Heterogeneous clasts as windows into magma mingling at Soufrière Hills volcano



[1] A clast erupted from Soufrière Hills volcano during the syn-collapse Vulcanian explosions of July 2003 reveals several incorporated silicic zones within a more mafic host. These incorporated silicics display variable microlite contents that reflect different stages of magma crystallization. One (zone A) shows evidence of heating in several mineralogical features, including completely pseudomorphed hornblende crystals that broke down over a minimum time period of 37–78 days. The lower limit of this time period roughly correlates with the onset of hybrid seismic swarms prior to the 2003 dome collapse. Element X-ray maps reveal percolation of Ca-rich melt into the incorporated silicic margins, suggesting sustained contact that allowed isolation of portions of the resident magma. These characteristics suggest that magma mixing at Soufrière Hills may not only result from disaggregation of mafic inclusions, but also from disaggregation and incorporation of the resident magma into the mafic intrusion.

1. Introduction

[2] Mingling of magmas of different compositions and temperatures occurs at many volcanoes and not only influences the composition of erupted products, but possibly initiates eruptive activity. Processes of magma mingling introduce thermal and chemical gradients into a system [Eichelberger, 1980; Sparks and Marshall, 1986; Huppert and Sparks, 1988], and may evolve to the point of complete mixing, where a single, uniform melt composition is formed from two distinct end members [Anderson, 1976]. The current eruption (1995–2010) of Soufrière Hills volcano (SHV) is thought to have initiated through the remobilization of andesite crystal mush by the intrusion of mafic magma into the chamber [Devine et al., 1998; Murphy et al., 1998, 2000]. Because thermal diffusivity is much faster than chemical diffusivity [Sparks and Marshall, 1986; Huppert and Sparks, 1988] heating of the resident magma should be the first consequence of mafic influx, and evidence of heating is observed in the petrologic record at SHV through the existence of sieve-textured zones in plagioclase phenocrysts, pseudomorphed hornblendes, and resorbed quartz crystals [Barclay et al., 1998; Devine et al., 1998; Murphy et al., 1998, 2000; Rutherford and Devine, 2003].

[3] Murphy et al. [2000] hypothesized underplating of the SHV andesite by the mafic intrusion, with dikes of the hotter, mafic magma ascending through the cooler, resident andesite, resulting in convective mixing. Studies conducted to examine methods of magma mixing in compositionally layered chambers have shown that the development of bubble-rich foams at layer boundaries may result in the formation of discrete plumes that rise upward from the mafic magma into the silicic magma [Eichelberger, 1980; Huppert and Sparks, 1988; Phillips and Woods, 2002]. In the case of convective self-mixing, mafic magma that underplates the resident andesite creates a thermal boundary layer, causing buoyant plumes of heated andesite at the magma interface to ascend [Couch et al., 2001; Devine et al., 2003]. The various magma mixing processes hypothesized for SHV attest to the complexity of the system.

[4] Mafic inclusions are observed to occupy a steadily increasing percentage of the SHV eruptive products. During 1995–1999, the proportion of mafic enclaves was estimated at <1 vol% [Murphy et al., 1998, 2000], with an increase to at least 5 vol% in the 2002 products (C. Mann, personal communication, 2009) and up to 7 vol% in 2007 products (J. Barclay et al., Caught in the act: Implications for the increasing abundance of mafic enclaves during the eruptions of the Soufrière Hills Volcano, Montserrat, manuscript in preparation, 2009). Recent studies suggest that mass may be transferred between the mafic and andesite magmas during ascent through the volcanic conduit, where disaggregation of mafic inclusions due to shear stress incorporates mafic microlites [Humphreys et al., 2009] or mafic melt [Humphreys et al., 2010] into the andesite groundmass. As this study reveals, other instances of magma mingling at SHV may be the result of the mafic intrusion incorporating portions of the cooler, more silicic magma from the chamber margins (walls and roof). As these incorporated silicics are isolated and disaggregated, they may serve as an additional source for many of the petrologic characteristics observed in the SHV eruptive products, including quartz with resorbed rims, pseudomorphed hornblendes, and a wide range of plagioclase compositions and textures within a single clast [Murphy et al., 1998, 2000].

