February 2017 Issue Table of Contents
The role of volatiles in magma dynamics and eruption style is fundamental. Magmatic volatiles partition between melt, crystal, and vapor phases and, in so doing, change magma properties. This has consequences for magma buoyancy and phase equilibria. An exsolved vapor phase, which may be distributed unevenly through reservoirs, contains sulfur and metals that are either transported into the atmosphere or into ore deposits. This article reviews the controls on volatile solubility and the methods to reconstruct the volatile budget of magmas, focusing particularly on the exsolved vapor phase to explore the role of volatiles on magma dynamics and on eruption style.
Keywords: exsolved volatiles, vapor, volcanic gases, magma reservoir, eruption style
Volcanic eruptions, in all their diverse forms, are driven by overpressure, buoyancy, and degassing. Subaerial eruptions produce vast clouds of volcanic gases, a process that has shaped our hydrosphere over Earth’s history. Many rocks produced by volcanic eruptions, particularly explosive eruptions, are dominated by vesicles—bubbles frozen in place during eruption (Fig. 1). There can be no doubt that to understand volcanic processes, the mechanisms and consequences of magmatic volatile degassing are paramount.
Magmatic volatiles are chemical constituents in silicate melts that partition into a magmatic vapor phase at low (crustal) pressures. The primary magmatic volatiles that drive volcanic eruptions are water (H2O) and carbon dioxide (CO2). Volatiles are only present in small amounts in magmas (a few wt%), but their influence belies their weight. The presence of dissolved volatiles in silicate melts controls the abundance and composition of crystal phases that grow in the magma during cooling and decompression, as well as the density of silicate melts (Ochs and Lange 1999). This has consequences for buoyant magma ascent through the crust, as well as for convection in magma reservoirs.
Within a few kilometers of the Earth’s surface, silicate melts typically reach saturation with respect to a magmatic vapor phase, which is disseminated as bubbles. Volatile exsolution can occur as a result of isobaric cooling and crystallization in magma reservoirs, and this can lead to an increase in reservoir pressure (because the surrounding country rocks are relatively incompressible) that can trigger eruptions (Tait et al. 1989). Volatile exsolution also occurs during decompression and eruption of magma, causing tremendous expansion of bubbly magma and its acceleration up the conduit. The rate and efficiency of exsolution and vapor loss during ascent, as well as the rheological properties of the magma, largely control eruption style.
The magmatic vapor phase in crustal magma reservoirs is predominantly CO2 at depth, becoming more H2O-rich at lower pressures (Fig. 2A). Volcanic gases (a term used to describe the vapor phase when discharged at the surface through a vent or fumarole) contain a myriad of other chemical species, chief of which are sulfur (Fig. 2B) and halogens. Sulfur partitions strongly into the vapor phase at low pressures (Scaillet et al. 1998; Fig. 2C). In relatively oxidised and low temperature magmas in the upper crust, most of the sulfur is likely to exist in the vapor phase rather than being dissolved in the melt (Fig. 2D). A consequence of this is that explosive volcanic eruptions typically release far more sulfur (as sulfur dioxide) than can be accounted for by the amount dissolved in melt inclusions (tiny aliquots of melt trapped in crystals) (Wallace 2001). Chlorine, which also partitions into the vapor, forms complexes with metals, a key part of the process of forming porphyry Cu deposits and epithermal Au deposits in the shallow plumbing systems of some volcanic systems. For highly saline magmas, a high-density magmatic vapor phase at depth separates into a lower density vapor and a brine, causing metals to partition between the two phases and helping to enrich the concentrations of these metals.
The presence of vapor bubbles changes a magma’s bulk physical and rheological properties. Bubbles make magma compressible. Compressible magma responds to injections or evacuations of magma (e.g. during eruption) by contracting or expanding, like a “magma sponge” (Voight et al. 2010), and this property has consequences for eruption magnitude and duration (Huppert and Woods 2002) and volcano monitoring. The exsolution of H2O from silicate melts removes a “network modifier”, resulting in the lengthening of chains of corner-sharing silicate tetrahedra in the melt, thereby increasing melt viscosity. The interplay between exsolution, viscosity change, and outgassing during magma ascent is, therefore, critical in determining eruption style.
The behavior of the exsolved vapor phase in long-lived, vertically extensive, mush-dominated magmatic systems beneath volcanoes remains poorly understood and is the focus of much recent research. Vapor bubbles may be retained in crystal-rich magmas by capillary forces at low gas fractions, or the bubbles may be transported through quasi-brittle fractures at high vapor mass fractions and might modify the bulk rheological properties of crystal mushes, rendering them able to respond to magma recharge (by mingling and reorganization) on relatively fast timescales (Huber et al. 2011).
