| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Petrology | Volume 44 | Number 8 | Pages 1503-1523 | 2003
© Oxford University Press 2003
Degassing at the Soufrière Hills Volcano, Montserrat, Recorded in Matrix Glass Compositions
1 DEPARTMENT OF EARTH SCIENCES, UNIVERSITY OF BRISTOL, BRISTOL BS8 1RJ, UK
2 SUERC, SCOTTISH ENTERPRISE TECHNOLOGY PARK, RANKINE AVENUE, EAST KILBRIDE G75 0QF, UK
* Corresponding author. E-mail: Steve.Sparks{at}bristol.ac.uk
RECEIVED JANUARY 11, 2002; ACCEPTED MARCH 19, 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONSAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUSIONSAPPENDIX A: OPTIMIZING THE...APPENDIX B: STEP-HEATING...REFERENCES |
|---|
Compositions of matrix glasses from the current eruption of Soufrière Hills Volcano indicate that decompression-driven crystallization results in 20–70 wt % groundmass crystallization during eruption and variable degassing. Variations in crystallinity and volatile contents (water and chlorine) of matrix glasses are attributed to variations in extrusion rates and residence times in the lava dome. Residual water contents in pumice clasts (0·2–0·6 wt %) indicate minimum pressures of 1·1–3·7 MPa in 1997 Vulcanian explosions. Residual water contents of 1·6 wt % in a ballistic block ejected in sub-Plinian explosive activity on 17 September 1996 imply larger pressure drops (
20 MPa). Variable residual water contents in dome samples are consistent with pressure variations of up to 9 MPa in the lava dome interior. Large variation in chlorine contents between lava blocks compared with explosion pumice clasts indicates that shallow-level processes dominate degassing. Low melt chlorine contents of dome samples cannot be explained by open- or closed-system degassing, especially when crystallization is taken into account. Instead, heterogeneous chlorine leaching by circulation of groundwater vapour in the dome is proposed. Variable and elevated matrix glass 
D values can also be attributed to interaction with isotopically heavy surface waters. HCl emission during the current eruption can be accounted for by Cl loss from the melt, consistent with the melt being undersaturated in chlorine. KEY WORDS: andesite; degassing; chlorine; lava dome
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUSIONSAPPENDIX A: OPTIMIZING THE...APPENDIX B: STEP-HEATING...REFERENCES |
|---|
The exsolution and escape of volcanic gas from magma are major controls on volcanic activity. Understanding degassing is an important goal in volcanology, because the fluxes and inventories of volcanic gases can be monitored by remote-sensing techniques and analytical methods. These data provide information to assist forecasting activity and mitigating the effects of volcanism. The current eruption of Soufrière Hills Volcano, Montserrat, has exhibited a wide variety of eruptive styles, ranging from sub-Plinian and Vulcanian explosive eruptions to extrusion of a viscous lava dome at rates from <0·5 m3/s to
9 m3/s (Robertson et al., 2000). Transitions in eruptive style have been attributed principally to ascent-driven degassing and crystallization, because of their enormous effect on magma rheology (Melnik & Sparks, 1999; Voight et al., 1999; Sparks et al., 2000). During magma ascent, decompression-driven volatile exsolution is accompanied by crystallization, because volatile loss causes a rise in the liquidus temperature of the magma (Blundy & Cashman, 2001). This paper presents a study of the volatiles (water and chlorine) and hydrogen isotope compositions of glasses from a suite of samples from the current eruption of Soufrière Hills Volcano to investigate degassing processes. The study complements and develops the study of degassing at the Soufrière Hills Volcano by Edmonds et al. (2001). The Soufrière Hills Volcano is well suited to such a study because of the availability of both a well-documented sample suite, which covers a variety of eruptive activities, and a wealth of complementary information on the volcanic system. Experiments simulating chlorine behaviour in magma of the Soufrière Hills Volcano (Signorelli & Carroll, 2001) allow comparisons to be made with samples from the current eruption. Degassing is assessed through analyses of the major elements and volatiles (chlorine and water) in matrix glasses. Comparing post-eruptive volatile concentrations in matrix glasses with pre-eruptive concentrations inferred from melt inclusions allows assessment of degassing during magma ascent. Modelling of water degassing is facilitated by analysis of the hydrogen isotope composition of matrix glass water. Hydrogen isotope compositional variations can be influenced by whether a degassing system is open or closed, and can be influenced by kinetic effects and interactions with external waters. Analysis of major elements and volatiles in the same matrix glass samples allows investigation of the relationship between crystallization and degassing processes. This information can help interpret investigations of gas fluxes using remote-sensing techniques. In particular, changes of Cl/S measured by Fourier transform IR spectroscopy (FTIR) (Edmonds et al., 2001) can be interpreted in the context of Cl degassing processes.
|
SAMPLING AND ANALYTICAL METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUSIONSAPPENDIX A: OPTIMIZING THE...APPENDIX B: STEP-HEATING...REFERENCES |
|---|
Samples were selected to cover a broad range of eruptive conditions, including explosion-derived pumice and ballistic clasts, and lava blocks from periods of varying dome extrusion rates (Table 1). For safety reasons it was impossible to sample the lava dome directly. Hence lava blocks were collected from block-and-ash-flow deposits formed by episodic dome collapse. Samples were selected from block-and-ash flows thought likely to have involved collapse of the active flow lobe of the lava dome and not to have remobilized large proportions of material from older parts of the dome.
|
|
Matrix glasses were analysed for major elements and chlorine at Bristol University on the JEOL JXA-8600 electron microprobe with four wavelength-dispersive spectrometers. Analytical techniques were optimized to minimize loss of sodium whilst at the same time ensuring good precision (Appendix A). Glass areas were selected so as to allow as large an analysis area as possible, often
5 µm x 4 µm, and always greater than 4 µm x 3 µm. Each glass area was analysed twice with a rastered beam and an accelerating voltage of 15 kV. First the glass was analysed for major elements at the low beam current of 2 nA. The background was measured only on the initial analysis, which was not included in the mean. Sodium stability tests indicate that sodium loss is thus minimized to an estimated 5–10% relative (Appendix A). Subsequent analysis of the same glass area for chlorine, present only in small quantities, was carried out at 10 nA with a peak count time of 120 s. Five to 10 glass analyses were made at each beam current to assess sample heterogeneity. For homogeneous samples multiple analyses also reduce uncertainties for the mean glass composition. Separate standardization routines were carried out at each beam current, using a beam diameter of 5 µm and the following standards: albite (Na), olivine (Mg), wollastonite (Si, Ca), synthetic Al2O3 (Al), adularia (K), synthetic SrTiO3 (Ti), hematite (Fe), synthetic MnO (Mn) and NaCl salt (Cl). To check the standardization, secondary standards KK1 albite and KN18 glass were analysed at 2 and 10 nA, respectively, with a 15 µm beam diameter. The predicted precision for a single microprobe analysis of chlorine, based on counting statistics, ranges from 0·02 to 0·04 wt % for glasses containing 0·33 and 0·05 wt % Cl, respectively. The measured standard deviation for chlorine in KN18 containing 0·32 wt % is 0·01 wt % (for 26 analyses during the course of this work; predicted standard error of the mean is 0·008). For major elements the single analysis precision based on counting statistics is better than 7% relative, except for Fe (20%) and minor elements (Ti, Mg, Mn). It is difficult to obtain a direct measure of water in matrix glasses. The glass areas are too small for FTIR analysis, and assessment of water by electron microprobe difference methods is not reliable for these tiny glass patches as a result of sodium loss. Instead, a selection of samples was analysed for water and hydrogen isotopes by bulk extraction at SUERC (Scottish Universities Environmental Research Centre). Anhydrous samples of the bulk matrix were prepared by coarse crushing and hand-picking to remove amphibole crystals. In some cases amphiboles were removed using heavy liquid (tetrabromoethane) separation and then hand-picking (Table 2). Coarse grain size fractions, generally 1–1·4 mm and occasionally as fine as 250–500 µm, were used to minimize surface water contribution (Newman et al., 1988), facilitate removal of amphiboles and minimize grain size fractionation of glass.
