Noritic anorthosite, gabbroic anorthosite and hornblende-gabbro xenoliths are ubiquitous in the host andesite at Montserrat. Other xenoliths include quartz diorite, metamorphosed biotite-gabbro, plagioclase-hornblendite and plagioclase-clinopyroxenite. Mineral compositions suggest a majority of the xenoliths are cognate. Cumulate, hypabyssal and crescumulate textures are present. A majority of the xenoliths are estimated to have seismic velocities of 6.7–7.0 km/s for pore-free assemblages. These estimates are used in conjunction with petrological models to constrain the SEA CALIPSO seismic data and the structure of the crust beneath Montserrat. Andesitic upper crust is interpreted to overlie a lower crust dominated by amphibole and plagioclase. Xenolith textures and seismic data indicate the presence of hypabyssal intrusions in the shallow crust. The structure of the crust is consistent with petrological models indicating that fractionation is the dominant process producing andesite at Montserrat.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The composition of arc crust is a fundamental issue in earth sciences. Seismic surveys suggest that arcs are significantly thicker than oceanic crust, with inferred compositions more mafic than continental crust [e.g., Christensen and Mooney, 1995]. Most of our knowledge derives from seismic surveys, exhumed crustal sections and crustal xenoliths. Here igneous xenoliths sampled from andesite lavas on Montserrat are described and used, together with petrological information, to help interpret seismic velocity data obtained from the SEA CALIPSO project. Together these data provide constraints on the crustal structure beneath Montserrat.
 Montserrat is located in the Lesser Antilles island arc, related to subduction of the Atlantic plate beneath the Caribbean plate. Arc volcanism has been active since the Cretaceous and shifted westward during the Miocene, producing a double island chain. Early integrated gravity and seismic surveys determined an average crustal thickness in the Lesser Antilles of ∼30 km, with a heterogeneous upper crust (Vp = 6.2 km/s) of variable thickness, and a higher velocity lower crust (Vp = 6.9 km/s [Westbrook and McCann, 1986, and references therein]). However a more detailed seismic survey by Christeson et al.  reported an average crustal thickness of 24 km.
 The four volcanic centres of Montserrat date back to 2.6 Ma and have generally produced andesitic deposits, with the exception of basaltic and basaltic andesite lavas from South Soufrière Hills [Harford et al., 2002]. The current eruption of Soufrière Hills Volcano (SHV) began in 1995, characterised by phases of dome growth and collapse. Deposits include andesitic domes and pyroclastic, lahar and debris avalanche deposits. Evidence from deformation studies [Mattioli et al., 1998] and melt inclusions [Devine et al., 1998] indicate a magma chamber at 5–6 km depth beneath the SHV. A deeper chamber (∼12 km) has been inferred from deformation modelling combined with magma extrusion volumes [Elsworth et al., 2008]. Mafic magmatic inclusions in the host andesite suggest repeated input of basalt and basaltic andesite magma accompanied by reheating [e.g., Sparks et al., 1998; Murphy et al., 1998].
 Crustal xenoliths were collected from all four volcanic centres. Modal mineral proportions were obtained for a representative selection of xenoliths, host andesite and mafic magmatic inclusions by point counting using an optical microscope. Mineral phases were analysed for major elements using the Cameca SX100 electron microprobe at Bristol University with a 20 kV accelerating voltage and 10 nA beam current. Textures were analysed using the Scanning Electron Microscope (SEM).
 Seismic velocities of xenoliths were estimated from modal proportions and elastic properties of individual minerals, assuming isotropic fabrics, using the following equation:
where M* is the bulk or shear modulus of the composite; vi is the volumetric proportion of the ith mineral; and Mi is the bulk or shear modulus of the ith mineral [Watt et al., 1976, and references therein]. The Voigt average (MV) assumes uniform strain (t = 1) and the Ruess average (MR) assumes uniform stress (t = −1) throughout the polymineralic rock. Because averaging schemes effectively provide upper and lower bounds, the arithmetic mean (MV + MR)/2, or the ‘Hill average’, is commonly used [Watt et al., 1976]. Unless stated otherwise, quoted velocities have been calculated using the arithmetic mean. Elastic constants were taken from the compilation of Hacker and Abers [2004, and references therein]. Average mineral compositions were used to extrapolate between end-member elastic constants, except for plagioclase, for which we used the relationship of Angel . The effect of porosity on seismic velocities was estimated using the following equation [Wyllie et al., 1958]:
where ϕ is fractional porosity, Vf and Vm are the seismic velocities of the pore fluid and the rock matrix respectively.
 Most xenoliths are intrusive igneous rocks. They can be classified as: noritic anorthosites, gabbroic anorthosites, and hornblende-gabbros. Other less common xenoliths include: quartz diorite, metamorphosed biotite-gabbro, and nearly pure (80–86%) monomineralic rocks including: plagioclase-hornblendite and plagioclase-clinopyroxenite. Xenolith modal mineralogy is characterised by varying proportions of plagioclase, amphibole, orthopyroxene, clinopyroxene, titanomagnetite, ilmenite and quartz, similar to mineral assemblages in the host andesite (Table 1). Montserrat xenoliths have similar mineral assemblages to xenoliths found throughout the Lesser Antilles [Arculus and Wills, 1980], apart from the absence of olivine.