The reconstruction of volatile budgets in magmas that feed volcanic eruptions has received much attention from the perspective of understanding climate impacts, mantle volatile systematics, magma storage, and volcanic processes. Various tools exist to determine volatile concentrations and speciation in silicate melts and their exsolution history, including melt inclusion geochemistry (Lowenstern 1995), phase equilibria experiments, and thermodynamic models. More challenging is the evaluation of how exsolved vapor is generated and distributed in magma reservoirs. Owing to buoyancy, vapor bubbles may segregate from their source magma. Long-lived reservoirs, which are subject to sporadic recharge and mingling of magmas, may develop complex reservoir architectures over time, with segregated regions of melts, mushes, and exsolved vapor (Christopher et al. 2015).
This article reviews the evidence for the formation, distribution, and form of exsolved vapor in magma reservoirs and the consequences for volcanic processes.
Abundance and Distribution of Exsolved Vapor in Magma Reservoirs
Vapor saturation of silicate melts occurs when the sum of the partial pressures of the dissolved volatiles in a silicate melt is equal to the confining pressure, at which point a multicomponent vapor phase will be in equilibrium with the magma. The solubility of the volatiles CO2 and H2O is mainly controlled by pressure (Fig. 2A). The much lower solubility of CO2 compared to H2O causes vapor at higher pressures to be more CO2 rich, and at lower pressures, more H2O rich. Bubbles are the consequence of vapor-saturation of magmas, and their nucleation and growth in silicate melts accommodates the exsolving vapor phase. Bubble nucleation may be homogeneous (in melt) or heterogeneous (on crystals) and requires volatile supersaturation to overcome surface tension. The extent of supersaturation is usually small and easily achieved in decompressing or crystallizing a magma, except in the case of homogeneous nucleation of bubbles in crystal-free rhyolitic magma, where strong melt supersaturations may develop.
Other volatiles, such as sulfur and halogens, partition into the exsolved vapor phase to varying extents. Sulfur and chlorine partition behavior is well understood for a wide range of oxidation states and silicate melt compositions (Zajacz et al. 2012). In general, experiments indicate that sulfur will partition strongly into the vapor phase, particularly for more reducing conditions below the sulfate–sulfide transition [at an oxygen fugacity of ~fayalite–magnetite–quartz (FMQ) + 1 log unit] (Fig. 2B, 2C), caused by the lower solubility of sulfur when it exists as sulfide (S2−) than when it occurs, under more oxidising conditions, as sulfate (S6+). Saturation of the silicate melt with sulfide melt or with anhydrite (at more oxidising conditions) limits the sulfur concentration in the coexisting vapor phase. Sulfur partitioning into the vapor phase is much more pronounced for more silica-rich compositions (Fig. 2C) (Zajacz et al. 2012). Chlorine partitions less strongly into a vapor phase at magma chamber conditions than sulfur. Chlorine-rich vapor is important for transporting metals to the sites of hydrothermal ore deposits. Experimental data are consistent with the idea that for long-lived magmatic systems in which evolved magmas are recharged by underplating mafic magmas, the vapor phase coexisting with the evolved magma will be more chlorine-rich, whilst the vapor phase supplied by the mafic magmas will be sulfur-rich, to a degree that depends on oxidation state and whether saturation with respect to a sulfur-bearing phase has been reached.
Observations Constraining Vapor Saturation of Silicate Melts in the Crust
Primitive mafic magmas in arc settings have an average H2O content of ~4 wt% (Plank et al. 2013), whereas mid-ocean ridge basalts have 0.3–0.5 wt% H2O and ocean island basalts associated with hotspots have 0.5–1.5 wt% H2O (Dixon et al. 2002). Concentrations of CO2 in different types of basaltic magma are more difficult to constrain using melt inclusions and submarine pillow basalts because low solubility causes deep degassing during ascent, but initial values are likely to be ~2,000 ppm CO2 for primary mid-ocean ridge basalt magmas (Michael and Graham 2015), up to several wt% for more trace-element-enriched oceanic and hotspot magmas, and possibly up to 1 wt% in primary arc magmas. In arc systems, the deep magmatic vapor phase released by magma intrusion into the crust has been linked to flux melting and assimilation of lower crustal rocks and the generation of intermediate magmas, as well as to the linked processes of lower crustal dehydration (by extraction of water-rich partial melts) and the formation of granulite terrains.