2. Methods

[5] Tephra fall samples erupted during the Vulcanian events of July 2003 [Herd et al., 2005; Edmonds and Herd, 2007] were obtained from the Montserrat Volcano Observatory's catalogued collection. Clasts were sectioned, photographed, and prepared for petrographic and electron microscopic analyses. Helium pycnometry was used to measure clast density and vesicularity. One sample erupted during the syn-collapse explosions (clast 15a) best displays the details of small-scale (<1 cm) mingling processes by which portions of the mafic intrusion incorporates the more silicic resident magma within the chamber. Observed zones within the clast were differentiated on the basis of variable textures and each is described in the following section. Number densities/morphologies of microlite phases were documented from backscattered electron images taken at 1500x magnification using a Jeol IC-845A scanning electron microscope (SEM) operating at an accelerating voltage of 15 kV and a working distance of 15 mm. SEM backscattered images were also used to measure the thicknesses of hornblende breakdown rims. SEM element X-ray maps were collected to reveal compositional variations within clast 15a and allow identification of different mineral phases. Chemical analyses were performed using a secondary ion mass spectrometer (SIMS) with a ∼25 μm-diameter 16O primary beam operating at a current of 2–5 nA. 42Ca/30Si ratios of plagioclase phenocrysts and microphenocrysts in zone A were analyzed and three standards (Amelia albite, Lake County plagioclase, Stillwater plagioclase) were used to calibrate the anorthite (An) values. The SEM and Cameca 3f SIMS are both housed within the National SIMS Facility at Arizona State University.

3. Results

[6] Sample 15a is a vesiculated clast containing more silicic portions (referred to henceforth as zones) of various colors in hand sample, some appearing red, similar to oxidized dome rock (Figures 1a and 1b). The poor polish of the mafic intrusion, compared to other zones within the clast, inhibits a more detailed analysis of the texture, but it displays no orientation of vesicles, which are spherical and range from 15–300 μm in diameter. The incorporated silicic zones show a distinct lack of vesicles, while the bulk vesicularity of the entire clast is 24.4 (±1.0) %. Boundaries between the mafic intrusion and the incorporated silicics vary from smooth to crenulate and are fluidal in the case of zone A. In the outer margins of zone A, Ca X-ray maps indicate that melt from the mafic intrusion has percolated into the edges of the zone (fluidal boundaries) and was in the process of isolating some portions prior to quench (Figures 1c1e).

Figure 1.

(a) Transmitted light image of clast 15a, with identified zones outlined in blue; (b) reflected light image of zone A, showing the location of fractured quartz phenocrysts (Qz), oxidized pyroxenes (Px), oxidized Fe-oxides (Ox), pseudomorphed hornblendes (Hb), and clear, fractured plagioclase (Pl), with white squares outlining some of the regions examined using element X-ray mapping. (c–e) Ca X-ray maps of the margins of zone A, where the more mafic melt is percolating into the edges of the incorporated silicic. Brighter areas of the X-ray maps indicate a higher concentration of the selected element.

[7] Zone A, measuring 6 mm in diameter, shows evidence of heating and oxidation in all contained phenocrysts. Zone A contains several Type 2 completely pseudomorphed hornblende microphenocrysts [Murphy et al., 2000], in which the shape of the original crystal is preserved, but the hornblende has been replaced by an interlocking microlite network dominated by optically continuous pyroxene with some plagioclase and Fe-oxides (Figure 2a). The widths of eight pseudomorphed hornblendes range from 98 (±10) to 384 (±15) μm, with only two showing any remnant hornblende remaining in the core (Figure 2a). Zone A also contains several pyroxene crystals that display symplectite lamellae of a Fe-rich phase throughout the entire crystal (in microphenocrysts) or in the outermost 50 μm (in phenocrysts) (Figure 2b). Fe-oxide phenocrysts and microphenocrysts are anhedral and show evidence of extensive oxidation (Figure 2c). Plagioclase phenocrysts are clear and heavily fractured (Figure 2d), with consistent An compositions throughout averaging 53 (±4) mol% (Table 1). Zone A also contains at least 15 euhedral quartz crystals 200 μm to >1 mm in diameter (Figure 2e), with thin (<20 μm) resorption rims but no overgrowth rims of pyroxene. Like the plagioclase, the largest quartz phenocryst is heavily fractured (Figure 2e). The groundmass contains Fe-oxide microlites with few to no plagioclase microlites and no observable pyroxene (see auxiliary material).

Figure 2.

Backscattered electron images of crystals within zone A: (a) pseudomorphed hornblende crystal with small patches of remnant hornblende remaining in the core; (b) pyroxene crystal containing symplectite lamellae; (c) anhedral Fe-oxide crystal; (d) highly fractured, dominantly unzoned plagioclase phenocryst; and (e) Si X-ray map of zone A with fractured quartz phenocryst and plagioclase phenocrysts labeled; (f) Ca X-ray map of same area, showing a high concentration of small apatite crystals surrounding the large Fe-oxide phenocryst.

