Chapter 7

Volatiles in Magmas
Paul J. Wallace
University of Oregon, Eugene, OR, USA

Terry Plank
Lamont-Doherty Earth Observatory of Columbia University, Palisades, NY, USA

Marie Edmonds
Earth Sciences Department, University of Cambridge, Cambridge, UK

Erik H. Hauri
Carnegie Institution of Washington, Washington, DC, USA

Chapter Outline
5.1. Midocean Ridge Basalts 172
1. Definition and Significance of Magmatic Volatiles 164 5.2. Ocean Island Basalts 173
2. Solubility of Volatiles in Silicate Melts 164 5.3. Island Arc and Continental Margin (Subduction-
2.1. Water and Carbon Dioxide 164 Related) Basalts 174
2.2. Sulfur 166 5.4. Backarc Basin Basalts 175
2.3. Halogens (Cl and F) 167 5.5. Flood Basalts and Large Igneous Provinces 176
3. Approaches for Assessing Magmatic Volatile Contents 168 6. Volatile Abundances in Silicic Magmas 177
3.1. Experimental Phase Equilibria 168 6.1. Andesites 177
3.2. Analysis of Quenched Glass (Submarine Glasses, 6.2. Dacites and Rhyolites 177
Melt Inclusions, Matrix Glass) 169 6.3. Gradients in Volatile Concentration and Vapor
3.3. Thermodynamic Calculation Based on Mineral Saturation in Magmas 179
Compositions 170 6.4. Volcanic Gases: Total Volatile Budgets and
3.4. Fluid Inclusions 171 Climate Impacts of Eruptions 180
3.5. Saturation with Volatile-Rich Minerals and 7. Origin of Volatiles, Earth Degassing, and
Immiscible Liquids 171 Volatile Recycling by Subduction 180
4. Analytical Techniques for Measuring Volatile 8. Future Directions 182
Abundances in Silicate Glasses 172 Further Reading 182
5. Volatile Concentrations in Basaltic Magmas 172

GLOSSARY glass If a liquid is cooled very rapidly, it may harden to a glass and
not crystallize. Magma is a silicate liquid which, if cooled rapidly
exsolved gas Gas that has been released from solution in a magma. enough, forms glass.
Gas exsolution occurs when pressure is decreased, such as when immiscible Some types of liquids, such as molten iron sulfide, will
the magma moves toward the Earth’s surface. not dissolve into a silicate melt to form a homogeneous solution.
fugacity A gas is considered to be perfect if its behavior can be Such coexisting liquids are described as immiscible.
described by the ideal gas law: PV ¼ nRT, where P is the pressure, magma Natural silicate melt with or without suspended crystals and
V is the volume, n is the number of moles, R is the ideal gas bubbles.
constant, and T is the temperature. Most gases are not perfect, melt (glass) inclusion During crystallization of magma, many crys-
especially at high pressures or low temperatures. The fugacity is a tals grow imperfectly, trapping small blebs of silicate melt inside
kind of equivalent pressure that accounts for the deviation of a real the crystals. If the magma is erupted and cooled rapidly, then these
gas from ideal gas behavior. trapped inclusions of silicate melt become quenched to glass.

The Encyclopedia of Volcanoes. http://dx.doi.org/10.1016/B978-0-12-385938-9.00007-9 163

Copyright Ó 2015 Elsevier Inc. All rights reserved.
164 PART | I Origin and Transport of Magma

saturation A melt can be saturated with respect to certain chemical deeper within the Earth, most of the water and carbon di-
elements in three ways: it may contain bubbles of a gas or gas oxide are dissolved in the silicate melt portion of the
mixture (e.g., H2OeCO2), it may contain crystals rich in the magma, and they have important effects on magma crys-
element such as anhydrite (CaSO4, rich in S), or it may contain tallization, temperature, density, and viscosity. In the upper
droplets of immiscible liquid rich in the element such as drops of
mantle, where basaltic magmas are generated by partial
sulfide melt (rich in S).
melting, volatiles affect the composition of melts and their
solubility The maximum amount of a volatile species or component
that can be dissolved in a silicate melt under a given set of con- physical segregation from the residual mantle minerals.
ditions such as pressure, temperature, and melt composition. After H2O and CO2, the most abundant volatiles are sulfur,
volatile An element or compound, such as H2O or CO2, that forms a chlorine, and fluorine; sulfur is particularly important for
gas at a relatively low pressure and magmatic temperature. Vol- understanding the composition and fluxes of volcanic gas,
atiles can be dissolved in silicate melts, can occur as bubbles of and all three can be important in the formation of ore de-
exsolved gas, and can crystallize in minerals such as biotite and posits. Many other volatiles, such as He, N, Ar, Br, and I,
amphibole. The gas phase is also commonly described as vapor. can dissolve in natural silicate melts (magmas), but they are
Because its density increases with pressure, making it more generally much less abundant. The noble gases, for
liquid-like at crustal and mantle depths, it is also sometimes example, are usually present at concentration levels
referred to as a fluid phase.
<1 ppm (0.0001% by weight) in natural magmas. Despite
such low concentrations, the isotopic compositions of noble
1. DEFINITION AND SIGNIFICANCE OF gases have been studied extensively because of their value
in understanding large-scale degassing of the Earth and the
MAGMATIC VOLATILES
formation of the atmosphere.
In this chapter, we focus on the most abundant volatiles:
H2O, CO2, S, Cl, and F. We review (1) experimental data on
The observatory worker who has lived a quarter of a century with
the solubility of volatiles in natural silicate melts, (2)
Hawaiian lavas frothing in action, cannot fail to realize that gas
analytical and sampling techniques used for measuring the
chemistry is the heart of the volcano magma problem.
abundances of volatiles in magmas, (3) volatile abundances
T.A. Jaggar, 1940
in basaltic and silicic magmas typical of various tectonic
The above quotation from T.A. Jaggar, founder of the Ha- environments, and (4) Earth degassing and volatile recy-
waiian Volcano Observatory and an influential volcanolo- cling by subduction. We also discuss the evidence for
gist in the first half of the twentieth century, attests to the preeruptive vapor saturation, as this is important for un-
importance of gases in governing volcanic eruptive phe- derstanding the explosive eruptive behavior of volcanoes,
nomena. As early as the middle of the nineteenth century, it fluxes of volcanic gases, and the formation of ore deposits.
was recognized that gas plays a fundamental role in forcing
magma to the Earth’s surface and generating explosive
eruptions. Many different species of gas can dissolve in a 2. SOLUBILITY OF VOLATILES IN
molten silicate liquid in the same way that carbon dioxide SILICATE MELTS
can dissolve in water. Such gas species are referred to as
volatile components or magmatic volatiles because of their
2.1. Water and Carbon Dioxide
tendency to form gas bubbles at a relatively low pressure. Solubility refers to the maximum amount of a volatile
The amount of a volatile component that can dissolve in species or component that can be dissolved under a given
silicate melt increases with pressure. As magma ascends set of conditions, such as pressure, temperature, and melt
toward the Earth’s surface, pressure decreases, thereby composition. Laboratory measurements of the solubilities
decreasing the solubility of volatiles, and causing them to of H2O, CO2, S, and many other volatile components have
come out of solution to form bubbles. In addition, the been made for a number of different melt compositions. In
bubbles expand with a further decrease in pressure. At one such laboratory experiments, it is possible to control the
atmosphere pressure and typical magma temperatures, an parameters that affect solubility. As an example, a silicate
amount of exsolved H2O equivalent to 0.1% of total magma melt can be equilibrated with H2O vapor at high tempera-
mass is sufficient to produce a magmatic foam that has tures and a variety of pressures in order to determine the
>90% bubbles by volume. Tremendous expansion occurs solubility of water (Figure 7.1). Such melt is described as
near the surface as a result of such volatile exsolution, and “H2O saturated” because in the experiment, the melt co-
this provides the driving force for volcanic eruptions, exists with water vapor, though it should be noted that the
similar to the fountaining of champagne from an uncorked vapor phase is not pure because it contains at least a minor
bottle. amount of other components from the silicate melt. As seen
Water and carbon dioxide are the most important vol- in the experimental data, the solubility of H2O in silicate
atile components in natural magmas. At higher pressures, melts is strongly dependent on pressure. This pattern, in
Chapter | 7 Volatiles in Magmas 165