There is strong evidence that magmas stored in the mid to upper crust prior to eruption are commonly vapor-saturated as a consequence of magma differentiation, prolonged storage, and recharge or underplating by CO2-rich mafic magma (Wallace 2001). The geochemistry of melt inclusions hosted by quartz in the Bishop Tuff (Long Valley, California, USA) have been used to reconstruct the distribution and abundance of the exsolved vapor phase (Wallace et al. 1999) within the Long Valley Caldera magma body. Volatiles and trace element concentrations are consistent with vapor-saturated crystallization in a magma chamber zoned with respect to exsolved vapor, varying from ~1 wt% exsolved vapor near the bottom of the body (at ~250 MPa) to ~6 wt% near the top (at ~150 MPa) (Wallace et al. 1999).
Another approach to quantifying the abundance of exsolved vapor in magma reservoirs is to compare the mass flux of volcanic gases with the flux of magma erupted. At the Soufrière Hills Volcano (Montserrat), measurements of the gas composition and flux during the eruption (1995–2011) permitted estimates to be made of 2–8 wt% exsolved vapor in the magma prior to eruption (Edmonds et al. 2014), similar to estimates for the Bishop Tuff. This amount of exsolved vapor would impart significant compressibility to the magma, which would cause the magma to behave like a sponge, easily compressed or expanded (Voight et al. 2010), thus buffering magma reservoir volume changes. A muted ground deformation signal was in fact observed during eruption of the Soufrière Hills Volcano: the observed volume decrease (deflation) during periods of eruption was only around one tenth of the volume erupted (Voight et al. 2010) (Fig. 3). The presence of exsolved vapor also has the effect of greatly increasing eruption longevity owing to the greater compressibility, leading to eruption of a larger mass to relieve the same overpressure (Huppert and Woods 2002).
Large sulfur dioxide clouds accompanying explosive eruptions require a preeruptive vapor phase containing sulfur in the magma reservoir (Wallace 2001). Notably, the eruptions associated with the largest sulfur clouds per erupted unit volume of magma appear to be intermediate arc magmas (andesites and dacites) (Wallace 2001). An example of this is the eruption cloud shown in Figure 3A, from Okmok Volcano (Alaska, USA) in 2008. Sulfur partitioning into the vapor phase is maximized when the magma is relatively cool, silica- and H2O-rich, and alkali-poor, conditions met by the vast majority of explosive arc eruptions. The opportunity to observe a contrast to this sulfur-rich case may have presented itself recently: the Plinian eruption of Chaitén Volcano (Chile) in 2008 was associated with a remarkably sulfur-poor cloud (Carn et al. 2009) that accompanied the eruption of crystal-poor rhyolite. This low abundance of gaseous sulfur may have been due to the lack of crystals, which may have played a role in restricting vapor storage prior to eruption. Understanding the spatial and temporal controls on the amount, composition, and mobility of exsolved vapor in magma reservoirs has important implications for our understanding of the climate and environmental effects of eruptions, the geochemical cycling of volatiles by plate tectonics, and the generation of ore deposits. The largest volcanic sulfur clouds cause global cooling of perhaps a few degrees centigrade over year-long timescales due to the stratospheric volcanic sulfate aerosol reflecting and absorbing solar radiation. These eruptions result in severe environmental degradation and perhaps even cessation of photosynthesis and mass extinction (Self 2006).
Dynamics of Vapor-Rich Magma in the Crust and Eruption Triggering
The role of bubbles in magma mixing, eruption triggering, and the dynamics of magma reservoirs has been studied using analog materials and modelling. Mafic magmas underplating more silicic, viscous magmas can produce a range of behaviors. If the mafic magma is vapor-saturated, bubbles may accumulate at the interface between the mafic and the felsic magmas, lowering the bulk density of the mafic magma, which then may induce overturn, mafic enclave formation, or bubble rise through the interface, depending on the viscosity contrast between the two magmas (Thomas and Tait 1997). Heating and remobilization of crystal-rich magma in the crust may take place by “gas sparging”, a process of mafic underplating, quenching, and outgassing of vapor that advects heat through the pore spaces of the crystal-rich magma, causing it to partially melt and perhaps trigger an eruption (10s of km3 in size) (Bachmann and Bergantz 2006).