Step-heating experiments were performed to assess water release as a function of temperature, to determine a suitable heating schedule for water extraction (Appendix B). The resultant extraction schedule involved overnight degassing at 200°C to remove surface water, before heating to over 1400°C for a period of around 2–6 h, until no further gas was evolved. Released water was converted to H2 by reaction with hot uranium and the yield measured manometrically (Fallick et al., 1993). Extractions were generally performed at least in duplicate. Hydrogen isotope results are expressed as D values in parts per thousand (
) relative to V-SMOW (Vienna Standard Mean Ocean Water). The water content of the matrix glass was estimated from the measured water content (relative error 10%), and applying a correction to account for the anhydrous crystals in the samples analysed. Estimates of matrix glass content based on electron microprobe analysis (EMPA; see below) were used where available. Where glass could not be analysed by electron microprobe, glass contents were estimated by comparison with other samples (12 wt % for dome samples, 32 wt % for pumice clasts), and these estimates were then used to calculate approximate glass water contents.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUSIONSAPPENDIX A: OPTIMIZING THE...APPENDIX B: STEP-HEATING...REFERENCES |
|---|
Matrix glass compositions
The electron microprobe major-element analyses show that the matrix glasses are rhyolitic and span the compositional range from 76·5 to 79·3 SiO2 (normalized to 100% anhydrous; Table 1). Repeat analyses on each sample gave major-element 1
variation comparable with counting statistics precision, indicating major element homogeneity within the samples studied. Here K2O is used as an indicator of the evolution of the melt phase in the magma as a result of crystallization, as K is nearly incompatible in the crystallizing mineral assemblage in the Soufrière Hills magma. K2O constitutes only around 0·1–0·4 wt % of plagioclase and 0·18 wt % of amphibole crystals (Murphy et al., 2000). Amphibole does not crystallize during magma ascent, as it is unstable at low pressures (Barclay et al., 1998). The matrix glass compositions vary from 2·0 to 5·9 wt % K2O, indicating a wide range of sample crystallinities. The matrix glasses are richer in K2O than most melt inclusions in phenocrysts (Barclay et al., 1998; Devine et al., 1998a; Edmonds et al., 2001), a groundmass composition (determined by rastered EMPA; Barclay et al., 1998), and the average whole-rock composition [determined by X-ray fluorescence (XRF); Murphy et al., 2000, G. Zellmer, personal communication, 2002], indicating that the matrix glasses have experienced a greater degree of crystallization than these materials (Table 1).
Matrix glass K2O contents were used to estimate the glass contents of samples by mass balance of K2O. The model involves a mass balance of K2O using observed proportions and K contents of the phenocrysts (principally plagioclase and amphibole) and bulk-rock K2O contents (Murphy et al., 2000). Estimated glass proportions vary from 34 to 10 wt % (Table 1). These glass estimates have small uncertainties (1 of ±0·5–4·6 wt % absolute) associated with errors in XRF and electron microprobe K2O measurements, and assumptions involved in the mass balance calculation (Table 1). The glass contents estimated are consistent with estimates by image analysis and results reported in Edmonds et al. (2001). The estimated glass content of MVO47 is 25 wt % by mass balance and 22 vol. % by image analysis (assuming an average density of 2400 kg/m3 for glass and 2700 kg/m3 for whole rock, this corresponds to 20 wt % glass). Many samples proved unsuitable for reliable EMPA of matrix glass because of the lack of sufficiently large matrix glass areas and are excluded from the dataset. Glass estimates for some of these samples by image analysis indicated glass proportions as low as 5 vol. %.
There is much greater variation in glass composition and groundmass crystallization in lava dome samples than in pumice clasts. The variation between two lava blocks from the same block-and-ash-flow deposit is, in fact, as great as the variation between lava blocks from different block-and-ash-flow deposits. This variability can be attributed to two factors: first, dome collapses sample regions in the dome with different degassing and cooling histories, and hence different ultimate crystallinities; second, block-and-ash flows can erode blocks deposited from previous flows (Cole et al., 2002).
An exception to major-element glass homogeneity occurs in samples where there appear to be two ‘glasses’, as illustrated by analyses MVO58K and MVO58Na (Table 1). Similar observations have been made for Mount St. Helens samples (Cashman, 1992). Blundy & Cashman (2001) recently examined samples with apparently heterogeneous glasses by transmission electron microscopy (TEM). They concluded that the heterogeneities constituted an intimate mixture of a K-rich true glass, and very finely crystalline intergrowths of feldspar and quartz. The Montserrat results provide additional evidence to support this interpretation. The K-rich zones contain chlorine, whereas the Na-rich zones contain no chlorine, consistent with the former zones comprising glass, and the latter zones comprising volatile-poor microcrystalline regions. Hence, where this texture was observed, only the K-rich, true glass was analysed.
Matrix glass chlorine contents
Matrix glass chlorine contents are highly variable, ranging from 0·31 to below the detection limit (Table 1, Fig. 1). Chlorine values are high in pumice clasts (0·31–0·27 wt %), and variable in ballistic and lava dome samples (0·28–0·03 wt %). The standard deviation of repeat chlorine analyses is generally comparable with counting statistics precision, indicating chlorine homogeneity within these samples. However, the 1 variation for MVO47 and MVO52 (0·08 and 0·10 wt %, respectively) significantly exceeds the counting statistics precision (0·04 wt %), indicating chlorine heterogeneity. Edmonds et al. (2001) also reported variable matrix glass chlorine contents in a similar range of 0·32 wt % to below the detection limit of 0·04 wt %. Matrix glass chlorine concentrations are mostly lower than values reported for plagioclase melt inclusions of up to 0·44 wt % (Devine et al., 1998a; Edmonds et al., 2001). Devine et al. (1998a) also measured chlorine contents of 0·26 wt %, in melt inclusions in quartz.
|
Both matrix glass values and melt inclusion chlorine contents are below the solubility limit of chlorine as determined experimentally (Signorelli & Carroll, 2001) for rhyolitic melt compositions similar to those found at Soufrière Hills Volcano in the presence of vapour and brine. Signorelli & Carroll (2001) used average groundmass composition as the starting material. Those workers found solubilities of 0·48–0·68 wt % Cl at pressures in the range 250–25 MPa, respectively. At pressures between 100 and 200 MPa, where the melt composition is relatively constant, chlorine solubility is inversely correlated with pressure (Signorelli & Carroll, 2001). Metrich & Rutherford (1992) also found a negative correlation between pressure and Cl solubility in experiments on hydrous peralkaline rhyolites. In contrast, Webster et al. (1999) found that Cl solubility decreases with decreasing pressure from 2 kbar to 1 atm with water-poor felsic, intemediate and mafic melts. These experiments involved coexistence of silicate melts with a brine but without a vapour phase, and are thus less relevant to the Montserrat situation.
Changing melt composition, in particular increasing silica activity or increasing [(Na + K)/Al]molar (from peraluminous to metaluminous compositions), can result in a moderate decrease in chlorine solubility (Metrich & Rutherford, 1992; Carroll & Webster, 1994; Webster & De Vivo, 2002). Experimental glasses of Signorelli & Carroll (2001) range in composition from 73·0 to 77·7 wt % SiO2 with [(Na + K)/Al]molar of 0·67–0·81, whereas natural samples include more evolved compositions, ranging from 76·5 to 79·3 wt % SiO2 (normalized to 100% anhydrous) with [(Na + K)/Al]molar of 0·67–0·99. Some of the dome sample matrix glasses therefore have slightly higher SiO2 and [(Na + K)/Al)]molar than the 25 MPa experimental glasses of Signorelli & Carroll (2001), which might result in slightly lower Cl solubility. Thus, for pressures between 25 and 100 MPa for the Soufrière Hills andesite, Cl solubility decreases with increasing pressure. The increase in SiO2 and [(Na + K)/Al)]molar as a result of crystallization will have a small counteracting effect. These effects are, however, minor, such that Cl solubility remains approximately constant during magma ascent with concurrent crystallization.
Matrix glass water and hydrogen isotopes
Bulk extraction from amphibole-free separates gave low water contents of 0·02–0·4 wt % and D values ranging from -40 to -100 (Table 2). Estimated glass proportions were used to calculate approximate glass water contents from bulk water values, as described above. Removal of amphibole and other heavy mineral phases, and fractionation of glass out of the coarse-grained separate analysed can cause the proportion of glass in the crushed sample analysed to be unequal to that in the whole-rock. However, K2O contents of glass residues after fusion (for hand-picked samples) were similar to whole-rock K2O by XRF (averages of 0·76 wt % and 0·77 wt %, respectively), indicating that effects caused by removal are minor. The uncertainties in K2O analyses by electron microprobe and XRF are larger than the error in glass content estimates introduced by these procedures.