4 4
Basalt (1200°C)

3 Basalt (1200°C) 3 Rhyolite (850°C)

Pressure (kbar)

Pressure (kbar)
2 Rhyolite (850°C) 2
Nephelinite (1200°C)

1 1

0 0
0 2 4 6 8 10
0.0 0.2 0.4 0.6 0.8
H2O (wt%) CO 2 (wt%)
FIGURE 7.1 H2O solubility in basaltic and rhyolitic melts at typical FIGURE 7.2 Carbon dioxide solubility in basaltic and rhyolitic melts at
magmatic temperatures based on experimental data. (See review of typical magmatic temperatures. Also shown for comparison is the solu-
experimental solubility data in Baker and Alletti, 2012.) bility in a nephelenitic melt to show the strong increase in solubility in
alkali-rich and silica-poor melts. (Solubilities calculated using Iacono-
which the solubility increases with increasing pressure, is Marziano et al., 2012, for mafic melts and Newman and Lowenstern,
observed for many volatile components, and occurs 2002, for rhyolitic melt.)
because the gaseous form of a volatile component has a
larger molar volume than does the same component when it solubilities are much lower than for H2O. The amount of
is dissolved in a silicate melt. The solubility of H2O in CO2 that dissolves into both basaltic and rhyolitic melts is
silicate melts is also dependent on both melt composition 50e100 times less, by weight, than the solubility of H2O at
and temperature. The H2O solubility is greater in rhyolitic a comparable pressure and temperature. Carbon dioxide
melts than in basaltic melts, but much of this difference is dissolves in silicate melts in two different species, but in
actually due to temperature differences rather than to contrast to water, the speciation is controlled by the bulk
compositional effects (Figure 7.1). silicate melt composition. Thus, in silica-poor glasses
Early studies on the solubility of water in silicate melts (basalts, basanites, and nephelinites), CO2 is present as
recognized that there is a linear relationship between the structurally bound carbonate (CO2 3 ), whereas in silica-rich
partial pressure (or fugacity) of H2O and the square of the glasses (rhyolite), it occurs as CO2 molecules. Glasses with
mole fraction of dissolved water in the melt. Such a rela- intermediate silica contents (andesitic) contain both spe-
tionship is consistent with a solution mechanism in which cies. Experimental studies show that the solubility of CO2
water dissolves by forming OH groups that are structur- is similar in basaltic, andesitic, and rhyolitic melts, despite
ally bound to the aluminosilicate network of the melt. More the differences in speciation across this compositional
recent studies using infrared spectroscopy have shown that range. However, increases in alkali content strongly in-
dissolved water in silicate melts occurs as two different crease CO2 solubility such that nephelinite and phonolite
species: OH groups and H2O molecules. The relative magmas have much higher solubilities (Figure 7.2). In
proportions of the two species vary systematically with addition, in basaltic melts, the melt CaO content has a
total water concentration. At low total dissolved water strong positive effect on solubility.
concentrations, virtually all the water is present as OH In discussing solubility, it is important to distinguish
groups. As total dissolved water concentration increases, between saturation and undersaturation with a particular
the relative proportion of molecular H2O increases as well. volatile component. Solubility refers to the maximum
In rhyolitic melts, the proportions of OH groups and H2O amount of a volatile species or component that can be
molecules are equal at about 3 wt% total dissolved water, dissolved in silicate melt at a given set of conditions. In
whereas in basaltic melt, the proportions of the two species both natural and experimental silicate melts, however, the
are equal at about 3.5 wt% total water. The speciation is melt can have a concentration of a dissolved volatile
also dependent on temperature and quench rate, and this component that is less than the maximum possible. Such a
forms the basis of a geospeedometer, from which the melt is described as undersaturated with respect to the
cooling rates of hydrous rhyolitic glasses can be calculated volatile component of interest. As an example, laboratory
using measurements of OH groups and H2O molecules. experiments show that a rhyolitic melt can contain about
As with H2O, the solubility of CO2 in silicate melts is 6 wt% dissolved H2O at 2 kbar, but this does not mean that
strongly dependent on pressure (Figure 7.2). However, CO2 all natural rhyolitic melts that are at a 2-kbar pressure in the
166 PART | I Origin and Transport of Magma

Earth’s crust contain this much water. Instead, they could 1600
3 kbar
have any amount from 0 wt% to 6 wt%. A rhyolitic melt
containing less than the dissolved H2O needed to saturate
1200
the melt at a given temperature and pressure is thus “H2O

CO 2 (ppm)
2 kbar XvCO2 = 0.8
undersaturated.”
It is possible in laboratory solubility experiments, as 800
0.6
described above, to equilibrate a melt with a single gas
component such as H2O. However, in magmas, many vol- 1 kbar 0.4
400
atile components are present together. It is therefore
0.2
important to distinguish between the dissolved concentra-
tions of individual volatile components and their concen- 0
trations in the separate exsolved gas phase. If the sum of 0 1 2 3 4 5 6
the partial pressures of all dissolved volatile components in H 2 O (wt%)
a magma is equal to the local confining pressure, then the FIGURE 7.3 H2O and CO2 solubility (heavy solid curves) in vapor-
magma is saturated with bubbles of a multicomponent saturated basaltic melts at 1200  C and 1e3 kbar pressure based on
vapor phase containing H2O, CO2, S, and other minor gases experimental data and thermodynamic models. The curvature of the
such as Cl, F, and noble gases. Because volatile species saturation curve for a given pressure results from the nonlinear dependence
of H2O solubility on pressure (Figure 7.1). The light curves show the
have different solubilities, their relative abundances dis- composition (in mole fraction CO2) of the vapor phase in equilibrium with
solved in the melt will differ from their relative abundances the melt. (Diagram calculated using Newman and Lowenstern, 2002.)
in the gas phase.
For vapor-saturated silicate melts, the solubilities of
individual volatile components are dependent in part on sulfate-bearing phases and because dissolved S can occur
the abundances of other volatiles that are present in the in multiple valence (oxidation) states. The speciation of
system. This is because the amount of a given component dissolved S in silicate melts is related to the relative
that is dissolved in the melt is proportional to the partial magmatic oxygen fugacity (Figure 7.4). At relatively low
pressure of the volatile component in the coexisting oxygen fugacities, S is present in solution mainly in the
equilibrium vapor. As an example, consider a silicate melt reduced form sulfide (S2), whereas under more oxidizing
that is saturated with vapor consisting only of H2O and conditions, sulfate (S6þ) appears to be the dominant spe-
CO2. The sum of the partial pressures (PH2O and PCO2) of cies. For basaltic to andesitic magmas in the range of
these components in the vapor phase must equal the total
pressure in a vapor-saturated melt. Thus, at a constant total 1.0
pressure, an increase in PH2O, and hence dissolved H2O in
the melt, must result in a decrease in PCO2 and dissolved
CO2. On a diagram of dissolved H2O versus dissolved 0.8
CO2 in silicate melts, vapor saturation isobars, which
show the maximum dissolved H2O and CO2 at a given
Sulfate/total sulfur

