from the northern islands, including St Kitts, and St Vincent lie above a 16 kyr reference line, whereas those from the southern islands, excepting St Vincent, scatter about a 45 kyr reference line (see Heath et al., 1998b). It should be noted that the observed ²²⁶Ra excesses can only be consistent with such ages via two-stage fluid addition models [see Turner et al. (2003b) for further discussion]; just how much significance can be attributed to the two arrays will have to await further investigations. Nevertheless, the present data make it increasingly clear that there are significant along-arc variations in mantle source composition, including the contribution from subducted sediment, in the Lesser Antilles.

	CONCLUSIONS

TOP ABSTRACT INTRODUCTION GENERAL GEOLOGY ANALYTICAL METHODS PETROLOGY AND MINERAL CHEMISTRY PETROGENESIS CONTRIBUTIONS TO THE MANTLE... CONCLUSIONS SUPPLEMENTARY DATA REFERENCES

 
(1) Primary magmas of the low-K Mt Liamuiga suite were generated in an N-MORB type mantle modified by sediment and fluid additions from the subducting slab. By analogy with rocks of Soufrière, St Vincent, the primary magmas were either extracted relatively dry (2 wt % H₂O) from depths of 35 km or, more probably, finally equilibrated with the mantle at these depths.

(2) Slightly different fractionation histories in the uppermost mantle and deep crust, resulting in differences in the temperature of clinopyroxene crystallization, produced higher-Al and lower-Al magma groups, which were available for eruption throughout the history of the centre.

(3) Both groups evolved dominantly by fractional crystallization of olivine–clinopyroxene–plagioclase–titanomagnetite–orthopyroxene assemblages, although field evidence of magma mingling and mineral chemical evidence of hybridization suggest that open-system processes were ubiquitous during magma evolution. Phenocryst core compositions are compatible with mixing between basalt, basaltic andesite and andesite in the hybrid rocks.

(4) Magmas of the higher-Al and lower-Al groups apparently did not mix, indicating the existence of at least two magma storage reservoirs under St Kitts.

(5) Pb isotopic and trace element (Ba/La, La/Sm) evidence points to minor assimilation of biogenic sediments within the crust, as recorded previously on the neighbouring island of Saba.

(6) U–Th disequilibria data show a marked distinction between the northern and southern parts of the Lesser Antilles, with St Vincent as an exception. Data from St Kitts range from near equilibrium to moderate U excesses but the north–south distinction in the data remains problematic. Mineral data are dominated by melt inclusions and require the entrainment of old crystals and/or that the bulk of the crystals crystallized from melts different from their groundmass. Thus, the scatter of the St Kitts U–Th data is inferred to reflect a combination of fluid addition of U, mixing and ageing.

	SUPPLEMENTARY DATA

TOP ABSTRACT INTRODUCTION GENERAL GEOLOGY ANALYTICAL METHODS PETROLOGY AND MINERAL CHEMISTRY PETROGENESIS CONTRIBUTIONS TO THE MANTLE... CONCLUSIONS SUPPLEMENTARY DATA REFERENCES

 
Supplementary data for this paper are available at Journal of Petrology online.

	ACKNOWLEDGEMENTS

 
This research was supported by the Royal Society (S.P.T.) and through NERC studentships to J. Toothill and C. A. Williams. We thank Peter Hill and Richard Hinton (Edinburgh) and Mabs Gilmore, Peter Evans, Louise Thomas and Peter van Calsteren (Open University) for help and guidance in the laboratory work, and John Taylor and Andy Lloyd (Open University) who produced the diagrams. Professors J. D. Devine, J. Gill and M. Wilson, and an anonymous referee provided rigorous and helpful reviews of the manuscript.

---

*Corresponding author. E-mail: r.macdonald{at}lancaster.ac.uk

	REFERENCES

TOP ABSTRACT INTRODUCTION GENERAL GEOLOGY ANALYTICAL METHODS PETROLOGY AND MINERAL CHEMISTRY PETROGENESIS CONTRIBUTIONS TO THE MANTLE... CONCLUSIONS SUPPLEMENTARY DATA REFERENCES

 
Arculus RJ. (1976) Geology and geochemistry of the alkali basalt–andesite association of Grenada, Lesser Antilles island arc. Geological Society of America Bulletin 87:612–624.[Abstract/Free Full Text]

Arculus RJ and Wills KJA. (1980) The petrology of plutonic blocks and inclusions from the Lesser Antilles island arc. Journal of Petrology 21:743–799.[Abstract/Free Full Text]

Bacon CR and Hirschmann MM. (1988) Mg/Mn partitioning as a test for equilibrium between coexisting Fe–Ti oxides. American Mineralogist 73:57–61.[Abstract]

Baker PE. (1968) Petrology of Mt. Misery volcano, St. Kitts, West Indies. Lithos 1:124–150.