Table 1. Vesicle-Free Mineral Proportions, Calculated Compressional Wave Velocities and Laboratory Velocity Measurements of the Main Rock Types
 The xenoliths are mostly unlayered and isotropic. Sharp contacts with the host andesite and absence of chilled margins suggest the xenoliths were largely or entirely consolidated prior to entrainment, in contrast to the mafic magmatic inclusions (Figure 1). Many of the xenoliths have orthocumulate textures with vesiculated groundmasses (<26 vol%) and abundant plagioclase microlites, similar to xenoliths from other Antilles islands [Arculus and Wills, 1980]. This indicates the presence of partially quench crystallised interstitial melt (<42%). Miarolitic cavities partially infilled with secondary cristobalite together with the fine-medium grain size indicate that most xenoliths crystallised in hypabyssal intrusions. Some samples display adcumulate and crescumulate textures. One sample consists of alternating pyroxene-anorthosite and plagioclase-clinopyroxenite layers, with crescumulate textures normal to the mineral layering.
 Plagioclase compositions of most xenoliths (An46–91) show similar variation to the host andesite and mafic inclusions [Murphy et al., 1998]. Crystals display normal, reverse and oscillatory zoning consistent with repeated injections of mafic magma, as interpreted for the host andesite [e.g., Sparks et al., 1998; Murphy et al., 1998; Zellmer et al., 2003b]. Sodic plagioclase microlites (An20–37) are consistent with crystallisation from a late-stage melt.
 Mafic phases include clinopyroxene (En37–40Wo38–45) ± orthopyroxene (En54–60Wo2–4) ± magnesio-hornblende [Leake et al., 1997]. No significant core-rim variations are observed in a majority of the xenoliths. Hornblende typically contains 6.3–8.8 wt% Al2O3, with magnesium numbers of 58–64, similar in range to the host andesite [Murphy et al., 1998]. A few xenoliths have more Al-rich hornblende similar to the mafic inclusions (<14 wt% Al2O3 [Murphy et al., 1998]). Mineral compositions of most xenoliths are similar to the host andesite, which is consistent with a co-genetic origin. The lack of suitable mineral assemblages has hindered estimates of xenolith equilibration pressures.
 Hornblende from plagioclase-hornblendite has significantly higher magnesium numbers of 70–76. The layered crescumulate sample is compositionally distinct, with highly calcic plagioclase (An79–87) and a small proportion of more magnesium-rich orthopyroxene (En65–67). Sodic plagioclase rims (An36–64) and Fe-rich orthopyroxene rims (En46–52) are present in the pyroxene-anorthosite layer. Crescumulate textures are interpreted as resulting from rapid crystal growth from a supercooled melt at the margins of a magma body [e.g., Donaldson, 1977], with rim compositions indicative of infiltration of the crystal mush by more evolved melts.
4.2. Velocity Estimates
 Velocity estimates of the xenoliths are very similar despite the range of mineral assemblages (Figure 2). Velocity of most xenoliths calculated from their primary modal mineralogy show good correlation with laboratory measurements of similar rocks (Table 1) [Christensen and Mooney, 1995]. However calculated velocities of the andesite and mafic magmatic inclusions are significantly higher. Alteration minerals and a small proportion of glass observed in these rocks may produce some of this variation.
 Velocities have been calculated for three scenarios: (1) no pores, assuming that pores are only abundant in the shallow crust; (2) all pores filled with water; and (3) all pores filled with secondary cristobalite, using elastic constants of Yeganeh-Haeri et al.  (Figure 2). Water-saturated pores dramatically reduce seismic velocity estimates by up to 2.9 km/s for highly porous samples. Secondary cristobalite reduces velocities by <0.7 km/s.
4.3. SEA CALIPSO Results
 The upper crust beneath Montserrat is thought to consist largely of andesite based on surface geology and our new observations of xenoliths from shallow intrusions. An intrusive complex could explain the high velocity core beneath the island imaged by the SEA CALIPSO project [Paulatto et al., 2010; E. Shalev et al., Three-dimensional seismic velocity tomography of Montserrat from the SEA-CALIPSO offshore/onshore experiment, submitted to Geophysical Research Letters, 2010]. At the greatest resolvable depth (∼8 km [Paulatto et al., 2010]) seismic tomography results yield values lower than all calculated pore-free velocities, indicating that porosity is an important control of velocity in the uppermost crust. Field observations indicate that volcaniclastic rocks likely dominate the near-surface, with primary and secondary porosity from vesicles and inter-particle spaces.