Table 1. Twenty Plagioclase Crystals From Zone A Were Analyzed, Near the Core or Near the Rim, Using SIMS, and the An Values of Those Spots Are Provideda
CrystalComposition (mol% An)Temperature (°C)
  • a

    In some cases, position on the crystal could not be determined (indeterminate) due to ambiguity in where the beam was striking the sample. The reported temperatures are calculated from the An content, assuming an H2O pressure of 130 MPa, and using the relationship determined in the experiments of Rutherford and Devine [2003] on a representative magma. Due to the ion beam diameter (∼25 μm), reported compositions may reflect an average of several individual zones of growth. Average values are provided at the bottom of the table, with numbers in parentheses representing 1σ. Precision in the An value is 1–2 mol%, based upon Poisson counting statistics, and resulting precision in the calculated temperature is 5–12 °C.

16257 878853 
2 54  839 
35653 849834 
45658 849858 
55157 824853 
65558 844858 
75053 819834 
85654 849839 
95044 819790 
105049 819814 
115156 824849 
12  48  810
134958 814858 
14 45  795 
1558  858  
16  52  829
17  50  819
18  47  805
19  46  800
205156 824849 
core average53 (4)  836 (19)  
rim average54 (5)  837 (23)  
average of all values53 (4)  833 (21)  

[8] Zone C is 2 mm by 1.5 mm and the groundmass contains numerous tabular plagioclase and pyroxene microlites. Zone E (2 mm x 1 mm) contains euhedral plagioclase, pyroxene, and Fe-oxide microphenocrysts. The dominantly glassy groundmass contains an uneven distribution of plagioclase and pyroxene microlites with skeletal morphologies. Zone F (1 mm x 0.5 mm) has a groundmass similar to zone A, with few plagioclase microlites difficult to distinguish from the glass, although unlike zone A, zone F contains pyroxene microlites. Zone G contains euhedral to subhedral (partially resorbed) plagioclase phenocrysts and microphenocrysts, several of which display ∼75 μm-thick calcic rims mantling sieve-textured zones. Zone G also contains phenocrysts and microphenocrysts of pyroxene. The groundmass of zone G is dominantly plagioclase which, in addition to pyroxene and Fe-oxides, forms a dense microlite network. Groundmass images of the different zones and number densities of microlite phases are provided for comparison in the auxiliary material.

4. Interpretations

[9] All silicic zones contain Fe-oxides in the groundmass, and in many cases, these microlites act as nucleation sites for pyroxene. Zone A is the only zone observed without pyroxene microlites, constraining the final depth of crystallization to 100–150 MPa [Martel and Schmidt, 2003] before incorporation into the mafic intrusion. The presence of euhedral quartz crystals constrains the temperature of zone A to <825°C assuming an H2O pressure of 130 MPa (top of magma chamber) during the bulk of crystallization [Barclay et al., 1998; Devine et al., 1998; Rutherford and Devine, 2003], which is confirmed by the limited range of plagioclase compositions (Table 1). Based upon the microlite number densities, and comparison with the decompression experiments of Martel and Schmidt [2003], we hypothesize that the various zones were incorporated into the mafic intrusion in the following order: zone A (deepest, within lower conduit or upper chamber), zone F, zone E, zone C, and zone G (highest position in conduit). We suspect that zone A was incorporated shortly after intrusion of the mafic magma into the chamber and that the other incorporated silicics (excluding zone A) were engulfed by the mafic intrusion later in time but prior to fragmentation, perhaps during rapid magma ascent through the conduit in response to dome collapse. Only zone A demonstrates sustained physical contact with the mafic intrusion.

[10] Zone A contains highly fractured plagioclase phenocrysts that display an absence of zoning that would result from ascent through the conduit and into the extruding lava dome (Figure 2f), suggesting crystallization at a single depth. This is confirmed through compositional analyses of the crystals, which show similar An contents in cores and rims (Table 1). Additionally, all hornblende crystals were pseudomorphed, even those directly abutting other crystals, indicating that breakdown did not involve decompression through the conduit, but was strictly the result of heating. This is further supported by the absence of pyroxene microlites in zone A, indicating that crystallization occurred at a final pressure greater than 100 MPa [Martel and Schmidt, 2003]. Quartz crystals have thin resorption rims and larger crystals are euhedral, suggesting that the heating of zone A was not sustained for a long enough period to significantly resorb the quartz, as the effects of heating will manifest in hornblendes (Figure 2b) and Fe-oxides (Figure 2c) prior to other mineral phases [Devine et al., 2003].