total pressure, have negative slopes (Figure 7.3). However,

0.6
there are some indications that at pressures greater than
approximately 2e3 kbar, low to moderate amounts of
dissolved H2O may actually increase CO2 solubility
slightly. In many common igneous systems, the partial 0.4
pressures of other volatile components (e.g., Cl, F, and S)
are probably relatively small compared with those of H2O
and CO2, but this may not always be the case. A knowl- 0.2 } Fe-poor & Fe-free
glasses

edge of the preeruptive concentrations of H2O and CO2 in Albite

Basalt
a magma as measured in melt inclusions can thus poten- Andesite
tially constrain the pressure at which the magma was 0.0
stored before eruption, provided the magma is vapor -3 -2 -1 0 1 2 3 4 5
saturated. Oxygen fugacity (∆QFM)
FIGURE 7.4 Fraction of total sulfur that is present as sulfate (sulfate/total
2.2. Sulfur sulfur) as a function of oxygen fugacity, relative to the quartze
fayaliteemagnetite buffer curve (DQFM). The curves show the experi-
The behavior of S in silicate melts is complex because the mental relationship for basaltic and andesitic melts compared to Fe-poor
solubility of S is controlled by the stabilities of sulfide and silicic melts. (Figure is modified from Klimm et al., 2012.)
Chapter | 7 Volatiles in Magmas 167

oxygen fugacities from slightly above the fayalitee 3000

magnetiteequartz (FMQ) buffer to approximately 1.5
Sulfide melt saturation
log10 units above FMQ, both dissolved sulfide and sulfate 2500
are present in significant quantities (Figure 7.4). In Fe-poor Loihi Seamount

silicic melts, the region where both S species are present is 2000

S (ppm)
shifted to lower oxygen fugacities by about 1.5 log10 units
1500 East Pacific Rise Galapagos
(Figure 7.4). The maximum S that can be dissolved in
silicate melt is controlled by saturation of the melt with
1000
an S-bearing phase. At relatively low oxygen fugacities,
this phase is either an immiscible iron-rich sulfide liquid
500
(with as much as 10% oxygen), which occurs in high-
temperature basaltic melts, or crystalline pyrrhotite
0
(Fe1xS) in intermediate (andesitic) and silicic (rhyolitic) 8 10 12 14 16 18 20
melts. A CueFe sulfide mineral may also crystallize from
FeOT (wt%)
andesitic to rhyolitic melts. At higher oxygen fugacities,
the sulfate mineral anhydrite (CaSO4) crystallizes in in- FIGURE 7.6 Covariation of sulfur and total iron (FeOT) in submarine
MORB glasses from the Galapagos spreading center (Perfit et al., 1983)
termediate to silicic melt compositions. In high-pressure
and the East Pacific Rise (le Roux et al., 2006) compared to submarine
laboratory experiments, mafic melts at high temperatures basaltic glasses from an oceanic island volcano (Loihi seamount). The line
can be saturated with an immiscible sulfate melt, but evi- shows the S saturation limit in MORB melts. The saturation limit for Fe-
dence for this in natural samples has only rarely been poor Galapagos basalt is calculated at 1170  C and an oxygen fugacity that
found. Silicate melts at intermediate oxygen fugacities can is 1.5 log units more reduced than the nickelenickel oxide (NNO) buffer.
For Fe-rich Galapagos basalt, the S concentration is calculated at 1070  C
crystallize both pyrrhotite and anhydrite, an assemblage
and an oxygen fugacity that is 0.5 log units more reduced than NNO.
that was found, for example, in trachyandesite tephra from (Modified from Wallace and Edmonds, 2010, using additional data from
the 1982 eruption of El Chichón in Mexico. Experimental Dixon and Clague, 2001.)
solubility studies show that S is more soluble at high ox-
ygen fugacities, under which conditions the dissolved S is (Figure 7.6), which indicates that S2 dissolves largely by
present as sulfate, and anhydrite is the stable S-bearing complexing with Fe2þ. Sulfur solubility under both
mineral (Figure 7.5). The solubility of reduced S (S2) oxidizing and reducing conditions increases with temper-
increases with the Fe concentration in the silicate melt ature (Figure 7.5). Pressure has contrasting effects,
depending on the oxidation state. For oxidizing conditions,
the solubility of S increases with increasing pressure, but
Sulfur solubility in silicate melts for reducing conditions, where S2 is the dominant spe-
4000 cies, solubility decreased with increasing pressure. As a
CaSO saturated
4 result of the latter effect, a basaltic magma with low ox-
S (ppm) dissolved in melt

FeS saturated ygen fugacity formed in equilibrium with residual sulfide

3000
in the mantle becomes undersaturated in sulfide during
adiabatic ascent.
2000 4 kbar
2.3. Halogens (Cl and F)
1000
As with S, the solubility of Cl is complex because silicate
2 kbar melt can be saturated with an immiscible alkali chloride
melt (molten salt). If water is present, then the immiscible
0 alkali chloride melt will also contain dissolved water and
750 800 850 900 950 1000 is referred to as a hydrosaline melt. Chlorine solubility in
Temperature (°C) silicate melts is strongly dependent on silicate melt
FIGURE 7.5 Experimental determinations of sulfur solubility versus composition, and the solubility generally increases as
temperature for hydrous silica-rich melts at different pressures and oxygen the ratio (Na þ K)/Al in the silicate melt increases. A
fugacities. Curves are shown for both sulfide-saturated melts (oxygen silicate melt saturated with molten salt has the maximum
fugacity at the NNO buffer) and sulfate-saturated melts (oxygen fugacity dissolved Cl. The saturation concentration of Cl varies
at the MnO-Mn3O4 [MNO] buffer). At 2-kbar pressure, S is more soluble
with pressure, temperature, dissolved H2O concentration,
under oxidizing conditions (MNO) than under reducing conditions (NNO).
Solubility curve for sulfate-saturated melts (MNO buffer) at 4 kbar shows and silicate melt composition. For silicate melts that
that sulfate solubility increases with increasing pressure. (Figure is based are saturated with an H2O- and CO2-rich vapor phase
on experimental data of Luhr, 1990.) at middle to upper crustal pressures, maximum Cl
168 PART | I Origin and Transport of Magma