Baker PE. (1969) The geological history of Mt. Misery Volcano, St. Kitts, West Indies. Overseas Geology and Mineral Resources 10:207–230.

Baker PE. (1980) Geology and geochemistry of the Mansion pyroclastic fall succession, St. Kitts. Bulletin Volcanologique 43:303–310.

Baker PE. (1984) Geochemical evolution of St Kitts and Montserrat, Lesser Antilles. Journal of the Geological Society, London 141:401–411.[Abstract/Free Full Text]

Baker PE. (1985) Volcanic hazards on St. Kitts and Montserrat, West Indies. Journal of the Geological Society, London 142:279–295.[Abstract/Free Full Text]

Baker PE and Holland JG. (1973) Geochemical variations in a pyroclastic succession on St. Kitts, West Indies. Bulletin Volcanologique 37:472–490.[CrossRef]

Baker PE, Buckley F, Padfield T. (1980) Petrology of the volcanic rocks of Saba, West Indies. Bulletin Volcanologique 43:337–347.

Barclay J, Rutherford MJ, Carroll MR, Murphy MD, Devine JD, Gardner J, Sparks RSJ. (1998) Experimental phase equilibria constraints on pre-eruptive storage conditions of the Soufrière Hills magma. Geophysical Research Letters 25:3437–3440.[CrossRef][Web of Science]

Belshaw NS, Freedman PA, O'Nions RK, Frank M, Guo Y. (1998) A new variable dispersion double-focusing plasma mass spectrometer with performance illustrated for Pb isotopes. International Journal of Mass Spectrometry and Ion Processes 181:51–58.[CrossRef]

Blundy JD and Wood BJ. (2003) In Bourdon B, Turner SP, Henderson G, Lundstrom C (Eds.). Mineral–melt partitioning of uranium, thorium and their daughters. Uranium Series Geochemistry. Reviews of Mineralogy and Geochemistry 52:59–123.

Brown GM, Holland JG, Sigurdsson H, Tomblin JF, Arculus RJ. (1977) Geochemistry of the Lesser Antilles island arc. Geochimica et Cosmochimica Acta 41:785–801.[CrossRef][Web of Science]

Chabaux F, Hemond C, Allègre CJ. (1999) ²³⁸U–²³⁰Th–²²⁶Ra disequilibria in the Lesser Antilles arc: implications for mantle metasomatism. Chemical Geology 153:171–185.[CrossRef][Web of Science]

Clift P and Vannucchi P. (2004) Controls on tectonic accretion versus erosion in subduction zones: implications for the origin and recycling of the continental crust. Reviews of Geophysics 42: doi:10.1029/2003RG000127.

Cooper KM and Reid MR. (2003) Re-examination of crystal ages in recent Mount St. Helens lavas: implications for magma reservoir processes. Earth and Planetary Science Letters 213:149–167.[CrossRef][Web of Science]

Davidson JP. (1985) Mechanisms of contamination in Lesser Antilles island arc magmas from radiogenic and oxygen isotope relationships. Earth and Planetary Science Letters 72:163–174.[CrossRef][Web of Science]

Davidson JF. (1986) Isotopic and trace-element constraints on the petrogenesis of subducted-related lavas from Martinique. Journal of Geophysical Research 91:5943–5962.

Davidson JP. (1987) Crustal contamination versus subduction zone enrichment: examples from the Lesser Antilles and implications for mantle source compositions of island arc volcanic rocks. Geochimica et Cosmochimica Acta 51:2185–2198.[CrossRef][Web of Science]

Davidson JP and Harmon RS. (1989) Oxygen isotope constraints on the petrogenesis of volcanic arc magmas from Martinique, Lesser Antilles. Earth and Planetary Science Letters 95:255–270.[CrossRef][Web of Science]

Davidson JP, Teply FJ III, Palacz Z, Meffan-Main S. (2000) Magma recharge, contamination and residence times revealed by in situ laser ablation isotopic analysis of feldspar in volcanic rocks. Earth and Planetary Science Letters 184:427–442.[Web of Science]

Defant MJ, Sherman S, Maury RC, Bellon H, de Boer J, Davidson J, Kepezhinskas P. (2001) The geology, petrology, and petrogenesis of Saba island, Lesser Antilles. Journal of Volcanology and Geothermal Research 107:87–111.[CrossRef][Web of Science]