The results confirm that matrix glass water contents are low: <1 wt % (Table 2, Fig. 2). One exception is MVO57, a glassy ballistic from the 17 September 1996 sub-Plinian explosive eruption (Robertson et al., 1998) for which six individual analyses gave an average matrix glass H2O content of 1·82 ± 0·09 wt %. D variation of up to 20 for splits of the same sample is significantly greater than expected from analytical precision (±3 for mineral separates) (Table 2). Variation of D is unlikely to represent incomplete water extraction for two reasons. First, there is no systematic relationship between D values and water content for repeat analyses of the same sample, which could have been related to processes of water release. Second, after water extraction, fused glasses had no detectable retained water by FTIR, i.e. <0·003 wt % H2O, which corresponds generally to less than a few percent of the total water extracted from the sample. The very low bulk water content of these samples may increase analytical uncertainties, but the magnitude of the variation suggests that there is D heterogeneity within the samples.
|
Ion microprobe studies (Harford & Sparks, 2001) have shown marked
D heterogeneity within and between amphibole crystals in early erupted samples (before April 1996). In contrast, ion microprobe analysis showed variations within the analytical precision of this technique (±8) in amphiboles erupted after spring 1996. Repeat analyses by bulk extraction of fresh amphibole separates (MVO121, 17 September 1996 explosion pumice) gave a mean D of -52 ± 10 (ranging between -36 and -64), comparable with values obtained for other volcanic amphiboles in subduction zone settings (Hildreth & Drake, 1992; Taran et al., 1997; Kusakabe et al., 1999). However, D variation of ±10 (1) is significantly greater than the analytical precision of ±3 (Harford & Sparks, 2001). |
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUSIONSAPPENDIX A: OPTIMIZING THE...APPENDIX B: STEP-HEATING...REFERENCES |
|---|
Eruptive conditions
The difference in composition and crystallinity observed between samples is interpreted as due to differences in eruptive conditions and associated time spent at shallow levels. The most K2O-poor, least crystalline samples are pumice clasts (with estimated glass contents of 34–31 wt %), erupted during explosions. During Vulcanian explosions, material from the uppermost several hundred metres of the conduit was evacuated on a timescale of less than a few minutes (Druitt et al., 2002), giving insufficient time for crystallization. During the periods of Vulcanian explosions, magma extrusion rate was estimated at
9 m3/s, comparable with the highest lava extrusion rates (Sparks et al., 1998; Druitt et al., 2002). For the pumice clasts, ascent from the magma reservoir at an estimated depth of >5 km (Barclay et al., 1998) to the surface, assuming a conduit area of 700 m2, as discussed by Melnik & Sparks (1999) and Robertson et al. (1998), would have taken only 4 days. These estimates are consistent with ascent speeds estimated by studies of amphibole breakdown reactions (Devine et al., 1998b). During the sub-Plinian explosion of 17 September 1996, when it is inferred that the majority of the conduit was evacuated (Robertson et al., 1998), pumice clasts may have spent as little as an hour in transit to the surface from depths of up to 4 km. In contrast, samples of the lava dome experienced a relatively slow transit to the surface. These samples are more crystalline, with estimated glass contents of 31–5 wt %. Magma ascent from the chamber to the surface would have taken between 5 and 80 days for typical extrusion rates of 8–0·5 m3/s (Sparks et al., 1998), assuming a conduit area of 700 m2. In addition, these samples would have spent an unknown period in the lava dome, probably ranging from a few days to several months. The amount of crystallization thus reflects both ascent time and residence time in the dome before collapse, explaining the lack of correlation between glass K2O content (a proxy for groundmass crystallinity) and magma extrusion rate (Fig. 3). A further modifying control on crystallization is each sample's individual thermal and degassing history, which may vary significantly dependent on position within the shear lobes of the lava dome (Watts et al., 2002). These factors account for the large variation in matrix glass composition and groundmass crystallization in lava dome samples compared with pumice clasts.
|
Chlorine degassing
In the following discussion glass K2O content is used as a proxy for crystallinity. The chlorine contents of matrix glasses do not show a simple relationship with either magma extrusion rate or K2O content. Pumice clasts erupted during periods of rapid magma extrusion have low K2O glass contents and crystallinities, and high chlorine contents, similar to those of most plagioclase melt inclusions. Data on melt inclusions in plagioclase (Devine et al., 1998a; Edmonds et al., 2001) extend the data to higher chlorine contents and lower glass K2O contents. These data are consistent with the interpretation that most of the melt inclusions were trapped before magma ascent. There is an approximate inverse correlation between K2O and Cl for these matrix glass samples (Fig. 1), indicating that Cl degassing accompanies crystallization. However, at higher crystallinities (high K2O), ballistics and lava dome samples have very variable chlorine contents, which do not relate to K2O. Both high and very low Cl contents are observed in highly crystalline samples (Fig. 1), indicating that the relationship between late-stage crystallization and Cl degassing is complex. Lack of correlation between matrix glass Cl content and magma extrusion rate for lava dome samples (Fig. 3) may in part be attributed to variable extrusion rates and residence times in the dome, resulting in variable degassing histories as well as variable crystallization histories as discussed above.
The chlorine contents of both matrix glasses and melt inclusions are all well below the solubility limit of chlorine in Soufrière Hills melts of 0·5–0·7 wt % in the pressure range 25–250 MPa (Signorelli & Carroll, 2001). Melt inclusions in plagioclase (Devine et al., 1998a; Edmonds et al., 2001) are assumed to represent melt compositions in the magma reservoir, and hence reflect pre-eruption volatile contents. In the absence of degassing, increasing crystallization of chlorine-incompatible groundmass phases during magma ascent should be accompanied by increasing chlorine in the melt. However, the chlorine contents of the matrix glasses are lower than those of the plagioclase melt inclusions. These observations, together with the tendency for Cl to decrease with higher crystallinity (higher glass K2O), indicate chlorine degassing on ascent. Edmonds et al. (2001) observed similar evidence for chlorine degassing during magma ascent for their studies of matrix and melt inclusion glass compositions.
Chlorine degassing can occur when the combined solubility limit of the volatile species present is reached. The lack of a simple relationship between Cl and H2O in matrix glasses (Fig. 4) suggests that a single process does not control chlorine degassing at Soufrière Hills Volcano. Cl–H2O degassing can be modelled in terms of equilibrium partitioning of chlorine (as a trace element) between the melt phase and an accompanying aqueous fluid phase or vapour plus brine (Villemant & Boudon, 1999) (Fig. 4). It is assumed that no significant fluid phase exists in the magma chamber and that a water-dominated fluid phase forms during ascent by decompression-induced water saturation. Starting melt volatile contents of 4·3 wt % H2O and 0·34 wt % Cl are assumed based on melt inclusions in plagioclase measured by FTIR (Barclay et al., 1998) and EMPA (Devine et al., 1998a). The chlorine partition coefficient between the aqueous phase and the melt, DCl, is defined as
 |
and 
denote chlorine concentration in the aqueous fluid phase (vapour in the chlorine-undersaturated case) and melt, respectively. There are no direct measures of DCl for the Soufrière Hills melt compositions. The 25–150 MPa experiments of Signorelli & Carroll (2001) were conducted in the subcritical region of the system NaCl–H2O, where melts are chlorine saturated and two immiscible fluids (vapour and brine) are present (Shinohara et al., 1989). Therefore only apparent DCl values can be determined. These values (21 at 25 MPa to 34 at 150 MPa) can be considered as a maximum for true DCl for the chlorine-undersaturated case, if there is only a vapour phase present. Villemant & Boudon (1999) modelled data for Mont Pelée, Martinique, largely using a DCl value of 10. DCl exhibits a strong positive pressure dependence (Shinohara et al., 1989). This is supported by experiments on rhyolitic glass at 850°C, which indicate that DCl decreases from 8 at 150 MPa to 2 at 20 MPa (Gardner et al., 1998; J. E. Gardner, personal communication, 2000). These results were used to estimate DCl at varying pressure for the models presented (Fig. 4b). Degassing models are also presented with constant DCl to illustrate key points (Fig. 4a).
|
Closed-system degassing is derived from a simple mass balance derived directly from the definition of DCl:
 | (1) |
 |
 | (2) |
= DCl - 1 and f as above.