solubilities range from nearly 3000 ppm in rhyolitic melts content of magma requires that both dissolved and exsolved
to nearly 3 wt% in basaltic melts. Thus, Cl solubility in volatile abundances are known. Because the solubility of
silicate melts is somewhat intermediate between that of most volatile components is strongly pressure dependent,
H2O and that of CO2. In addition, at low pressures magmas exsolve gas during ascent to the surface. In order
(1.2 kbar atw800  C), a silicate melt can be saturated to understand the crystallization and physical properties of
with both hydrosaline melt and an H2O- and CO2-rich magma, as well as to understand eruptive behavior, it is
vapor phase. important to know the total volatile content of magma
A natural silicate melt may contain less dissolved Cl during its preeruptive storage within the Earth’s crust.
than the amount needed to saturate the melt with either Dissolved volatiles may be estimated in many ways, but
molten salt or hydrosaline melt. Many common magma exsolved volatiles are difficult to assess. The following
types do not contain Cl-rich mineral phases or evidence of discussion will focus primarily on dissolved volatiles; we
hydrosaline melt. By comparison with experimental Cl will discuss the relationship to exsolved volatiles in a
solubility studies, most magmas probably have insufficient subsequent section, and the reader is also referred to the
dissolved Cl to be saturated with hydrosaline melt. In some Chapter 45 on Volcanic, Magmatic and Hydrothermal
Cl-rich magmas, however, formation and separation of a Gases for additional information.
hydrosaline melt phase or alkali-chloride-rich vapor may
play an important role in the formation of magmatic-
hydrothermal ore deposits. For magmas that are saturated 3.1. Experimental Phase Equilibria
with an H2O- and CO2-rich vapor phase, some Cl partitions One of the most valuable methods for estimating dissolved
into the vapor. Experimental studies suggest that the con- volatile concentrations in magmas, particularly H2O and
centration of Cl in the vapor phase coexisting with melt CO2, is through the use of phase equilibrium experiments.
compositions ranging from basalt to rhyolite is about 1e8 The assemblage of minerals that crystallizes from a melt of
times greater than the concentration dissolved in the melt. a given bulk composition and the sequence and temperature
This results in fairly minimal loss of Cl from melt by of appearance of those minerals during cooling vary
degassing until very low (near-surface) pressures are depending on pressure and dissolved H2O concentration
reached, at which point large mass fractions of H2O are (Figure 7.7). By equilibrating a given rock composition at
exsolved and some Cl is lost.
The solubility and speciation of dissolved F in silicate
melts are complex and still not well understood, but Phase equilibria for Mount St Helens dacite
available data clearly show that solubility is strongly
dependent on silicate melt composition. In general, F is 3.0
highly soluble in silicate melts, similar to H2O. In granitic
Pressure (kbar)

composition melts, as much as 10 wt% F can be dissolved.

Basaltic and andesitic magmas typically have low dissolved
2.0
F, on the order of a few hundred parts per million or less.
op plag

Melt + Vapor
x

Rare melts, such as those in topaz-bearing rhyolitic

magmas, may contain as much as 5 wt% F, and ultra- ph
am Fe
-Ti
potassic mantle-derived magmas can contain as much as 1.0 ox
2 wt% F. Many granitic magmas have sufficient F to id
es
become saturated with fluorite. Because of its high solu-
bility, F is much more strongly partitioned into the liquid
0.0
phase than in the vapor phase, and therefore experiences 850 950 1050 1150
minimal degassing even upon eruption, in contrast to the
other volatiles described above. Temperature (°C)
FIGURE 7.7 Experimental phase equilibria for 1980 Mount St Helens
dacite as a function of temperature and pressure for water-saturated melts.
Because dissolved H2O increases with pressure (Figure 7.1), the water
3. APPROACHES FOR ASSESSING contents of the experimental melts increase along the vertical axis of the
MAGMATIC VOLATILE CONTENTS diagram. With increasing pressure (and dissolved H2O), the temperatures
at which anhydrous minerals (pyroxene, plagioclase, and FeeTi oxides)
All terrestrial magmas contain dissolved volatiles, the most begin to crystallize are significantly diminished. In contrast, increasing
pressure and dissolved H2O increases the stability of amphibole, which
abundant being H2O, CO2, S, Cl, and F. If the dissolved
therefore crystallizes at higher temperatures as pressure increases. Com-
concentrations of volatiles are sufficiently high, a magma parison with the actual observed mineral assemblage in the 1980 Mount St
may be vapor saturated such that the magma also contains Helens dacite indicates preeruptive temperature and pressure of 920  C
an exsolved vapor (gas) phase. Assessing the total volatile and approximately 2.2 kbar. (Modified from Rutherford et al., 1985.)
Chapter | 7 Volatiles in Magmas 169

various temperatures, pressures, and H2O concentrations, it assess total volatile concentration in the erupted magma,
is frequently possible to find a relatively narrow range of but the magma may have gained or lost bubbles before
conditions under which the observed phenocryst assem- eruption. Bulk analyses of low solubility volatiles such as
blage in the rock is in stable equilibrium. This technique is CO2 in samples of glass with trapped bubbles typically give
particularly suited to constraining the dissolved concen- greater concentrations than do direct microanalyses of the
trations of H2O during preeruptive crystallization and glass.
storage of magma within the Earth’s crust. Experimental Another important source of information on the dis-
studies have been applied with great success to a number of solved volatile concentrations of magmas is through the
volcanic eruptions, most notably, the eruptions of Katmai analysis of melt (glass) inclusions trapped inside crystals.
(1912), Parı́cutin (1943e1951), Mount St Helens (1980), When melts cool they crystallize, and the crystals often
El Chichón (1982), Mt Pinatubo (1991), Soufriere Hills grow imperfectly, causing small amounts of melt to be
(1995e2011), and Chaitén (2008). trapped inside the crystals. If the magma ascends, erupts,
In addition to its effect on phase stability, dissolved and cools rapidly, then these trapped melt inclusions
H2O also affects the composition of minerals that crys- quench to glass (Figure 7.8). Because the crystalline host
tallize from silicate melts. The most useful effect related for the inclusions is relatively rigid, they act as tiny
to magmatic H2O concentrations is the change in plagio-
clase composition that occurs with variations in dissolved
H2O. As the dissolved H2O concentration in a melt is
increased, the equilibrium plagioclase composition be- (A)
comes more calcic (anorthite-rich) and also crystallizes at
a lower temperature than in dry magmas. As will be
discussed below, anorthite-rich plagioclase phenocrysts
are common in subduction-related basaltic magmas,
providing evidence of relatively high concentrations of
H2O compared to basaltic magmas from other tectonic
environments.

3.2. Analysis of Quenched Glass

(Submarine Glasses, Melt Inclusions,
Matrix Glass)
Subaerially cooled volcanic rocks and tephra have lost
nearly all their original magmatic volatiles because of the (B)
low pressures of Earth’s atmosphere. In contrast, during
deep submarine eruptions, the confining pressure exerted
by the overlying water may prevent exsolution and loss of
many volatile species. Such pressures can be as great as
500e600 bars. Submarine-erupted magmas typically form
glassy rinds where hot magma is rapidly quenched against
cold seawater. Volatiles may be retained dissolved within
such glassy basalt, which can be analyzed for volatiles
directly. H2O is relatively soluble in basaltic melt and as
much as 2 wt% may be retained in deep sea basaltic glass
without eruptive loss. In contrast, for relatively insoluble
CO2, the pressures on the seafloor are commonly insuffi-
cient to prevent exsolution. This results in small vesicles in
FIGURE 7.8 Glass (melt) inclusions in crystals from volcanic rocks.
quenched submarine glasses, usually <1% by volume, (A) Basaltic glass inclusion with a vapor bubble in an olivine pheno-
except in submarine island arc volcanoes and some backarc cryst from Mauna Loa, Hawaii. The bubble likely formed during
regions, where initial melts are more H2O rich and hence cooling after the inclusion was trapped because the melt in the inclusion
exsolve substantial gas at seafloor pressures. For typical, contracts more on cooling than does the surrounding olivine host crystal
and because the crystallization of olivine along the inclusionehost
vesicle-poor glasses, the small size of the vesicles and their
interface causes additional volume loss from the inclusion. (B) Rhyo-
relatively low volumetric abundance result in isolated litic glass inclusions in quartz phenocryst from the Huckleberry Ridge
(noninterconnected) bubbles trapped in the quenched glass. Tuff, Yellowstone. A 100-mm scale bar (black) is shown in both
Bulk analyses of these glassy rocks can therefore be used to photographs.
170 PART | I Origin and Transport of Magma