DePaolo DJ. (1981) Trace element and isotopic effects of combined wallrock assimilation and fractional crystallisation. Earth and Planetary Science Letters 53:189–202.[CrossRef][Web of Science]

Devine JD and Sigurdsson H. (1983) The liquid composition and crystallization history of the 1979 Soufrière magma. Journal of Volcanology and Geothermal Research 16:1–31.[CrossRef][Web of Science]

Droop GTR. (1987) A general equation for estimating Fe³⁺ concentrations in ferromagnesian silicates and oxides from microprobe analyses, using stoichiometric criteria. Mineralogical Magazine 51:431–435.[Web of Science]

Elliott T, Plank T, Zindler A, White WM, Bourdon B. (1997) Element transport from slab to volcanic front at the Mariana arc. Journal of Geophysical Research 102:14991–15019.[CrossRef]

Ghiorso MS and Sack RO. (1995) Chemical mass transfer in magmatic processes IV. A revised and internally consistent thermodynamic model for the interpolation and extrapolation of liquid–solid equilibria in magmatic systems at elevated temperatures and pressures. Contributions to Mineralogy and Petrology 119:187–212.

Gill JB and Williams RW. (1990) Th isotope and U-series studies of subduction-related volcanic rocks. Geochimica et Cosmochimica Acta 54:1427–1442.[CrossRef][Web of Science]

Harkness DD, Roobol MJ, Smith AL, Stipp JJ, Baker PE. (1994) Radiocarbon redating of contaminated samples from a tropical volcano—the Mansion Series of St Kitts, West Indies. Bulletin of Volcanology 56:326–334.[Web of Science]

Hawkesworth CJ and Powell M. (1980) Magma genesis in the Lesser Antilles island arc. Earth and Planetary Science Letters 51:297–308.[CrossRef][Web of Science]

Heath E, Macdonald R, Belkin HE, Hawkesworth CJ, Sigurdsson H. (1998a) Magmagenesis at Soufrière volcano, St Vincent, Lesser Antilles arc. Journal of Petrology 39:1721–1764.[CrossRef][Web of Science]

Heath E, Turner SP, Macdonald R, Hawkesworth CJ, van Calsteren P. (1998b) Long magma residence times at an island arc volcano (Soufrière, St. Vincent) in the Lesser Antilles: evidence from ²³⁸U–²³⁰Th isochron dating. Earth and Planetary Science Letters 160:49–63.[CrossRef][Web of Science]

Higgins MD. (2000) Measurement of crystal size distributions. American Mineralogist 85:1105–1116.[Abstract/Free Full Text]

Jarrard RD. (1986) Relations among subduction parameters. Reviews of Geophysics 24:217–284.[Web of Science]

Jerram DA, Cheadle MC, Philpotts AR. (2003) Quantifying the building blocks of igneous rocks: are clustered crystal frameworks the foundation? Journal of Petrology 44:2033–2051.[Abstract/Free Full Text]

Kay RW. (1980) Volcanic arc magmas: implications of a melting–mixing model for element cycling in the crust–upper mantle system. Journal of Geology 88:497–522.[Web of Science]

LePage LD. (2003) ILMAT: an Excel worksheet for ilmenite–magnetite geothermometry and geobarometry. Computers and Geosciences 29:673–678.[CrossRef]

Lindsley DH. (1983) Pyroxene thermometry. American Mineralogist 68:477–493.[Abstract]

Macdonald R, Hawkesworth CJ, Heath E. (2000) The Lesser Antilles volcanic chain: a study in arc magmatism. Earth-Science Reviews 49:1–76.

Marsh BD. (1998) On the interpretation of crystal size distributions in magmatic systems. Journal of Petrology 39:553–599.[CrossRef][Web of Science]

Martin-Kaye PHA. (1959) Reports on the Geology of the Leeward and British Virgin Islands(St Lucia, Voice Publishing Co.) pp. 117.

Maury RC and Westercamp D. (1990) Geology of the Lesser Antilles. In Dengo JE (Ed.). The Geology of North America. Volume H, The Caribbean Region(Geological Society of America, Boulder, CO) pp. 141–166.