The fractionation effects of both open- and closed-system processes become more marked as f decreases. In the early stages, when f is large (<0·75), open-system trends resemble closed-system equilibrium trends (e.g. Wilson, 1989). Because the melt contains only 4·3 wt % water, the minimum f attained through water degassing is 0·96, and evolution never extends beyond the early stages. Therefore chlorine fractionation during water loss is not markedly greater in an open system than in a closed system (Fig. 4).
The samples with high observed chlorine melt contents can be reasonably well accounted for using realistic values for DCl of Gardner et al. (1998) and J. E. Gardner (personal communication, 2000) and an open- or closed-system degassing model (Fig. 4b). However, these models cannot generate the low to very low melt chlorine contents of some dome samples, without requiring very high DCl values of 50 (Fig. 4a).
Villemant & Boudon (1999) proposed a model of simultaneous crystallization and open-system degassing, with DCl of 10–30, to account for low chlorine contents of dense samples from Mt Pelée, Martinique. The degassing models presented above require modification to account for concurrent crystallization because chlorine is incompatible in the crystallizing phases. Two important resultant effects must be taken in account. First, the effective bulk partition coefficient for the crystalline assemblage removed from the melt, 
, is less than the partition coefficient for the aqueous phase alone, 
(referred to as DCl above), as follows:
 | (3a) |
 | (3b) |
 | (4) |
for DCl into equations (1) and (2) gives expressions for crystallization accompanying closed- and open-system degassing, respectively. Assuming an initial groundmass glass content of 43% (from mass balance using average groundmass composition, Table 1), shallow-level groundmass crystallization varies between 21 and 77 wt % (for end glass contents of 34 and 10 wt %, respectively). Around 3·5–4 wt % of water is degassed from the melt on ascent, resulting in bulk k values varying from about 5 to 20 (Fig. 4b). The two effects of crystallization counteract one another. Higher f for a given degree of H2O degassing means that higher degrees of fractionation are reached. However, low means that less Cl is partitioned out of the melt. For high k, becomes <1 and Cl partitions preferentially into the melt, resulting in Cl enrichment with water degassing and crystallization (Fig. 4c). In all cases, melt Cl is elevated over that in the open-system model with no degassing (i.e. k = 0) (Fig. 4b and c). Therefore simultaneous degassing and crystallization cannot account for the low melt chlorine concentrations at Soufrière Hills Volcano, unless the value of DCl exceeds 50, as also recognized by Edmonds et al. (2001). Kinetics cannot account for enhanced chlorine loss either, because diffusion of chlorine is several times slower than that of water in granitic melts (Bai & van Groos, 1994). In explosive degassing, where diffusion controls may become dominant (Hort & Gardner, 2000) chlorine may be left behind in the melt relative to water. Thus, if water escaped rapidly in an open system, volatile differentiation would involve melt chlorine enrichment rather than depletion.
There are five possible explanations for the low melt chlorine concentrations observed. First, a marked reduction in chlorine solubility at low pressures could result in low melt chlorine. Changing melt composition, in particular increasing silica activity or increasing [(Na + K)/Al]molar (from peraluminous to metaluminous compositions) or decreasing Mg, Ca and Fe, can result in a decrease in chlorine solubility (Metrich & Rutherford, 1992; Carroll & Webster, 1994). However, these effects could only lower chlorine solubility moderately, and in addition the inverse dependence of chlorine solubility on pressure deduced from experiments in the presence of vapour would counteract this. Second, at low pressures, where no experimental constraints are available, an elevated DCl (of order 50 or more) could result in late-stage chlorine depletion (Fig. 4b). However, DCl depends positively on pressure (Shinohara et al., 1989; Gardner et al., 1998) so very high values of D seem unlikely. Further, if this were the correct interpretation then a problem would emerge of how to explain the high chlorine contents of other glasses with high K2O. Third, there is the possibility that a Cl-rich brine can separate at low pressures, resulting in very high distribution coefficients between the bulk fluid and silicate melt. J. E. Gardner (personal communication, 2000) has found values of 100–300 for DCl for such systems. Unfortunately, there are no experiments under pertinent conditions to test this hypothesis. Fourth, late-stage formation of a chlorine-rich mineral phase, for example apatite, could account for elevated bulk partitioning out of the melt. However, considering apatite comprises only << 1 wt % of the lava, it would require both an unfeasibly high DCl for apatite of >150 and an unachievable >7 wt % chlorine concentration in the apatite.
A final possibility is that enhanced water circulation results in additional chlorine degassing from the melt without requiring significantly elevated DCl. Assuming DCl = 2, a mass of water equivalent to 90% of the melt mass would be sufficient to reduce the melt Cl from 0·3 to 0·05 wt %. In fact, many samples with low melt Cl are highly microcrystalline with effective bulk k [equation (3)] of up to 20 as discussed above. Degassing with k of 20 would result in 0·7 wt % Cl at 0·5 wt % H2O with open-system degassing (Fig. 4c). We have assessed the effects of flushing of meteoric water by a finite-difference calculation in which a small amount of water is equilibrated with a unit mass of melt and then outgassed and replaced by another small amount of water. This is repeated until the chlorine content has reduced to the observed values of the most chlorine-poor glasses. Assuming that k = 20 and DCl = 2, reduction to 0·05 wt % Cl requires flushing by 130% of the melt mass of water. This is equivalent to 13% of the lava mass for a sample with 10% final melt, or 300 kg water/m3 magma. Water rising from magma at a lower level in the conduit is unlikely to provide this flux, as such water should already be chlorine charged. It is also unlikely that such a large water flux could be accounted for by an excess vapour phase sourced from the magma chamber. However, groundwater and hydrothermal circulation could account for such volumes. Assuming an average rainfall of 2 m/year and that magma extrusion over the eruption has averaged 75 x 106 m3/year, an area of 10 km2 would receive sufficient rainwater to leach chlorine to this extent from all the magma erupted.
Samples have variable melt Cl, therefore we propose that leaching of chlorine by escaping steam is heterogeneous. Vapour passes preferentially through permeable areas of the dome, such as cracks, as evidenced by localized areas of steam emission from the dome. The continuation of steam venting, at times very vigorous, during the period of no dome extrusion (Norton et al., 2002) is consistent with a model of circulation of surface waters through the dome. Another possibility is that interaction of water with the glass and removal of Cl occur in the pyroclastic flow deposits by infiltration of rain and circulation of surface waters. However, our samples were collected within a few days of emplacement, limiting the time available for interaction.
This model has implications for the cooling and crystallization history of the lava dome. Hydrothermal circulation would lead to localized cooling, although this has to be limited to avoid hydration of glass. Some cooling-driven crystallization may thus be important in some parts of the dome. Below, our discussion of the hydrogen isotope data provides independent support for involvement of external water.
Water degassing
Matrix glass water contents are low (<1 wt %) compared with water contents of plagioclase melt inclusions (4·3 ± 0·5 wt % by FTIR, Barclay et al., 1998), which are inferred to represent pre-eruption melt concentrations. This indicates that water degasses efficiently during eruption at Soufrière Hills Volcano.
Matrix glass water contents can be interpreted as minimum pressures at which the melt is water saturated. The solubility model of Moore et al. (1998) is used for the Soufrière Hills melt compositions at an estimated magma temperature of 860°C (Murphy et al., 2000). The most water-rich matrix glass is that of the dense glassy ballistic sample MVO57, erupted during the 17 September 1996 sub-Plinian explosive eruption, and interpreted as part of the conduit wall broken off from a relatively deep level (Robertson et al., 1998). The glass water content of 1·6 wt % H2O corresponds to a minimum pressure of 20 MPa. Assuming conduit pressure equals magmastatic pressure, and an average magma density of 2400 kg/m3, this pressure equates to a depth of 830 m. However, several lines of evidence suggest that overpressures of several MPa develop in the conduit at Soufrière Hills Volcano, and that this therefore represents a maximum depth estimate. Ground deformation during 1996 has been modelled in terms of a pressure source at 700 m depth with an overpressure of 10 MPa and total pressure (overpressure plus magmastatic) of 25 MPa (Shepherd et al., 1998). Tilt-meter cycles have been interpreted in terms of a pressure source at 400 m depth (Voight et al., 1999). Melnik & Sparks (1999) developed dynamic flow models of lava dome extrusion, which suggest that overpressures of up to 4–8 MPa develop at depths of a few hundred metres in the conduit. A pressure of 20 MPa from the water content of a small ballistic can be compared with the peak explosion pressure of 27·5 MPa estimated for this eruption from the largest ballistics (Robertson et al., 1998).