pressure vessels and keep the trapped melt at high pres- CO2 values, reflecting trapping at different pressures from
sure, even though the bulk magma decompresses to sur- melts that have variably degassed before entrapment.
face pressure during eruption. For this reason, trapped Basaltic melt inclusions from arc volcanoes also commonly
melt (glass) inclusions can retain a record of dissolved show a range of H2O values that can also be caused by
volatiles at the time of trapping. However, caution must variable preentrapment degassing as well as postentrap-
be exercised when interpreting melt inclusion data ment diffusive H loss. Sulfur concentrations can also be
because the trapped volatile concentrations in the melt variable, reflecting partial degassing before trapping or loss
can be modified after entrapment by several processes. of S to sulfides. Various methods have been used to assess
One such process is the formation of a shrinkage bubble the original H2O content of magmas from such data sets,
during postentrapment cooling, which causes the loss of but the simplest approach is to select the maximum H2O
low solubility CO2 from the trapped melt into the vapor concentration or H2O/K2O ratio recorded within a melt
bubble. Another effect is the potential loss or gain of H2O inclusion population as representative of the parental
in an inclusion due to diffusion of H through the mineral magma.
host. Laboratory experiments have found that diffusive Quenched glassy scoria or pumice that formed during
reequilibration of H2O in melt inclusions can occur subaerial eruptions can also be analyzed for volatiles.
through mineral hosts on the order of hours to days Generally, such material has extensively degassed during
(olivine) and days to weeks (quartz) at magmatic tem- eruption and quenching and can also be affected by post-
peratures. Finally, melt inclusions may in some instances emplacement, low temperature hydration, so measured
leak volatiles during explosive eruption due to cracks in volatile contents do not reflect those before eruption. An-
the host mineral, in which case textural evidence such as alyses of relatively fresh (nonhydrated) scoria glass or
cracks or a large vapor bubble or multiple bubbles are pumice can, nevertheless, be useful for interpreting
typically present. This is especially problematic for degassing and eruptive processes.
inclusions in minerals with relatively good cleavages,
such as hornblende and plagioclase. 3.3. Thermodynamic Calculation Based on
Melt inclusions in phenocrysts are trapped during
crystal growth and thus sample the liquid from which the
Mineral Compositions
crystals grew at the time of entrapment. Inclusions that are One method that has commonly been used to estimate
trapped at different times during crystallization may pre- volatile concentrations in magmas is the application of
serve a record of changing magma compositions due to thermodynamic calculations. For example, reactions
processes such as crystal fractionation, magma mixing, and involving hydrous minerals and their anhydrous breakdown
degassing. Glassy reentrants and hourglass inclusions, products can be used along with thermodynamic properties
which are connected to the outside of the host crystal by a of the phases to estimate water concentrations. One such
melt-filled neck, reveal the composition of melt just before reaction is the equilibrium involving

KFe3 AlSi3 O10 ðOHÞ2 þ ½ O2 ¼ KAlSi3 O8 þ H2 O þ Fe3 O4

biotite þ gas ¼ feldspar þ gas þ magnetite

eruption. Pressure-dependent concentrations of volatiles in This equation can be used to calculate the water
melt inclusions, hourglass inclusions, and reentrants can fugacity at the time of crystallization for any volcanic rock
provide a record of decompression, degassing, and crys- containing biotite, feldspar, and magnetite, provided the
tallization during magma ascent. These processes can make oxygen fugacity can also be estimated.
it difficult to assess the original volatile content of basaltic The most commonly used thermodynamic method is a
magmas before their ascent into the crust. Suites of melt hygrometer based on the equilibrium exchange of anorthite
inclusions from a single eruption often have a wide range of and albite components between plagioclase and silicate melt:

CaAl2 Si2 O8 þ NaAlSi3 O8 ¼ CaAl2 Si2 O8 þ NaAlSi3 O8

plagioclase melt melt plagioclase
Chapter | 7 Volatiles in Magmas 171

Increases in dissolved H2O in silicate melt shift the during crystallization, but the multicomponent nature of
equilibrium to the left, favoring an increase in anorthite such inclusions (H2OeCO2eNaCl) makes it difficult to
content in plagioclase. Calibration of this relationship infer the original dissolved volatile concentrations in the
using experimental data for a wide range of melt com- coexisting silicate melt.
positions from basalt to rhyolite yields a relationship from
which one can calculate magmatic H2O concentrations 3.5. Saturation with Volatile-Rich Minerals
with an uncertainty of about 0.25 wt% if the tempera-
ture and composition of plagioclase and coexisting melt
and Immiscible Liquids
are known. Hygrometers based on amphibole composi- The most common volatile-rich minerals and immiscible
tions have also been experimentally calibrated, but have liquids that occur in magmas are sulfide and sulfate min-
larger uncertainties than those of the plagioclase-melt erals and immiscible sulfide (FeeSeO) liquid. The pres-
hygrometer. ence of one or more of these phases in a quenched volcanic
A relatively new approach to estimating preeruptive rock indicates preeruptive saturation with an S-rich phase.
H2O contents of magmas is to study the H2O contents of By comparison with experimental solubility data, this
phenocrysts that grew in the magma during ascent. Small makes it possible to infer a preeruptive dissolved S con-
amounts of H2O can be contained in the structure of silicate centration. There are both thermodynamic and empirical
minerals (olivine, clinopyroxene, orthopyroxene, and feld- solution models available that can be used to calculate the
spar), and the equilibrium partitioning of H2O between dissolved S content for either sulfide- or sulfate-saturated
melts and minerals has been experimentally studied for melts for a wide range of bulk compositions and oxygen
basaltic compositions over a wide range of pressures. These fugacity conditions.
studies have determined the mineralemelt partition coef- For volatile species whose solubility in a melt is
ficient for H2O: controlled by the presence of a volatile-rich mineral or
melt, such as a sulfide phase, complex changes in solubility
mineral=melt H2 OðmineralÞ
D H2 O ¼ : can occur during magma cooling and differentiation. As an
H2 OðmeltÞ example, it has been observed in quenched glassy sub-
The H2O partition coefficients are dependent on pressure marine pillow rinds from midocean ridge basalts (MORBs)
and mineral composition, particularly for pyroxenes, that only a small amount (<1.5% by mass) of the total S in
where higher Al contents result in higher partition co- the magma is present as quenched sulfide globules. This
efficients. The use of these partition coefficients permits an observation may seem somewhat surprising when one
estimate of the magmatic H2O content if the H2O con- considers that the strong temperature dependence of S
centrations of coexisting minerals can be accurately solubility should cause considerable precipitation of
determined. immiscible sulfide liquid during cooling. However, the
observed low quenched sulfide globule abundance is
consistent with thermodynamic and experimental assess-
3.4. Fluid Inclusions
ment of the effects of differentiation on S solubility. Dur-
The presence of primary fluid inclusions in minerals from ing the cooling and crystallization of midocean ridge
volcanic or plutonic rocks provides evidence that crystal- basaltic magmas, residual liquids become increasingly Fe-
lizing melts were vapor and/or brine saturated. However, enriched due to fractional crystallization of silicate phases.
it can be difficult to distinguish between primary fluid The increase in Fe content increases the S solubility in the
inclusions, trapped during igneous crystallization, and melt by an amount that is greater than the decrease in S
secondary inclusions formed by lower temperature hydro- solubility brought about by cooling. This results in
thermal processes. Some volcanic phenocrysts from differentiated FeeTi basalts that have a higher dissolved S
basaltic rocks contain dense CO2-rich, primary fluid in- than do the parental (unfractionated) magmas. This effect,
clusions, indicating that the basaltic magma was CO2- combined with the increase in residual melt S brought
saturated during crystallization. Such inclusions can be about by crystallization, keeps midocean ridge basaltic
used to estimate the pressure of inclusion entrapment. A magmas finely balanced just at the S saturation point,
knowledge of this, coupled with experimentally determined causing the total mass of S in immiscible sulfide liquid
CO2 solubilities in basaltic melt, can be used to estimate the to be small throughout the course of differentiation. In
dissolved CO2 concentration of the basaltic magma at contrast, Kilauean tholeiitic magmas do not show an Fe-
depths. Phenocrysts in more evolved volcanic rocks enrichment trend because plagioclase crystallization
(dacites, rhyolites) may contain inclusions containing one occurs later than in MORBs. In the absence of Fe-
or more of the following phases: fluid, vapor, hydrosaline enrichment, if the melt is saturated with immiscible sul-
melt, and daughter crystals. The presence of these in- fide liquid, then the solubility of S will decrease as the
clusions also can be taken as evidence for vapor saturation temperature decreases.
172 PART | I Origin and Transport of Magma