Miyashiro A. (1974) Volcanic rock series in island arcs and active continental margins. American Journal of Science 274:321–355.[Abstract]

Morgan DJ and Jerram DA. (2006) On estimating crystal shape for crystal size distribution analysis. Journal of Volcanology and Geothermal Research 154:1–7.[CrossRef][Web of Science]

Nielsen C and Sigurdsson H. (1981) Quantitative methods for electron microprobe analysis of sodium in natural and synthetic glasses. American Mineralogist 66:547–552.[Abstract]

Ottley CJ, Pearson DG, Irvine GJ. (2003) A routine method for the dissolution of geological samples for the analysis of REE and trace elements via ICP-MS. In Holland JG and Tanner SD (Eds.). Plasma Source Mass Spectrometry: Applications and Emerging Technologies(Royal Society of Chemistry, London) pp. 221–230.

Pichavant M and Macdonald R. (2003) In Larter RD and Leat PT (Eds.). Mantle genesis and crustal evolution of primitive calc-alkaline basaltic magmas from the Lesser Antilles arc. Intra-Oceanic Subduction Systems: Tectonic and Magmatic Processes. Geological Society, London, Special Publications 219:239–254.

Pichavant M, Mysen BO, Macdonald R. (2002a) Source and H₂O content of high-MgO magmas in island arc settings: an experimental study of a primitive calc-alkaline basalt from St Vincent, Lesser Antilles arc. Geochimica et Cosmochimica Acta 66:2193–2209.[CrossRef][Web of Science]

Pichavant M, Martel C, Bourdier J-L, Scaillet B. (2002b) Physical conditions, structure, and dynamics of a zoned magma chamber: Mount Pelée (Martinique, Lesser Antilles Arc). Journal of Geophysical Research 107:B5 10.1029/2001JB000315.

Plank T. (2005) Constraints from thorium/lanthanum on sediment recycling at subduction zones and the evolution of the continents. Journal of Petrology 46:921–944.[Abstract/Free Full Text]

Plank T and Langmuir CH. (1998) The chemical composition of subducting sediment and its consequences for the crust and mantle. Chemical Geology 145:325–394.[CrossRef][Web of Science]

Rea WJ and Baker PE. (1980) The geochemical characteristics and conditions of petrogenesis of the volcanic rocks of the northern Lesser Antilles—a review. Bulletin Volcanologique 43:325–336.

Roeder PL and Emslie RF. (1970) Olivine–liquid equilibrium. Contributions to Mineralogy and Petrology 29:275–289.[CrossRef][Web of Science]

Rollinson H. (1993) Using Geochemical Data: Evaluation, Presentation, Interpretation(Longman, Harlow) pp. 352.

Roobol MJ, Smith AL, Wright JV. (1981) Revisions in the pyroclastic stratigraphy of Mt. Misery volcano, St. Kitts, Lesser Antilles—C¹⁴ ages and recognition of pyroclastic flow deposits. Journal of the Geological Society , London 138:713–718.[Abstract/Free Full Text]

Rosencrantz E and Sclater JG. (1986) Depth and age of the Cayman Trough. Earth and Planetary Science Letters 79:133–144.[CrossRef][Web of Science]

Saunders AD, Norry MJ, Tarney J. (1988) Origin of MORB and chemically depleted mantle reservoirs: trace-element constraints. Journal of Petrology, Special Lithosphere Issue 415–445.

Sigurdsson H, Tomblin JF, Brown GM, Holland JG, Arculus RJ. (1973) Strongly undersaturated magma in the Lesser Antilles island arc. Earth and Planetary Science Letters 18:285–295.[CrossRef][Web of Science]

Sisson TW and Grove TL. (1993) Experimental investigations of the role of H₂O in calc-alkaline differentiation and subduction zone magmatism. Contributions to Mineralogy and Petrology 113:143–166.[CrossRef][Web of Science]

Smith TE, Thirlwall MF, MacPherson C. (1996) Trace-element and isotope geochemistry of the volcanic rocks of Bequia, Grenadine Islands, Lesser Antilles arc: a study of subduction enrichment processes. Journal of Petrology 37:117–143.[Abstract/Free Full Text]

Spencer KJ and Lindsley DH. (1981) A solution model for coexisting iron–titanium oxides. American Mineralogist 68:586–594.

Sun S-S and McDonough WF. (1989) In Saunders AD and Norry MJ (Eds.). Chemical and isotopic systematics of oceanic basalts: implications for mantle composition and processes. Magmatism in the Ocean Basins. Geological Society, London, Special Publications 42:313–345.

Tatsumi Y and Eggins SM. (1995) Subduction Zone Magmatism(Blackwell, Oxford) pp. 211.