A great deal of attention has been devoted to the microphysics of multiphase crystal mushes. The presence of an exsolved vapor phase has implications for bulk mush rheological properties: only a few weight per cent of exsolved vapor is needed to substantially reduce the effective bulk viscosity of a crystal-rich magma (e.g. by four orders of magnitude for the addition of 9 vol% vapor in a 70% crystallinity magma) and to induce shear-thinning behavior, where the viscosity decreases with increasing strain rate (Pistone et al. 2013). Vapor bubbles could make mushes more mobile in response to magma recharge events, potentially allowing larger volumes of magma to be erupted, whereas the removal of such a vapor phase (through outgassing) could result in “viscous death” and the formation of plutonic bodies. Injection of vapor bubbles into crystal-rich suspensions in analogue materials shows that mushes may also behave in a brittle way with a yield strength, allowing the magmatic vapor phase to migrate relatively rapidly through them (Fig. 4) (Oppenheimer et al. 2015).
Exsolution of Volatiles During Eruption
Eruptions may be triggered when overpressures exceed the tensile strength of the country rocks, allowing magma to ascend along fractures towards the surface. These overpressures may be caused by magma recharge or by isobaric vesiculation during crystallization. As magma ascends, preexisting bubbles grow, or a new population of bubbles may nucleate, owing to the continued lowering of H2O and CO2 solubilities in silicate melts at low pressures (Fig. 2A). Bubble growth during magma decompression is limited by the rate of diffusion of volatiles into bubbles, by the rate of viscous deformation of melt as bubbles expand, and by the decompression rate.
The growth of vapor bubbles driven by H2O exsolution has multiple immediate consequences: a rise in both the melt viscosity (Dingwell et al. 1996) and the solidus temperature, inducing a rapid burst of crystallization; and lowering of the magma’s bulk density, particularly in the uppermost few kilometers of the conduit, causing acceleration of the magma (through conservation of mass). This combination of processes makes for a rich variety of volcanic eruption styles (Fig. 5), which are dependent on the interplay between magma decompression rate and the rheological properties of the melt.
Influence of Volatile Exsolution and Outgassing on Volcanic Eruption Styles
For basaltic eruptions, which involve relatively low viscosity melt, eruption style is governed by ascent rate and the dynamics of two-phase flow (Houghton et al. 2016). At low ascent rates, bubbles rise through melts in the conduit, yielding an eruptive spectrum from quiescent degassing to effusive activity to mild Strombolian eruptions, accompanied by persistent tropospheric gas plumes. This range of activity also broadly includes magma convection in a conduit, supplying volatiles to the atmosphere through outgassing of rising and bursting bubbles on the surface of a lava lake, followed by sinking of denser degassed magma.
Explosive styles of basaltic volcanic eruptions fall into the categories of Hawaiian and Strombolian (Fig. 5) and may transition between the two. Observations, analog experiments, and textural studies (Houghton et al. 2016) demonstrate that these styles display a continuum in eruption intensity and magnitude. The styles can range from Strombolian eruptions that are discrete in duration (typically < 100 s) and have mass fluxes of 102–104 kg/s, to Hawaiian eruptions that are more long-lived (typically > 2 h) and have mass fluxes of 104–106 kg/s (Houghton et al. 2016). The differences in duration and vigor are caused by fluctuations in eruption rate (driven by the pressure regime in a magma chamber). At low magma ascent rates, rising gas slugs and bubbles dominate, with their mass growth limited by diffusion. Strombolian activity is caused by the bursting of segregated single slugs or trains of bubbles. At higher magma ascent rates, continuous vesiculation and expansion cause inertia-dominated magma fragmentation (Namiki and Manga 2008) and Hawaiian fountaining. Violent Strombolian activity occurs in more H2O-rich basaltic magmas at mass fluxes of 104–105 kg/s and is, thus, intermediate between the Strombolian and sub-Plinian regimes (Pioli et al. 2008).
For high viscosity magmas (crystal-rich andesites, dacites, and rhyolites), magma decompression rate and the rheological properties of the magma control the style of eruption. Here, the viscous retardation of bubble growth generates bubble overpressure. Magma fragmentation is driven by overpressure overcoming the tensile strength of the surrounding melt. For the case of Vulcanian eruptions, the high bulk viscosity of the magma precludes conduit refilling on the timescale of the eruption, rendering the eruption discrete in duration and limited in magnitude. For slightly lower bulk viscosity magmas (e.g. crystal-free rhyolites) and large magma chamber overpressures, Plinian eruptions, with magma column heights of 10s of km that often penetrate the stratosphere, are driven by continuous magma fragmentation and magma flow, refilling the conduit on timescales of eruption (Fig. 5). The primary mode of magma fragmentation here might be brittle failure caused by rapid strain rates experienced by the rapidly vesiculating and expanding magma (Papale 1999).