4. ANALYTICAL TECHNIQUES FOR approximately 70e160 ppm H2O, based on studies of wa-
MEASURING VOLATILE ABUNDANCES ter in MORB glasses. If these concentrations occur
throughout the entire mantle, then an amount of water
IN SILICATE GLASSES
equivalent to approximately 10e50% of the mass of water
A large number of analytical techniques have been used to in the ocean could be contained in the upper mantle. The
measure volatile concentrations in silicate glasses. These mantle source regions for enriched MORBs (E-MORB)
techniques fall into two main categoriesdbulk extraction probably contain 300e500 ppm H2O.
techniques, in which a bulk rock or glass sample is crushed During partial melting, water and other volatiles
and analyzed, and microanalytical techniques, in which a generally behave as incompatible elementsdelements that
tiny portion of glass is analyzed. Bulk extraction techniques are enriched in the melt phasedbecause they are not stoi-
can make it difficult to distinguish between the volatiles chiometric constituents of the major mantle silicates. Even
that are actually dissolved in quenched magmatic liquid S, which is contained in sulfides, behaves as an incom-
(e.g., glassy submarine pillow rinds) and the volatiles that patible element because of the low modal abundance of
may be present in unopened vesicles or networks of cracks sulfide. Thus, basaltic magmas act as agents for outgassing
formed during magma solidification. Furthermore, it is the Earth’s mantle by carrying volatiles to the surface.
critical to select material for analysis that is free of alter-
ation. Microanalytical techniques offer a high spatial res-
olution, making it possible to analyze a tiny portion of 5.1. Midocean Ridge Basalts
alteration-free glass. The most commonly used microana- Midocean ridge basaltic magmas (N-MORB) have rela-
lytical techniques are secondary ion mass spectrometry tively low H2O contents, typically <0.4 wt%, but
(SIMS; H2O, CO2, S, Cl, and F), Fourier transform infrared frequently as low as 0.1e0.2 wt% (Figure 7.9). Enriched
spectroscopy (FTIR; H2O, CO2), Raman spectroscopy MORB magmas (E-MORB), which form a volumetrically
(H2O), and electron microprobe analysis (S, Cl, and F). less abundant component of the oceanic crust, have sys-
Technical advances, improved calibrations, and increasing tematically higher concentrations, as great as 1.5 wt%.
availability of FTIR, SIMS, and Raman in the last 15 years Water concentrations from comagmatic suites of MORB
have led to a dramatic increase in the number of studies of glasses commonly show positive covariation with nonvol-
volatiles in melt inclusions and submarine glasses as well atile, incompatible trace elements such as K2O and Ce. In
as experimental studies of volatile solubilities and solution fact, the H2O/Ce ratio is remarkably uniform in MORB
mechanisms in silicate melts and nominally anhydrous glasses (and also ocean island basalt glasses), generally
minerals. falling within values of 100e300. This uniformity indicates
that H2O behaves as an incompatible element during partial

5. VOLATILE CONCENTRATIONS IN 2.5

BASALTIC MAGMAS Back-arc basin

As an introduction to discussing volatiles in basaltic 2.0

magmas, we first briefly review the abundance and storage
of volatiles in the Earth’s mantle, from which basaltic Samoa
H2O (wt%)

magmas are generated by partial melting. Volatiles in the 1.5

Earth’s mantle can reside in a number of potential minerals.
Recent experimental studies have demonstrated that sig- 1.0
nificant quantities of water can be present as structurally North Arch
bound OH in the nominally anhydrous minerals olivine,
pyroxene, and garnet. Chlorine solubility in these nomi- 0.5 Loihi

nally anhydrous minerals is quite low, but F solubility is

MORB
relatively high, sufficient for F to be primarily stored in 0.0
olivine and orthopyroxene in the upper mantle. Water, Cl, 0.0 0.5 1.0 1.5 2.0 2.5 3.0
and F could also be stored in hydrous minerals such as
K2O (wt%)
amphibole, phlogopite, and apatite. Carbon is present pri-
marily in magnesian carbonate minerals, low degree car- FIGURE 7.9 Variations in H2O and K2O in submarine basaltic glasses
from different tectonic environments. Data are shown for MORBs, backarc
bonatitic melts, or in the reduced form as diamond,
basin basalts, and ocean island glasses from Hawaii (Loihi and North
graphite, or carbide. Sulfur is bound in sulfide minerals. Arch) and Samoa. (Data are from le Roux et al. (2006), Taylor and
Estimates of the water content of the depleted upper mantle Martinez (2003), Dixon and Clague (2001), Dixon et al. (1997), and
that is the source region for normal MORBs (N-MORB) are Workman et al. (2006).)
Chapter | 7 Volatiles in Magmas 173