Minimum pressures of 1·1–3·7 MPa are estimated for pumice clasts based on matrix glass water contents of 0·3–0·6 wt % (Table 2). If the pumices were quenched on fragmentation, this represents lower fragmentation pressures compared with the rapid decompression experiments of Aldibirov & Dingwell (1996), in which a pressure drop of 12 MPa resulted in spontaneous fragmentation of Mount St. Helens dacite. If pumice expansion continues during ejection then 1·1–3·7 MPa represents minimum fragmentation pressures (Gardner et al., 1996). Water contents of glasses from lava dome samples of 0·14–1·0 wt % can be related to equilibrium minimum pressures of 0·3–9 MPa. This pressure range is consistent with that which might be expected within a thick lava dome with deep regions of overpressure, which is suddenly sampled by a large dome collapse. The Soufrière Hills dome has reached heights of up to 300 m during its growth (Watts et al., 2002). Overburden pressures of up to 7 MPa can be expected at the base of the dome. Overpressures above magmastatic pressure can be generated during extrusion (Melnik & Sparks, 1999), so that pressure estimates using residual water contents of glasses cannot be simply used to infer depths.
Hydrogen isotope compositions of matrix glasses are controlled by the original isotopic composition, whether degassing occurs as an open or a closed system, whether kinetic effects are important, and whether isotopic exchange occurs with external waters. Taylor et al. (1983) first related hydrogen isotope compositions of lava to degassing processes, modelling degassing in terms of open- and closed-system end-members. Standard expressions (e.g. Holloway & Blank, 1994) equivalent to those used to model chlorine degassing above are used to calculate the effect of degassing on the end isotopic composition of melt. Using mass balance considerations, closed-system degassing is described using the approximation
 | (5) |
Df and Di are the final and initial hydrogen isotopic compositions in the melt, F is the mole fraction (normalized) of the volatile species remaining dissolved in the melt (unity at the beginning, zero at the end), and 
v–m is the bulk vapour–melt fractionation factor. Open-system degassing (Rayleigh fractionation), is modelled by an expression derived from equation (2) above:
 | (6) |
v–m, varies with water content (Newman et al., 1988; Dobson et al., 1989). The bulk fractionation factor between water vapour and melt can be expressed as (Taylor, 1986)
 |
v-OHm was taken as 1·040, from the experiments of Dobson et al. (1989) at 750–850°C. v-H2Om was calculated using this value for v-OHm, and a value of bulk v–m of 1·0239 [Taylor (1986) at 950°C with 3·1 wt % H2O]. Speciation of water in the melt for this bulk v–m value was estimated from Nowak & Behrens (1995) for 4·14 wt % H2O at 850°C, adjusted to 3·1 wt % H2O based on the speciation model in glass of Silver et al. (1990). v-H2Om was thus estimated as 0·957. For degassing models, v–m was estimated for each water content, using the same procedure, and varies from 1·020 to 1·037 at water content from 4 to 0·1 wt %, respectively. The parent magma chamber melt before degassing is assumed to have a water content of 4·3 wt % based on the water content of melt inclusions in plagioclase (Barclay et al., 1998). Using hbl–v (hornblende–vapour) of 0·983 at 750°C from Suzuoki & Epstein (1976), and v–m at melt water content of 4 wt % estimated above as 1·020, hbl–m is estimated to be 1·003. Assuming hydrogen isotope equilibrium between the melt and hornblende phenocrysts, the starting melt is estimated to have a D composition of -55 ± 10.
The Soufrière Hills matrix glasses do not show a systematic H2O–D relationship (Fig. 2). There is a broad similarity between these results and those for similar eruptions of crystal-rich andesites and dacites. Matrix glass D results from the 1991 eruption of Unzen Volcano, Japan, also show a non-systematic H2O–D relationship, interpreted as a result of late-stage kinetic degassing (Kusakabe et al., 1999). Varying combinations of closed- and open-system and kinetic degassing, linked to repose period before eruption, were invoked to account for complex H2O–D relationships of Mount St. Helens matrix glasses (Anderson & Fink, 1989). The Soufrière Hills data are also consistent with degassing by varying combinations of open-system, closed-system and kinetic degassing. This result is not surprising, as a very wide range of isotopic values can be generated by these processes; open- and closed-system degassing result in D depletion, whereas kinetic degassing results in D enrichment through faster diffusion of hydrogen- than deuterium-bearing water. However, most of the samples lie above the model D–H2O curves, even considering the uncertainty in initial D, suggesting that a D-enrichment process occurs. Kinetic degassing is one possible explanation. However, interaction with surface waters could also result in D enrichment. Meteoric waters in the region average at around -10 (International Atomic Energy Agency, 1992) and seawater has a D value close to 0. Chiodini et al. (1996) measured D of fumarolic condensates and spring waters on Montserrat, and found values varying between -13 and +22. Circulation of ground water through the dome and upper conduit has already been proposed to account for enhanced chlorine degassing of matrix glasses. It is therefore proposed that heterogeneous shallow-level interaction between matrix glass and groundwater vapour accounts for the observed D enrichment, non-systematic D–H2O relationship, and D heterogeneity within samples. This is consistent with a tendency for dome lava samples to be more enriched in D than pumice samples (Fig. 2). The large variation (±10) in bulk D analyses of fresh amphiboles may also be attributed to such processes.
Estimation of chlorine degassed from glass data
The matrix glass chlorine results, glass contents, and estimated magma ascent rates allow quantification of chlorine degassed at Soufrière Hills Volcano (Table 1). We do not include any chlorine-bearing fluid phase in the magma chamber. If there was a free vapour phase in the chamber into which chlorine was partitioned then this would also contribute to the chlorine emissions at the surface. Therefore our estimates give only the contribution of chlorine dissolved in the magma chamber to the total gas emissions. We also assume that there is no significant chlorine contribution from decompression amphibole breakdown (Devine et al., 1998b). The proportion of initial melt chlorine degassed at the Soufrière Hills Volcano can be calculated from the difference in the pre-eruption and post-eruption chlorine contents of the magma, based on estimates of melt proportions and melt chlorine concentrations. Pre-eruption chlorine content is based on an assumption of 43 wt % melt (Table 1) with chlorine concentration of 0·34 wt % (Devine et al., 1998a). An estimated 31–97% of chlorine in the melt is degassed during magma ascent, with the explosion pumice blocks degassing the least, estimated at <43%. The dense lava dome blocks degas 51–97% of the original chlorine. Therefore, even though the final chlorine content of the matrix glass varies significantly from 0·03 and 0·31 wt %, the proportion of chlorine degassing during magma ascent varies by a factor of only around three. The magma ascent rate varies by an order of magnitude from <1 to 9 m3/s, and therefore operates a major control on the rate of chlorine emission. However, a second potential major control is the continuation of chlorine degassing from previously erupted magma, through fluxing of the hot lava dome with external water vapour.
These observations can be used to constrain the chlorine emission from melt degassing during the eruption. We assume chlorine loss can be approximated as that from a typical lava dome sample that degasses 70% of its chlorine on eruption, as explosion-derived tephra accounts for only a very small proportion of the erupted magma volume. Assuming a total erupted volume of 300 x 106 m3 (dense rock equivalent, DRE) by March 1998, and 43 wt % melt with chlorine concentration of 0·34 wt % in the magma source region, the total HCl emission is 0·8 Mt. These estimates are in agreement with the model of Edmonds et al. (2001).
The observation that the melt is undersaturated in chlorine implies that only a vapour phase exsolves, rather than being accompanied by an additional chlorine-rich brine phase, as is the case for a chlorine-saturated melt (Shinohara et al., 1989). Therefore all the chlorine lost from the melt is transported into the escaping vapour phase and the total HCl emission estimated above should be equivalent to that measured in the volcanic plume. Remote-sensing data [Fourier transform IR spectroscopy (FTIR) and correlation spectrometer] indicate a total HCl emission of 0·7 Mt by March 1998 (Edmonds et al., 2001). The similarity of these estimates indicates that atmospheric HCl emission can be well accounted for by a combination of melt degassing during ascent and exchange with circulating water vapour. This supports the inference that any chlorine-bearing vapour phase present in the magma chamber or deeper in the system is not required to explain the emissions.