 W. Sevilla et al. (Crustal structure beneath the Montserrat region of the Lesser Antilles Island Arc, manuscript in preparation, 2010), have produced a receiver function profile that resolves the Moho at ∼30 km depth. A mid-crustal layer ∼1 km thick with velocities of 6.0–6.7 km/s may be a transitional layer between the upper and lower crust. Lower crustal velocities of 6.7–7.0 km/s (Sevilla et al., manuscript in preparation, 2010) match estimated velocities of plagioclase-hornblendite and basaltic to basaltic andesite mafic inclusions, and measured velocities of gabbro, norite, anorthosite and hornblendite (Figure 3) [Christensen and Mooney, 1995]. Relict oceanic crust may also be present in the lower crust.
4.4. Constraints From Petrology
 Several igneous processes can lead to layering of island arc crust, including partial melting of older crust and crystallisation of basalt, together with segregation of residual evolved melts. These scenarios produce physically similar layering comprising differentiated upper crust and denser more mafic cumulates or restite in the lower crust [Annen et al., 2006]. Thus it is difficult to distinguish between these processes from seismic velocity data alone.
 Magmatic parents to the andesite magma could be represented by the mafic inclusions or the South Soufrière Hills basalt and basaltic andesite. Zellmer et al. [2003a] calculated that the most evolved mafic inclusion could be produced from the least evolved mafic inclusion by fractional crystallisation of 49% amphibole and 21% plagioclase. A further ∼10% crystallisation is necessary to produce the Th concentration of the most evolved andesite. The andesite can also be produced by 65% crystallisation of plagioclase and amphibole from the least evolved South Soufrière Hills lava [Zellmer et al., 2003a].
 Fractionation models [Zellmer et al., 2003a] thus imply that an andesitic upper crust should be complemented by cumulates of plagioclase and amphibole in the lower crust. Upper/middle crust is ∼10 km thick to the island surface based on the receiver function profile (Sevilla et al., manuscript in preparation, 2010), and lower crust is ∼21 km thick. The ratio of upper to lower crust is therefore 1:2. This is consistent with the 65% South Soufrière Hills fractionation model [Zellmer et al., 2003a]. Assuming the lower crust dominantly comprises plagioclase and amphibole, observed seismic velocities (Sevilla et al., manuscript in preparation, 2010) correspond to 30–80% hornblende and corresponding plagioclase. These fractionation models are based on a parental magma that is evolved with respect to primitive melts, therefore additional cumulates of pyroxene, olivine and plagioclase are likely to be present in the lower crust. The calculated velocity of plagioclase-pyroxenite is 7.5 km/s (Figure 3) therefore the velocities of ultramafic cumulates could be >7.7 km/s and thus classed as sub-Moho.
 Some evidence indicates that the andesite is at least partially produced by anatexis. Hydrogen isotope analyses of amphibole [Harford and Sparks, 2001] and bulk U/Th ratios indicate partial melting and remobilisation of previous intrusions [Zellmer et al., 2003a]. Tatsumi et al.  suggest that ∼30% melting of basaltic crust could produce andesitic magma. This would produce an upper to lower crust ratio of 1:2 also similar to the observed crustal structure. Oxygen isotopes are consistent with a contribution of 10–20% hydrothermally altered crust. However trace element concentrations can largely be modelled by fractional crystallisation [Zellmer et al., 2003a]. The presence of cumulate textured xenoliths, together with petrological and geochemical evidence, favours a dominant role of intrusion and fractionation of incremental additions of basalt in the model of arc crust formation at Montserrat.
 Igneous xenoliths are present in the host lavas at Montserrat, with cumulate and hypabyssal textures. Most have mineral assemblages and mineral compositions indicating that they represent hypabyssal intrusive equivalents of the andesite, or cumulate rocks formed by fractionation of basaltic andesite and andesite. The presence of partially infilled vesicles in most xenoliths indicates shallow crystallisation. Surface geology and seismic velocities [Paulatto et al., 2010] are most consistent with an upper crust composed of andesitic volcanic and related intrusive rocks. The xenoliths and fast velocity regions beneath the volcanic centres of Montserrat [Paulatto et al., 2010; Shalev et al., submitted manuscript, 2010] support the interpretation of intrusive complexes. Petrological models [Zellmer et al., 2003a] suggest the lower crust contains cumulate rocks dominated by amphibole and plagioclase, which is consistent with seismic velocities (Sevilla et al., manuscript in preparation, 2010). The proportion of upper intermediate crust to cumulate or restitic lower crust indicates that andesite has been produced by fractionation of the South Soufrière Hills lava [Zellmer et al., 2003a] or partial melting of an initial basaltic crust [Tatsumi et al., 2008]. A model of arc crust formation largely by fractionation is consistent with xenoliths, petrology and the SEA CALIPSO seismic velocity data.
 This work is funded by the Natural Environment Research Council and British Geological Survey. EJK acknowledges N. Starkey, A. Dabrowa and the Montserrat Volcano Observatory for help in the field. EJK also thanks Stuart Kearns for help with EMPA analyses and Richard Arculus for useful discussion. RSJS acknowledges support of a European Research Council Advanced grant. BRE acknowledges support from the Dickinson College Research and Development Committee and C. Endress for field assistance. The authors also thank Georg Zellmer and J.S. Daly for helpful and constructive reviews.