melting of the mantle and fractional crystallization in ridge basaltic glasses (MORB; see Figure 7.9) but extends
crustal magma reservoirs and that significant H2O is neither to higher values. Typical values for tholeiitic basalts are
lost by degassing nor gained by assimilation in magma 0.3e0.7 wt% H2O (Hawaii), 0.6e1.0 wt% (Galapagos),
reservoirs. Rare andesites and dacites along midocean and 0.5e0.7 wt% (Iceland). Alkalic compositions gener-
ridges can contain 1.0e2.5 wt% H2O, and their geochem- ally have higher values, and some well-studied examples
ical characteristics indicate that assimilation of altered include alkali basalts from Samoa (0.6e1.5 wt% H2O),
oceanic crust is important along with fractional mildly alkalic basalts from the 2010 Eyjafjallajökull
crystallization. eruption in Iceland (0.8e1.0 wt%), basanites recovered
Dissolved CO2 concentrations in quenched MORB from Loihi Seamount, Hawaii (0.9e1.0 wt%), and basan-
glasses vary from 50 to nearly 400 ppm. Most MORB ites and nephelinites recovered from the submarine North
glasses contain small vapor-filled bubbles as well, indi- Arch volcanic field, Hawaii (0.8e1.4 wt%). The system-
cating vapor saturation of the melt at the time of eruption atics between incompatible trace elements and H2O con-
and quenching. Equilibrium thermodynamic calculations, tents in some of the tholeiites and alkalic compositions
using measured H2O and CO2 contents and experimental indicate higher H2O concentrations in their mantle source
solubility data, suggest that the coexisting vapor phase regions compared to MORB. Sulfur concentrations in
would be 93e100 mol% CO2. Actual measurements of the ocean island basalts are commonly similar to those in
vapor composition in MORB vesicles are limited, but most MORB, suggesting they are saturated with immiscible
have >90 mol% CO2. Because MORB basaltic magmas are FeeSeO liquid before eruption (Fig. 7.6). Some ocean
saturated with a nearly pure CO2 vapor, the dissolved CO2 island basaltic magmas contain as much as 3000 ppm dis-
measurements can be used to estimate a pressure of solved S but are not highly enriched in FeO, in contrast to
quenching. Such calculations indicate that many MORB S-rich MORB magmas, which are FeO-rich. This contrast
melts were supersaturated with CO2, that is, they contain is consistent with higher oxygen fugacities for the S-rich
more CO2 than should be dissolved given the pressure ocean island basaltic magmas. The higher oxygen fugac-
equivalent to their depth of eruption on the seafloor. This ities decrease the stability of immiscible FeeSeO liquid,
supersaturation is caused by slow diffusion of CO2 in the and a higher dissolved S concentration is required before
basaltic melt compared to the timescale of ascent, eruption, the basaltic melt is saturated. Chlorine concentrations in
and quenching. ocean island basaltic magmas are quite variable, from low
Sulfur concentrations in MORB glasses are typically values similar to uncontaminated MORB to much higher
between about 800 and 1400 ppm, but may reach as high as values. Some higher Cl values have been interpreted as the
2500 ppm in FeeTi-rich basalts. Sulfur contents show a result of the presence of deeply recycled crustal compo-
strong positive correlation with the FeO content of the nents, which are enriched in Cl by hydrothermal alteration,
melt, consistent with experimental solubility relations in the mantle plume sources of hotspots. However, Cl/K
(Figure 7.6). Most MORB glasses contain small amounts of ratios in some ocean island basalts are similar to those in
quenched immiscible FeeSeO liquid, indicating that the MORB, suggesting that recycled ocean crust in ocean is-
melts were S-saturated at the time of eruption. Primitive, land basalt mantle sources may not be uniformly Cl-
uncontaminated MORB magmas contain 10e50 ppm Cl, enriched. Fluorine concentrations in ocean island basaltic
but Cl is particularly susceptible to the effects of assimi- magmas are mostly between 100 and 1000 ppm and show
lation of wall rocks that have been hydrothermally altered considerable overlap with the values found in MORB
by seawater, and values up to several hundred parts per glasses. In general, F concentrations increase with differ-
million Cl are commonly found. For example, highly entiation, as indicated by K2O contents, and higher values
differentiated FeeTi rich basalts contain as much as occur in more alkalic compositions. The F/Nd ratio (Nd has
1100 ppm Cl, an amount that greatly exceeds what would a similar incompatibility as F during melting and crystal-
be expected during closed-system fractional crystallization lization) is nearly constant in MORB and ocean island
of basaltic magma. Fluorine contents of MORB glasses are basalts.
low, generally about 100e600 ppm. Fluorine contents in The best understood oceanic island volcano, and prob-
suites of comagmatic glasses show F increasing with ably the most thoroughly studied volcano on Earth, is
increasing fractionation, in accordance with enrichment in Kilauea on the island of Hawaii. The dissolved H2O con-
residual liquids during closed-system fractionation. centrations in a variety of basaltic melt inclusions and
submarine glasses from Kilauea are shown in Figure 7.10.
The variations in H2O are controlled by five main factors:
5.2. Ocean Island Basalts
(1) heterogeneity in mantle-derived magma compositions,
Dissolved H2O in melt inclusions and submarine basaltic (2) fractional crystallization, (3) variable degassing before
glasses from ocean island volcanoes range from about 0.2 entrapment of melt inclusions, (4) mixing of surface
to 1.5 wt%, a range that overlaps with normal midocean degassed and undegassed magmas at depths as a result of
174 PART | I Origin and Transport of Magma

1.0 part low concentrations of H2O (0.1e0.85 wt%) and S

Puna Ridge glasses
(0.02e0.15 wt%), have been erupted on the deep subma-
Melt inclusions
0.8
rine flanks of Kilauea. The submarine basalts erupted and
Fractional crystallization were sampled from depths at which hydrostatic pressure
would have prevented significant coeruptive degassing of
H O (wt%)

0.6 H2O and therefore the low concentrations of H2O have

heterogeneity
been interpreted as reflecting low-pressure degassing prior

Primary
to submarine eruption at higher pressures. Such low pres-
2

0.4
Mixing

sure degassing could have occurred by drainback and

Degassing
ing
Mix
0.2
mixing prior to magma being transported down the rift. The
shallow mixing interpretation is borne out by recent ob-
Approx. 1 bar solubility servations from the 2008e2013 summit eruptions at
0 Halemaumau, which show this process at work: melt in-
0 10 15
MgO (wt%) clusion data from this period show that degassed magma
FIGURE 7.10 H2O contents of olivine-hosted melt inclusions in tephra sank down the conduit and mixed with undegassed magma
erupted from the summit of Kilauea and submarine glasses from the Puna at a depth of approximately 1 km.
Ridge. Lines show the effects of fractional crystallization, magma mixing The compositions of primitive melts at Kilauea influ-
(typically between more evolved degassed and primitive undegassed ence the style of volcanic activity at the surface. The
melts), and degassing or diffusive loss of H from the melt inclusions.
magmas that feed the high fountain eruptions that typically
Magma mixing may be due to drain back after fountaining episodes or
vertical convection in the magma storage system beneath the summit. It is occur just outside the caldera have higher H2O and CO2 and
likely that a degree of compositional heterogeneity exists in primary melts are more primitive. In contrast, the magmas that erupt from
(with MgO >15 wt%). The melt inclusion compositions have been cor- the summit magma reservoir and feed effusive and lava
rected for the effects of postentrapment crystallization. (Figure is modified lake eruptions have lower volatile contents and are more
from Edmonds et al., in press.)
evolved, having mixed extensively with summit-stored
magmas. There are broad changes in parental melt com-
drainback, and (5) H loss from melt inclusions by diffusion positions at Kilauea that occur over 50- to 100-year time-
through the olivine host. scales and are thought to be caused by the changes in the
The interpretations of the H2O data fit with our general degree of partial melting coupled to small-scale mantle
understanding of degassing at Kilauea, which is inferred from source heterogeneities.
gas monitoring studies to occur mainly in two stages that are
related to subsurface storage and transport of magma. New 5.3. Island Arc and Continental Margin
batches of mantle-derived magma that ascend beneath
(Subduction-Related) Basalts
Kilauea are vapor saturated with a gas rich in CO2. The sol-
ubility of CO2 at the pressures of the summit magma reservoir It was first suggested in the early 1960s, before the advent
is low, and the first stage of degassing occurs as most of the of the modern concept of plate tectonics, that subduction of
CO2 is lost through noneruptive degassing, generating a altered oceanic rocks beneath the Aleutian arc was
persistent CO2-rich plume at the summit. The amount of H2O responsible for recycling water into the mantle source re-
that is lost with the CO2 in the first stage depends on the gion of arc magmas. Early work on melt inclusions in arc
amount of CO2 and the pressure, but is generally small, basalts during the 1970s used the deficit from 100% on the
because H2O is highly soluble in the melt. The second stage sum of the oxides measured by electron microprobe to
of degassing occurs when CO2-depleted magma from the estimate H2O concentrations. Such sum deficits were not
summit reservoir is transported laterally through the rift very precise, but they turned out to provide an accurate
system in dikes, at depths of generally >1e2 km, and then view of the water-rich nature of arc magmas. FTIR and
erupts, degassing the more soluble volatile components and SIMS ion microprobe techniques now provide a much
generating a gas plume rich in H2O and SO2. higher precision on H2O analyses. As a result, there has
As mentioned above, the volatile contents of magmas at been an explosion of data, both from melt inclusions and
depths may be modified by drainback of surface-degassed experiments, on the volatile contents of basaltic magmas
magma. Drainback was well documented for the 1959 parental to arc volcanoes because of interest in geochem-
Kilauea Iki eruption: magma that degassed during eruption ical recycling at subduction zones. The results of these
collected into a surface lava lake and then drained back into studies show that mafic arc magmas typically contain
the vent. Magma that has degassed in a lava lake and then 2e6 wt% H2O (Figure 7.11). These values are much higher
drained back down into the conduit can mix with unerupted than the H2O contents of MORBs and ocean island basalts,
magma stored at depths to yield partially degassed hybrid supporting the interpretation that water is recycled into the
magmas. Degassed hybrid magmas, with variable and in mantle by subduction, where it fluxes the overlying
Chapter | 7 Volatiles in Magmas 175