This model for chlorine degassing has implications for HCl/SO2 ratios of volcanic plumes, measured by FTIR gas monitoring. Gas emissions during the current eruption are interpreted in terms of HCl emission from ascent-driven Cl degassing closely correlating with extrusion rate, and emission of an SO2-rich phase sourced from the magma chamber controlled by edifice permeability (Edmonds et al., 2001). Although extrusion rate offers the strongest control on HCl emission, our model of Cl degassing suggests three important controls. First, variations in shallow-level crystallization and chlorine degassing result in a factor of three variation in HCl emission rate per volume of erupted magma (Table 1). Second, chlorine leaching by heterogeneous water vapour flux through the dome may increase HCl emission above that related to eruption of new magma. Third, chlorine degassing from hot block-and-ash-flow deposits may also increase HCl emission. In practice, at Soufrière Hills Volcano, the HCl emission for a given magma volume is largely controlled by the crystallinity of the samples, with the Cl content of the remaining melt being of secondary importance for calculation of total Cl degassed.
Relationship between degassing and crystallization
There are marked variations in both the chlorine contents and crystallinities of dome samples. Assuming that the magma chamber is relatively homogeneous in chlorine content and crystallinity, as suggested by the plagioclase melt inclusion data of Devine et al. (1998a), then these variations developed during magma ascent. The pumice fragments derived from relatively shallow levels in small-volume Vulcanian explosions (<300 m depth, assuming an average volume of 200 000 m3 (Druitt et al., 2002), show relative homogeneity of crystallinity and glass chlorine content. This suggests that the heterogeneities in the lava dome samples develop at a very shallow level, within the upper conduit and dome. Observations at Soufrière Hills Volcano show that lava dome growth involves extrusion of shear lobes (Watts et al., 2002). We suggest that variations in temperature and pressure across an individual shear lobe result in variations in crystallization and degassing histories between areas of the lobe. Variable residence times in the dome further contribute to these heterogeneities. Leaching of chlorine by localized flux of high-temperature groundwater vapour is an important factor in generating chlorine heterogeneity. Circulation of such water also has important implications for cooling of the lava dome and indicates that cooling-induced crystallization may be important in some areas of the dome. These factors account for the much greater variation in glass composition and groundmass crystallization in lava dome samples than in pumice clasts.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUSIONSAPPENDIX A: OPTIMIZING THE...APPENDIX B: STEP-HEATING...REFERENCES |
|---|
- Matrix glass compositions of glasses from the current eruption of Soufrière Hills Volcano indicate 20–70 wt % groundmass crystallization.
- Large variations in crystallinity and chlorine contents between lava blocks compared with explosion pumice clasts indicate that very shallow-level processes dominate degassing. The large spectrum of compositions and crystallinities observed is attributed to variation in extrusion rates and in residence times and locations within the lava dome.
- Models of open- and closed-system degassing cannot explain those samples with low melt Cl contents, particularly when simultaneous crystallization is taken into account. D values for matrix glasses, in particular those of dome lava samples, are heterogeneous and tend to be isotopically heavier than would be expected from simple open- or closed-system degassing models. Heterogeneous leaching of Cl by groundwater circulation within the lava dome and upper conduit is proposed. Isotopic exchange between melt and external water vapour would also lead to both D-enriched and heterogeneous melt D.
- The measurements of residual glass water contents in samples indicate minimum fragmentation pressures in the 1997 Vulcanian explosions of 1·1–3·7 MPa. A sample of a ballistic block from the sub-Plinian explosive eruption of 17 September 1996 has glass residual water content of 1·6%, consistent with rather higher pressure drops of 20 MPa in comparison with the 1997 Vulcanian explosions. Water contents in glasses from lava dome samples are low but variable and have implied pressures of up to 9 MPa.
- Chlorine emission is dependent on: degassing–crystallization history; rate of eruption; presence of a hot lava dome, which may continue to crystallize and degas chlorine; and leaching by water vapour flux through dome rock. HCl emission during the current eruption is well accounted for by degassing of chlorine from the melt during magma ascent and crystallization.
|
APPENDIX A: OPTIMIZING THE ELECTRON MICROPROBE OPERATING CONDITIONS FOR SOUFRIÈRE HILLS VOLCANO MATRIX GLASSES |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUSIONSAPPENDIX A: OPTIMIZING THE...APPENDIX B: STEP-HEATING...REFERENCES |
|---|
The loss of NaK
X-ray intensity during electron microprobe analysis (EMPA) of silicate glasses can result in underestimation of Na and overestimation of Si contents (through Si grow-in). Morgan & London (1996) found that routines applied to correct for this phenomenon may still underestimate sodium contents, and instead recommended EMPA conditions involving a 2 nA, 20 µm defocused beam to minimize Na loss to the level where no correction is necessary. However, the microcrystallinity of Soufrière Hills samples means that a beam size as large as 20 µm is impossible for the matrix glasses. The optimum conditions for EMPA of Soufrière Hills matrix glasses were therefore evaluated to minimize Na loss. Total sodium counts of Soufrière Hills matrix glasses were recorded on the Cameca Camebax electron microprobe electronically in blocks of 5 s, and the time-averaged count rate was calculated. The total sample exposure time is greater than the time measured as a result of small gaps unavoidably introduced between each 5 s block (Fig. A1). Two Soufrière Hills glasses were selected with different glass water contents: MVO47 (estimated <1 wt % H2O in glass) and MVO57 (1·6 wt % H2O in glass through bulk analysis). Analytical conditions were varied in terms of beam current (2, 5 and 10 nA), and rastered beam area (square of side 1, 2 and 5 µm). Glass analyses were subsequently carried out using the JEOL JXA-8600 electron microprobe. Use of four spectrometers allows analysis of elements that may be affected by problems associated with sodium loss (Na, Si, Al) to be completed rapidly. There is no significant difference between the results for samples MVO47 and MVO57 (Fig. A1), indicating that water content at these relatively low levels has no great effect on sodium stability. Sodium loss was, however, significantly increased when either beam current is increased or beam size is decreased.
|
Water content had no significant effect on Na loss at the relatively low levels in these two samples, and therefore all the samples can be considered together. There is a plateau lasting a total count time of
30 s (Fig. A1) where there is no count loss early in the analysis. Thus to minimize Na loss and achieve good precision, the optimum conditions for analysis of major element components of Soufrière Hills glasses involve a 2 nA, 5 µm width rastered area, with early analysis of Na, Si and Al. Na peak analysis is limited to 30 s count time, and the background reading from an initial reliable measurement is used for subsequent analyses on the same sample, with the first peak measurement discarded from the sample mean. For a 5 µm x 5 µm area Na loss is estimated to be <5%, and for smaller areas <10%. For analysis of chlorine, a minor component, a second analysis is carried out on the same spot, with 10 nA beam and a long Cl count time to optimize counting statistics precision. In practice, a rastered beam is used covering as large as area as possible, generally 5 µm x 4 µm, and always at least 4 µm x 3 µm (rastered areas on the JEOL probe are rectangular whereas on the Cameca probe they are square).
|
APPENDIX B: STEP-HEATING EXPERIMENTS TO ASSESS RELEASE OF WATER |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUSIONSAPPENDIX A: OPTIMIZING THE...APPENDIX B: STEP-HEATING...REFERENCES |
|---|
There are two difficulties in analysis of the water in glass by bulk extraction. First, complete water extraction is essential for accurate determination of both concentration and
D value of water in silicate glasses. However, heating silicate glasses does not always result in complete water release. For example, water solubility models have been corrected following recognition of this problem (Moore et al., 1998). A variety of heating schedules have been used for volcanic glasses but few details of rationale can be found in the literature. Second, efficient removal of surface-bound water must be ensured. Newman et al. (1988) found that the two components of water could be separated into the low- and high-temperature heating steps, provided that a grain size of >150 µm was used to reduce surface area effects. However, in large fragments of vesicular natural samples the effective surface area may be large.
Step-heating experiments were carried out to assess water and D release at various temperatures. The water produced at each step was converted to hydrogen, measured, collected and analysed for hydrogen isotopes. Periods of heating at each temperature were as follows: 
200°C, 7 h; 300–800°C, 1 h; 1400°C (fusion step), until no water remained. Two samples with different characteristics were selected. SSMON23 is a dense, dark grey ballistic sample from the 17 September 1996 explosion. MVO244 is a pumice block from a pumice-flow deposit formed by Vulcanian explosions in August 1997. MVO244 represents a sample with a high surface area. A grain size of 1–17 mm was selected to minimize the relative contribution of surface-adsorbed component. Hydrous amphibole crystals were removed by hand-picking. To test for complete water extraction FTIR analysis was carried out on glass residues following extraction. As thick a chip as possible (1 mm) was used to improve detection limit.