At low magma ascent rates, when relaxation of viscous stresses in the silicate melt can keep pace with bubble growth, permeable bubble networks develop as the melt vesiculates, allowing the magma to effectively outgas volatiles both upward into the atmosphere and laterally into shear-fractured conduit margins and country rocks, thus preventing magma fragmentation and explosive eruptions. Effective magma permeabilities for outgassing may develop at porosities of around 30%, but perhaps at much lower porosities for sheared magmas (Rust and Cashman 2004). Under conditions of high magma porosity and efficient gas loss, magmas erupt effusively in the form of steep-sided lava flows or domes.
Werner Giggenbach (1937–1997, German-born New Zealand-based chemist and volcanologist) proposed that andesitic volcanoes are “ventholes” that allow excess subducted volatiles to be recycled to the surface, where the term “venthole” implies the rise of a free volatile phase to the surface from a zone of arc magma generation. While this may be an extreme view, the observations suggest that, in most cases, magma bodies in the crust not only are vapor-saturated but also need to become charged with a magmatic vapor phase before they can erupt. Indeed, vapor saturation in crustal magma bodies may be an inevitable consequence of the high CO2 contents of most mantle-derived mafic magmas, causing continuous fluxing of overlying silicic magmas by CO2-rich vapor to occur. Within this framework, long periods of relatively small-scale eruptions and unrest at volcanoes such as Montserrat or Popocatépetl (Mexico) may be viewed as essentially intrusive events in which the mass of erupted magma, while potentially devastating locally, is small in comparison to the likely masses of both stored, differentiated magma and mafic, recently intruded magma at depth. Explosive eruptions of much larger magma bodies, however, contain within them the exsolved gases released from underlying mush zones and recharging mafic magmas (Parmigiani et al. 2016): these might reflect thousands or even tens of thousands of years of vapor-phase accumulation (Christopher et al. 2015).
The next key step in understanding volcanic systems is to develop methods for tracking the movement of the magmatic vapor phase independently of magma movement. Four potential methods are here proposed: (1) tracking of the concentrations and fluxes of volatile components of different solubilities, using the composition and density of melt inclusions and fluid inclusions to reconstruct equilibration depths, and using volcanic emissions to estimate total depth-averaged volatile fluxes; (2) phase equilibrium experiments and textural studies that can reveal processes such as gas fluxing; (3) seismic, ground deformation, and other geophysical methods to track mass and density changes that might be linked to outgassing; and (4) use of volatile isotopes such as 210Po that can reveal information about the volumes and timescales of degassing magma bodies.
We thank Keith Putirka, Jake Lowenstern and Rosario Esposito for their detailed and thoughtful reviews, which improved this paper immensely.
Bachmann O, Bergantz GW (2006) Gas percolation in upper-crustal silicic crystal mushes as a mechanism for upward heat advection and rejuvenation of near-solidus magma bodies. Journal of Volcanology and Geothermal Research 149: 85-102
Carn SA and 7 coauthors (2009) The unexpected awakening of Chaitén volcano, Chile. EOS, Transactions, American Geophysical Union 90: 205-206
Christopher TE and 7 coauthors (2015) Crustal‐scale degassing due to magma system destabilization and magma‐gas decoupling at Soufrière Hills Volcano, Montserrat. Geochemistry, Geophysics, Geosystems 16: 2797-2811
Dingwell DB, Romano C, Hess K-U (1996) The effect of water on the viscosity of a haplogranitic melt under P-T-X conditions relevant to silicic volcanism. Contributions to Mineralogy and Petrology 124: 19-28
Dixon JE, Leist L, Langmuir C, Schilling J-G (2002) Recycled dehydrated lithosphere observed in plume-influenced mid-ocean-ridge basalt. Nature 420: 385-389