Thirlwall MF and Graham AM. (1984) Evolution of high-Ca, high-Sr C-series basalts from Grenada, Lesser Antilles: the effects of intra-crustal contamination. Journal of the Geological Society, London 141:427–445.[Abstract/Free Full Text]

Thirlwall MF, Smith TE, Graham AM, Theodorou N, Hollings P, Davidson JP, Arculus RJ. (1994) High field strength element anomalies in arc lavas: source or process? Journal of Petrology 35:819–838.[Abstract/Free Full Text]

Turner S, Hawkesworth CJ, van Calsteren P, Heath E, Macdonald R, Black S. (1996) U-series isotopes and destructive plate margin magma genesis in the Lesser Antilles. Earth and Planetary Science Letters 142:191–207.[CrossRef][Web of Science]

Turner S, Hawkesworth CJ, Rogers NW, Bartlett J, Worthington T, Hergt J, Pearce JA, Smith I. (1997) ²³⁸U–²³⁰Th disequilibria, magma petrogenesis, and flux rates beneath the depleted Tonga–Kermadec island arc. Geochimica et Cosmochimica Acta 61:4855–4884.[CrossRef][Web of Science]

Turner SP, Peate DW, Hawkesworth CJ, Eggins SM, Crawford AJ. (1999) Two mantle domains and the time scales of fluid transfer beneath the Vanuatu arc. Geology 27:963–966.[Abstract/Free Full Text]

Turner S, van Calsteren P, Vigier N, Thomas L. (2001a) Determination of thorium and uranium isotope ratios in low-concentration geological materials using a fixed multicollector ICP-MS. Journal of Analytical Atomic Spectroscopy 16:612–615.[CrossRef]

Turner S, Evans P, Hawkesworth C. (2001b) Ultra-fast source-to-surface movement of melt at island arcs from ²²⁶Ra–²³⁰Th systematics. Science 292:1363–1366.[CrossRef][Web of Science][Medline]

Turner S, George R, Jerram DA, Carpenter N, Hawkesworth C. (2003a) Case studies of plagioclase growth and residence times in island arc lavas from Tonga and the Lesser Antilles, and a model to reconcile discordant age information. Earth and Planetary Science Letters 214:279–294.[CrossRef][Web of Science]

Turner S, Bourdon B, Gill J. (2003b) In Bourdon B, Henderson G, Lundstrom C, Turner S (Eds.). Insights into magma genesis at convergent margins from U-series isotopes. Uranium Series Geochemistry. Reviews of Mineralogy and Geochemistry 52:255–313.

van Calsteren P and Schwieters JB. (1995) Performance indications for TIMS with a deceleration lens system and post-deceleration detector selection. International Journal of Mass Spectrometry and Ion Processes 146–147:119–129.[CrossRef]

Wadge G. (1984) Comparison of volcanic production rates and subduction rates at the Lesser Antilles and Central America. Geology 12:555–558.[Abstract/Free Full Text]

Wadge G and Shepherd JB. (1984) Segmentation of the Lesser Antilles subduction zone. Earth and Planetary Science Letters 71:297–304.[CrossRef][Web of Science]

Westermann JH and Kiel H. (1961) The geology of Saba and St Eustatius. Publications of the Foundation for Science in Surinam and the Netherlands Antilles 24:1–175.

White WM and Dupré B. (1986) Sediment subduction and magma genesis in the Lesser Antilles: isotopic and trace element constraints. Journal of Geophysical Research 91:5927–5941.

Williams CA. (2000) The geochemistry of Mt. Misery volcano, St. Kitts, Lesser Antilles. Ph.D. thesis The Open University.

CiteULike    Connotea    Del.icio.us    What's this?

This Article Abstract FREE Full Text (PDF) Supplementary data All Versions of this Article: 48/1/3    most recent egl052v1 Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Add to My Personal Archive Download to citation manager Request Permissions Google Scholar Articles by Toothill, J. Articles by Tindle, A. G. Search for Related Content GeoRef GeoRef Citation Social Bookmarking          What's this?

Online ISSN 1460-2415 - Print ISSN 0022-3530

Copyright © 2010 Oxford University Press

Oxford Journals Oxford University Press

• Site Map
• Privacy Policy
• Frequently Asked Questions

Other Oxford University Press sites: Oxford University Press Oxford Journals China Oxford Journals Japan American National Biography Booksellers' Information Service Children's Fiction and Poetry Children's Reference Corporate & Special Sales Dictionaries Dictionary of National Biography Digital Reference English Language Teaching Higher Education Textbooks Humanities International Education Unit Law Medicine Music Online Products Oxford English Dictionary Reference Rights and Permissions Science School Books Social Sciences Very Short Introductions World's Classics