The bulk water contents of both samples are low: 0·23 wt % for MVO244 and 0·05 wt % for SSMON23. Spectra of water release with rising temperature show that water is released over a range of temperatures in both samples, with both a low- and a high-temperature peak of water release (Fig. A2). The release spectra of the two samples are, however, different. SSMON23 has two distinct phases of water release, one below and the other above 200°C, with no significant water release at 200–300°C. Further low-temperature step heating shows that the low-temperature water is released in approximately equal amounts at 120 and 200°C. The low- and high-temperature components of water are interpreted as surface-derived water and melt-bound water, respectively. The bulk melt-bound water content of this sample is therefore 0·027 wt %. MVO244 released more water than SSMON23, particularly at low temperatures. Almost half the total water content is released at 200°C. Medium temperatures give relatively little water. Water loss at high temperatures forms a peak at 800°C resembling that of SSMON23. Release of surface-bound and melt-bound water is thus harder to distinguish. However, the isotopic value of water released at mid-temperatures is similar to the average of those at high temperatures. Therefore, the total calculated melt D is little affected if one includes the mid-temperature water down to 200, 300 or 400°C (Table A2). FTIR analysis of glass chips following extraction showed no detectable water. This indicates <0·003 wt % H2O, which corresponds generally to less than a few percent of the total water extracted from the sample.
|
|
Dense lava samples release two distinct components of surface-adsorbed and melt-bound water. Measurement of melt-bound water can thus be readily achieved by degassing overnight at 200°C before collecting water from higher-temperature release. Pumiceous samples release large amounts of surface-adsorbed water at low to medium temperature. However, the total
D calculated is little affected if water released down to 200°C is included. Because of the variety of Montserrat samples, we decided to use a uniform procedure of overnight degassing at 200°C and then water collection at higher temperatures. We recognize that the water content of pumiceous samples may be slightly overestimated and that uncertainty in D for such values is greater. |
ACKNOWLEDGEMENTS |
|---|
C.H. was supported by an NERC studentship and R.S.J.S. was supported by an NERC Fellowship. We acknowledge the important contributions of many colleagues at the Montserrat Volcano Observatory, who provided many of the samples used in this study. Joe Devine and Jim Gardner are thanked for their helpful comments on an earlier version of the manuscript. The involvement of SUERC was facilitated by approved support to the project through the NERC Argon isotope facility (Dr M. Pringle); this support is acknowledged. The manuscript was improved by the reviews of D. Pyle, B. Villemant and J. Webster.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUSIONSAPPENDIX A: OPTIMIZING THE...APPENDIX B: STEP-HEATING...REFERENCES |
|---|
Alidibirov, M. & Dingwell, D. B. (1996). Magma fragmentation by rapid decompression. Nature 380, 146–148.[CrossRef]
Anderson, S. W. & Fink, J. H. (1989). Hydrogen-isotope evidence for extrusion mechanisms of the Mount St Helens lava dome. Nature 341, 521–523.[CrossRef]
Bai, T. B. & van Groos, A. F. K. (1994). Diffusion of chlorine in granitic melts. Geochimica et Cosmochimica Acta 58, 113–123.[CrossRef][Web of Science]
Barclay, J., Rutherford, M. J., Carroll, M. R., Murphy, M. D., Devine, J. D., Gardner, J. & Sparks, R. S. J. (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]
Blundy, J. & Cashman, K. (2001). Ascent-driven crystallisation of dacite magmas at Mount St Helens, 1980–1986. Contributions to Mineralogy and Petrology 140, 631–650.[Web of Science]
Carroll, M. R. & Webster, J. D. (1994). Solubilities of sulfur, noble-gases, nitrogen, chlorine, and fluorine in magmas. In: Carroll, M. R. & Holloway, J. R. (eds) Volatiles in Magmas. Mineralogical Society of America, Reviews in Mineralogy 30, 231–279.
Cashman, K. V. (1992). Groundmass crystallization of Mount St Helens dacite, 1980–1986—a tool for interpreting shallow magmatic processes. Contributions to Mineralogy and Petrology 109, 431–449.[CrossRef][Web of Science]
Chiodini, G., Cioni, R., Frullani, A., Guidi, M., Marini, L., Prati, F. & Raco, B. (1996). Fluid geochemistry of Montserrat Island, West Indies. Bulletin of Volcanology 58, 380–392.[CrossRef][Web of Science]
Cole, P. D., Calder, E. S., Sparks, R. S. J., Clarke, A. B., Druitt, T. H., Young, S. R., Herd, R. A., Harford, C. L. & Norton, G. E. (2002). Deposits from dome-collapse and fountain-collapse pyroclastic flows at Soufrière Hills Volcano, Montserrat. In: Druitt, T. H. & Kokelaar, B. P. (eds) The Eruption of Soufrière Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoir 21, 231–262.
Devine, J. D., Murphy, M. D., Rutherford, M. J., Barclay, J., Sparks, R. S. J., Carroll, M. R., Young, S. R. & Gardner, J. E. (1998a). Petrologic evidence for pre-eruptive pressure–temperature conditions, and recent reheating, of andesitic magma erupting at the Soufrière Hills Volcano, Montserrat, WI. Geophysical Research Letters 25, 3669–3672.[CrossRef][Web of Science]
Devine, J. D., Rutherford, M. J. & Gardner, J. E. (1998b). Petrologic determination of ascent rates for the 1995–1997 Soufrière Hills Volcano andesitic magma. Geophysical Research Letters 25, 3673–3676.
Dobson, P. F., Epstein, S. & Stolper, E. M. (1989). Hydrogen isotope fractionation between coexisting vapor and silicate-glasses and melts at low-pressure. Geochimica et Cosmochimica Acta 53, 2723–2730.[CrossRef][Web of Science]
Druitt, T. H., Young, S. R., Baptie, B., Bonadonna, C., Calder, E. S., Clarke, A. B., Cole, P. D., Harford, C. L., Herd, R. A., Luckett, R., Ryan, G. & Voight, B. (2002). Episodes of repetitive Vulcanian explosions and fountain collapse at Soufrière Hills Volcano, Montserrat. In: Druitt, T. H. & Kokelaar, B. P. (eds) The Eruption of Soufrière Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoir 21, 281–306.
Edmonds, M., Pyle, D. & Oppenheimer, C. (2001). A model for degassing at the Soufrière Hills Volcano, Montserrat, West Indies, based on geochemical data. Earth and Planetary Science Letters 186, 159–173.[CrossRef][Web of Science]
Fallick, A. E., Macaulay, C. I. & Haszeldine, R. S. (1993). Implications of linearly correlated oxygen and hydrogen isotopic compositions for kaolinite and illite in the Magnus sandstone, North Sea. Clays and Clay Minerals 41, 184–190.[Abstract]
Gardner, J. E., Thomas, R. M. E., Jaupart, C. & Tait, S. R. (1996). Fragmentation of magma during Plinian volcanic eruption. Bulletin of Volcanology 58, 144–162.[CrossRef][Web of Science]
Gardner, J. E., Rutherford, M. & Hort, M. (1998). Degassing of trace gases during volcanic eruptions. EOS Transactions, American Geophysical Union 79, F936.
Harford, C. L. & Sparks, R. S. J. (2001). Recent remobilisation of shallow-level intrusions on Montserrat revealed by hydrogen isotope composition of amphiboles. Earth and Planetary Science Letters 185, 285–297.[CrossRef][Web of Science]
Hildreth, W. & Drake, R. E. (1992). Volcan Quizapu, Chilean Andes. Bulletin of Volcanology 54, 93–125.[CrossRef][Web of Science]
Holloway, J. R. & Blank, J. G. (1994). Application of experimental results to C–O–H species in natural melts. In: Carroll, M. R. & Holloway, J. R. (eds) Volatiles in Magmas. Mineralogical Society of America, Reviews in Mineralogy 30, 187–230.
Hort, M. & Gardner, J. E. (2000). Constraints on degassing of pumice clasts during Plinian eruptions based on model calculations. Journal of Geophysical Research 105, 25981–26001.[CrossRef]
International Atomic Energy Agency (1992). Statistical Treatment of Data on Environmental Isotopes in Precipitation, 331. Vienna: IAEA.
Kohn, S. C. (2000). The dissolution mechanisms of water in silicate melts; a synthesis of recent data. Mineralogical Magazine 64, 389–408.