Edmonds M and 12 coauthors (2014) Pre-eruptive vapour and its role in controlling eruption style and longevity at Soufrière Hills Volcano. In: Wadge G, Robertson REA, Voight B (eds) The Eruption of Soufriére Hills Volcano, Montserrat from 2000 to 2010. Geological Society, London, Memoir 39, pp 291-315
Houghton B and 11 coauthors (2016) Stronger or longer: Discriminating between Hawaiian and Strombolian eruption styles. Geology 44: 163-166
Huber C, Bachmann O, Dufek J (2011) Thermo-mechanical reactivation of locked crystal mushes: Melting-induced internal fracturing and assimilation processes in magmas. Earth and Planetary Science Letters 304: 443-454
Huppert HE, Woods AW (2002) The role of volatiles in magma chamber dynamics. Nature 420: 493-495
Jugo PJ (2009) Sulfur content at sulfide saturation in oxidized magmas. Geology 37: 415-418
Lowenstern JB (1995) Applications of silicate-melt inclusions to the study of magmatic volatiles. In:
Thompson JFH (ed) Magmas, Fluids, and Ore Deposits. Mineralogical Association of Canada Short Course 23, pp 71-99
Michael PJ, Graham DW (2015) The behavior and concentration of CO2 in the suboceanic mantle: Inferences from undegassed ocean ridge and ocean island basalts. Lithos 236-237: 338-351
Namiki A, Manga M (2008) Transition between fragmentation and permeable outgassing of low viscosity magmas. Journal of Volcanology and Geothermal Research 169: 48-60
Newman S, Lowenstern JB (2002) VolatileCalc: a silicate melt–H2O–CO2 solution model written in Visual Basic for excel. Computers & Geosciences 28: 597-604
Ochs FA (III), Lange RA (1999) The density of hydrous magmatic liquids. Science 283: 1314-1317
Oppenheimer J, Rust AC, Cashman KV, Sandnes B (2015) Gas migration regimes and outgassing in particle-rich suspensions. Frontiers in Physics 3, doi: 10.3389/fphy.2015.00060
Papale P (1999) Strain-induced magma fragmentation in explosive eruptions. Nature 397: 425-428
Parmigiani A, Faroughi S, Huber C, Bachmann O, Su Y (2016) Bubble accumulation and its role in the evolution of magma reservoirs in the upper crust. Nature 532: 492-495
Pioli L and 6 coauthors (2008) Explosive dynamics of violent Strombolian eruptions: the eruption of Parícutin Volcano 1943–1952 (Mexico). Earth and Planetary Science Letters 271: 359-368
Pistone M, Caricchi L, Ulmer P, Reusser E, Ardia P (2013) Rheology of volatile-bearing crystal mushes: mobilization vs. viscous death. Chemical Geology 345: 16-39
Plank T, Kelley KA, Zimmer MM, Hauri EH, Wallace PJ (2013) Why do mafic arc magmas contain ~4wt% water on average? Earth and Planetary Science Letters 364: 168-179
Rust AC, Cashman KV (2004) Permeability of vesicular silicic magma: inertial and hysteresis effects. Earth and Planetary Science Letters 228: 93-107
Scaillet B, Clemente B, Evans BW, Pichavant M (1998) Redox control of sulfur degassing in silicic magmas. Journal of Geophysical Research: Solid Earth 103: 23937-23949
Self S (2006) The effects and consequences of very large explosive volcanic eruptions. Philosophical Transactions of the Royal Society A 364: 2073-2097
Tait S, Jaupart C, Vergniolle S (1989) Pressure, gas content and eruption periodicity of a shallow, crystallising magma chamber. Earth and Planetary Science Letters 92: 107-123
Thomas N, Tait SR (1997) The dimensions of magmatic inclusions as a constraint on the physical mechanism of mixing. Journal of Volcanology and Geothermal Research 75: 167-178
Voight B and 5 coauthors (2010) Magma‐sponge hypothesis and stratovolcanoes: Case for a compressible reservoir and quasi‐steady deep influx at Soufrière Hills Volcano, Montserrat. Geophysical Research Letters 37, doi: 10.1029/2009GL041732
Wallace PJ (2001) Volcanic SO2 emissions and the abundance and distribution of exsolved gas in magma bodies. Journal of Volcanology and Geothermal Research 108: 85-106
Wallace PJ, Anderson AT, Davis AM (1999) Gradients in H2O, CO2, and exsolved gas in a large‐volume silicic magma system: Interpreting the record preserved in melt inclusions from the Bishop Tuff. Journal of Geophysical Research: Solid Earth 104: 20097-20122
Zajacz Z, Candela PA, Piccoli PM, Sanchez-Valle C (2012) The partitioning of sulfur and chlorine between andesite melts and magmatic volatiles and the exchange coefficients of major cations. Geochimica et Cosmochimica Acta 89: 81-101