[11] The phenocryst and groundmass phases present (or absent) in zone A suggest that this represents a portion of the chamber roof or walls that cooled at depth (100–150 MPa) to a temperature sufficient for quartz growth (<825°C) [Barclay et al., 1998; Murphy et al., 2000]. Compositions of the plagioclase in zone A show no difference between core and rim values (Table 1), and assuming an H2O pressure of 130 MPa, provide an average crystallization temperature of 833 (±21)°C using the experimental trends of Rutherford and Devine [2003]. This temperature range safely brackets the value required for quartz precipitation at pressures equivalent to the upper chamber/lower conduit and also indicates that heating of zone A occurred after crystallization of the phenocryst phases, as heating during plagioclase crystallization would result in either sieve textures or more An-rich rims (as observed in zone G). Evidence of Ca-rich melt percolation into the margins of zone A (Figure 1c1e) suggests that convecting or upwelling plumes of the intruding mafic magma isolated portions of the chamber roof and walls. Murphy et al. [2000] suggested incorporation of the chamber margins as a possible explanation for the existence of quartz and low-temperature orthopyroxene crystals in erupted samples, and clast 15a provides evidence of this process. The high vesicularity and low crystallinity of the mafic intrusion (compared to the incorporated zones) suggests it represents a foam layer that rose up through the chamber from the magma interface as a low density, small-scale, discrete plume [Phillips and Woods, 2002] and incorporated zone A. Diktytaxitic texture was not observed under optical or electron microscopy, but some lath-shaped microlites (<50 μm) are visible, indicating that cooling of the mafic intrusion due to contact with the incorporated silicics was minor.

[12] Hornblende breakdown as a result of heating may be 1.3–2.7 times faster than the rate due to decompression [Browne et al., 2003]. Rutherford and Devine [2003] suggest that the rate difference is even greater for SHV, but as hornblende breakdown is also a function of melt viscosity (and the melt viscosity of zone A was high at the onset of incorporation due to its lower temperature), we assume the rate difference of Browne et al. [2003] is a reasonable approximation. We therefore used 1.3–2.7 times the experimentally determined rate of decompression-induced hornblende breakdown at SHV [Rutherford and Devine, 2003]. Taking this rate, and half the thickness of the largest observed pseudomorphed crystal, we estimate that zone A spent a minimum of 37–78 days heated by the mafic intrusion at a temperature above 860°C, which agrees with previous time estimates for heating [Devine et al., 2003]. The lower calculated value for the minimum time period of heating (37 days) roughly correlates with the onset of hybrid seismic swarms prior to the 2003 collapse [Herd et al., 2005], which are interpreted to represent the movement of fluids through solids and are regularly recorded at SHV prior to dome collapse events [Miller et al., 1998; Neuberg et al., 1998]. These hybrid events are routinely attributed to the movement of gases through the conduit margins [e.g., Miller et al., 1998], and although not the sole cause, convection or ascent of a mafic foam layer (or volatiles derived from it) through the cooler, rigid margins of the chamber or lower conduit may contribute to these seismic events.

[13] Results of this particular study show that characteristics of oxidized clasts in hand sample typically attributed to hydrothermal alteration or slow ascent into the lava dome [e.g., Devine et al., 2003] may actually be the result of heating at depth within the Soufrière Hills plumbing system, and future samples displaying such features should be examined more closely for evidence of small-scale magma mingling processes. Features within clast 15a indicate that, in addition to disaggregation of mafic inclusions during transport through the volcanic conduit [Humphreys et al., 2009, 2010] or convective self-mixing of the resident andesite [Couch et al., 2001], portions of the intruding mafic magma may also disaggregate margins of the resident magma within the chamber, contributing quartz phenocrysts, completely pseudomorphed hornblendes, and dominantly unzoned plagioclase crystals to products eventually erupted at the surface. Measurement of thermally-induced hornblende breakdown rims provides a minimum time period between intrusion of the mafic magma and eruption, helping to constrain a fundamental timescale for magma mixing processes at Soufrière Hills volcano.


[14] This research was supported by NSF EAR 0607229. Some travel costs were also supported by the Sigma Xi Grant-in-Aid of Student Research. Thanks to staff at the Montserrat Volcano Observatory for collecting and cataloguing the samples used in this study and to Marie Edmonds for providing sample information. The geochemical analyses performed at Arizona State University were possible due to the technical assistance of Richard Hervig, Klaus Franzreb, and Steve Guggino. Thanks to two anonymous reviewers for comments on an earlier version of this manuscript.