Kusakabe, M., Sato, H., Nakada, S. & Kitamura, T. (1999). Water contents and hydrogen isotopic ratios of rocks and minerals from the 1991 eruption of Unzen volcano, Japan. Journal of Volcanology and Geothermal Research 89, 231–242.[CrossRef][Web of Science]
Melnik, O. & Sparks, R. S. J. (1999). Non-linear dynamics of lava dome extrusion. Nature 402, 37–41.[CrossRef]
Metrich, N. & Rutherford, M. J. (1992). Experimental study of chlorine behavior in hydrous silicic melts. Geochimica et Cosmochimica Acta 56, 607–616.[CrossRef][Web of Science]
Moore, G., Vennemann, T. & Carmichael, I. S. E. (1998). An empirical model for the solubility of H2O in magmas to 3 kilobars. American Mineralogist 83, 36–42.[Abstract]
Morgan, G. B. & London, D. (1996). Optimizing the electron microprobe analysis of alkali aluminosilicate glasses. American Mineralogist 81, 1176–1185.[Abstract]
Murphy, M. D., Sparks, R. S. J., Barclay, J., Carroll, M. R. & Brewer, T. S. (2000). Remobilization of andesite magma by intrusion of mafic magma at the Soufrière Hills Volcano, Montserrat, West Indies. Journal of Petrology 41, 21–42.
Newman, S., Epstein, S. & Stolper, E. (1988). Water, carbon-dioxide, and hydrogen isotopes in glasses from the ca 1340 AD eruption of the Mono Craters, California—constraints on degassing phenomena and initial volatile content. Journal of Volcanology and Geothermal Research 35, 75–96.[CrossRef][Web of Science]
Norton, G. E., Aspinall, W. P., Baptie, B., Herd, R. A., Harford, C. L., Loughlin, S. C., Luckett, R., Ritchie, L., Rowley, K. C., Sparks, R. S. J., Watts, R. & Young, S. R. (2002). Post-dome-growth activity at Soufrière Hills volcano, Montserrat: March 1998–November 1999. In: Druitt, T. H. & Kokelaar, B. P. (eds) The Eruption of Soufrière Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoir 21, 467–481.
Nowak, M. and Behrens, H. (1995). The speciation of water in haplogranitic glasses and melts determined by in-situ near-infrared spectroscopy. Geochimica et Cosmochimica Acta 59, 3445–3450.[CrossRef][Web of Science]
Robertson, R. E. A., Cole, P., Sparks, R. S. J., Harford, C. L., Lejeune, A. M., McGuire, W. J., Miller, A. D., Murphy, M. D., Norton, G., Stevens, N. F. & Young, S. R. (1998). The explosive eruption of Soufrière Hills Volcano, Montserrat, West Indies, 17 September, 1996. Geophysical Research Letters 25, 3429–3432.[CrossRef][Web of Science]
Robertson, R. E. A., Aspinall, W. P., Herd, R. A., Norton, G. E., Sparks, R. S. J. & Young, S. R. (2000). The 1995–1998 eruption of the Soufriere Hills volcano, Montserrat, WI. Philosophical Transactions of the Royal Society of London, Series A 358, 1619–1637.[CrossRef]
Shepherd, J. B., Herd, R. A., Jackson, P. & Watts, R. (1998). Ground deformation measurements at the Soufrière Hills volcano, Montserrat: II: rapid static GPS measurements June 1996–June 1997. Geophysical Research Letters 25, 3413–3416.[CrossRef][Web of Science]
Shinohara, H., Iiyama, J. T. & Matsuo, S. (1989). Partition of chlorine compounds between silicate melt and hydrothermal solutions 1. Partition of NaCl–KCl. Geochimica et Cosmochimica Acta 53, 2617–2630.[CrossRef][Web of Science]
Signorelli, S. & Carroll, M. R. (2001). Chlorine solubility in peraluminous rhyolites from Soufriere Hills volcano, Montserrat: implications for degassing. Bulletin of Volcanology 62, 431–440.[CrossRef][Web of Science]
Silver, L. A., Ihinger, P. D. & Stolper, E. (1990). The influence of bulk composition on the speciation of water in silicate-glasses. Contributions to Mineralogy and Petrology 104, 142–162.[CrossRef][Web of Science]
Sparks, R. S. J., Young, S. R., Barclay, J., Calder, E. S., Cole, P., Darroux, B., Davies, M. A., Druitt, T. H., Harford, C., Herd, R., James, M., Lejeune, A. M., Loughlin, S., Norton, G., Skerrit, G., Stevens, N. S., Toothill, J., Wadge, G. & Watts, R. (1998). Magma production and growth of the lava dome of the Soufriere Hills volcano, Montserrat: November 1995 to December 1997. Geophysical Research Letters 25, 3421–3424.[CrossRef][Web of Science]
Sparks, R. S. J., Murphy, M. D., Lejeune, A. M., Watts, R. B., Barclay, J. & Young, S. R. (2000). Control on emplacement of the andesite lava dome of the Soufrière Hills volcano by degassing-induced crystallisation. Terra Nova 12, 14–20.[CrossRef][Web of Science]
Suzuoki, T. & Epstein, S. (1976). Hydrogen isotope fractionation between OH-bearing minerals and water. Geochimica et Cosmochimica Acta 40, 1229–1240.[CrossRef][Web of Science]
Taran, Y. A., Pokrovsky, B. G. & Volynets, O. N. (1997). Hydrogen isotopes in hornblendes and biotites from Quaternary volcanic rocks of the Kamchatka–Kurile arc. Geochemical Journal 31, 203–221.[Web of Science]
Taylor, B. E. (1986). Magmatic volatiles: isotopic variation of C, H, and S. In: Valley, J. W., Taylor, H. P. & O'Neil, J. R. (eds) Stable Isotopes in High Temperature Geological Processes. Mineralogical Society of America, Reviews in Mineralogy 16, 185–225.
Taylor, B. E., Eichelberger, J. C. & Westrich, H. R. (1983). Hydrogen isotopic evidence of rhyolitic magma degassing during shallow intrusion and eruption. Nature 306, 541–545.[CrossRef]
Villemant, B. & Boudon, G. (1999). H2O and halogen (F, Cl, Br) behaviour during shallow magma degassing processes. Earth and Planetary Science Letters 168, 271–286.[CrossRef][Web of Science]
Voight, B., Sparks, R. S. J., Miller, A. D., Stewart, R. C., Hoblitt, R. P., Clarke, A., Ewart, J., Aspinall, W. P., Baptie, B., Calder, E. S., Cole, P., Druitt, T. H., Harford, C., Herd, R. A., Jackson, P., Lejeune, A. M., Lockhart, A. B., Loughlin, S. C., Luckett, R., Lynch, L., Norton, G. E., Robertson, R., Watson, I. M., Watts, R. & Young, S. R. (1999). Magma flow instability and cyclic activity at Soufrière Hills Volcano, Montserrat, British West Indies. Science 283, 1138–1142.
Watts, R. B., Herd, R. A., Sparks, R. S. J. & Young, S. R. (2002). Growth patterns and emplacement of the andesitic lava dome at Soufrière Hills Volcano, Montserrat. In: Druitt, T. H. & Kokelaar, B. P. (eds) The Eruption of Soufrière Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoir 21, 115–152.
Webster, J. D. & De Vivo, B. (2002). Experimental and modeled solubilities of chlorine in aluminosilicate melts, consequences of magma evolution, and implications for exsolution of hydrous chloride melt at Somma–Vesuvius. American Mineralogist 87, 1046–1061.
Webster, J. D., Kinzler, R. J. & Mathez, E. A. (1999). Chloride and water solubility in basalt and andesite liquids and implications for magma degassing. Geochimica et Cosmochimica Acta 63, 729–738.[CrossRef][Web of Science]
Wilson, M. (1989). Igneous Petrogenesis. London: Unwin Hyman, 466 pp.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  S. J. Lane, J. C. Phillips, and G. A. Ryan Dome-building eruptions: insights from analogue experiments Geological Society, London, Special Publications, January 1, 2008; 307(1): 207 - 237. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Pichavant, F. Costa, A. Burgisser, B. Scaillet, C. Martel, and S. Poussineau Equilibration Scales in Silicic to Intermediate Magmas Implications for Experimental Studies J. Petrology, October 1, 2007; 48(10): 1955 - 1972. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. B. Watson and T. M. Harrison Zircon Thermometer Reveals Minimum Melting Conditions on Earliest Earth Science, May 6, 2005; 308(5723): 841 - 844. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||