Mexico Although arc magmatism is dominantly calc-alkaline in

Cascades
2500 composition, some arcs contain a spectrum of basaltic
Shasta high-Mg andesite
magma types that extends to tholeiitic magmas, which show
Arenal
2000 Irazu
an increase in FeOT content during differentiation. Detailed
Stromboli
Fuego melt inclusion studies of Aleutian volcanoes show that the
calc-alkaline types have >2 wt% H2O, whereas the tholei-
Cl (ppm)

1500
itic types have <2 wt%. A combination of higher H2O,
Etna

Kamchatka
Ijen Cerro
which suppresses plagioclase crystallization, and higher
1000 Guat Negro oxygen fugacity, which enhances FeeTieoxide crystalli-
(BVF)
zation, is responsible for the calc-alkaline fractionation
500 BAB Marianas trend of decreasing FeOT. Both the Cascades and Trans-
Klyuchevskoy
Kilauea Central
Mexican Volcanic Belt also contain some basaltic magmas
N-MORB America
0 with lower H2O (1.5 wt%), and they likely form by
0 1 2 3 4 5 6 7 decompression melting of the mantle rather than by fluxing
H2O (wt%) with a hydrous component from the subducted slab.
FIGURE 7.11 Variations in H2O and Cl in basaltic melt inclusions from Arc basaltic magmas also contain significant amounts
various subduction-related arcs and specific arc volcanoes. Also shown for of dissolved CO2 (present as carbonate), with reported
comparison is a shaded field for the Lau and Marianas back-arc basins concentrations as high as 2500 ppm. Isotopic data for C in
(BAB). Guat (BVF) is for cinder cones behind the volcanic front in volcanic gases suggest that the CO2 is partly mantle derived
Guatamala. (Figure is modified from Metrich and Wallace, 2008.)
and partly recycled from subduction of carbonate-rich
marine sediments and organic carbon. Because melt in-
asthenospheric mantle wedge to generate hydrous basaltic clusions are trapped at crustal pressures after at least some
melts. CO2 has been degassed, it is likely that the initial CO2
Given the large number of volcanoes that have now contents of arc basaltic magmas are higher than the melt
been studied, average H2O values can be calculated for inclusion values, possibly as high as approximately 1 wt%.
individual arcs, and the results show a surprisingly narrow Olivine-hosted melt inclusions from basaltic arc magmas
range, from 3.2 wt% (for the Cascades) to 4.5 wt% (for the typically have 900e2500 ppm S, with some magmas ranging
Marianas), with a global average of 3.9  0.4 wt% H2O. up to 5000 ppm. These values are higher than for MORB
The narrow range and common average value for H2O in magmas of the same FeOT but are comparable to some ocean
each arc are unexpected because the concentrations of most island basalts. The high S contents of many arc basaltic
other incompatible elements in arc magmas, such as Nb or magmas require that they have a higher oxygen fugacity than
Ba, vary by orders of magnitude, and H2O might be ex- do MORB magmas and also require enrichment of S in the
pected to show similar large variations. The narrow range mantle wedge above subducting slabs. Sulfur isotope data for
of H2O concentrations could reflect a crustal control, in fumarolic gases, submarine basaltic glasses, and subaerial
which magmas commonly stall at depths of 2e10 km, whole rock samples have been interpreted to indicate recy-
allowing more deeply formed melt inclusions in olivine to cling of seawater sulfate. The Cl concentrations of basaltic
diffusively reequilibrate via H loss. Alternatively, the nar- arc magmas (mostly between 250 and 2500 ppm) and Cl/K
row range might reflect a mantle control, in which the ratios are also much higher than those of MORBs and ocean
addition of H2O to the mantle causes larger degrees of island basalts, consistent with recycling of seawater-derived
melting, thus diluting the concentration of H2O in partial Cl from subducted sediments or altered oceanic crust.
melts. Such a negative feedback could cause a narrow range
of H2O contents for mantle melts compared to the ranges
5.4. Backarc Basin Basalts
observed for incompatible trace elements.
Despite the similarities in averages and ranges in H2O Basaltic magmas from backarc basins show wide variations
contents from different arcs, when data for individual vol- in H2O concentration, ranging from values as low as those
canic centers within a given arc are examined, there are in the most H2O-poor N-MORB (0.1 wt%) to values as high
important variations that in some cases correlate with other as 2.3 wt% (Figure 7.9). Such variations reflect the range of
geochemical parameters. For example, in the Aleutians, bulk chemical composition of backarc basin basalts, which
Cascades, Central America, and Trans-Mexican Volcanic can vary from MORB-like to showing many similarities to
Belt, there are correlations between magmatic H2O con- island arc volcanics, reflecting their formation in spreading
centrations and the abundances of many incompatible trace centers that overlie subduction zones. Dissolved CO2 in
elements, indicating that transfer of a hydrous component submarine basaltic glass samples from the Mariana Trough
from the subducted plate upward into the mantle wedge also are between 100 and 200 ppm. Negative covariation be-
controls many trace element characteristics of arc magmas. tween H2O and CO2 in these glasses and pressure estimates
176 PART | I Origin and Transport of Magma

based on experimental solubility data suggest that the igneous provinces, LIPs) have occurred throughout the
samples were vapor saturated at their depth of eruption and history of the Earth and are likely linked to plumes of hot,
quenching, consistent with the presence of vesicles in the upwelling mantle. Data on the volatile contents of LIP
glassy pillow rims. Therefore, the bulk magmatic concen- basaltic magmas, which are dominantly tholeiitic in
trations of H2O and CO2 may be significantly greater than composition but include alkalic and more silicic compo-
the dissolved concentrations. Chlorine concentrations vary sitions in some cases, are still relatively limited. Water
from 80 ppm to nearly 1 wt% and show a positive corre- concentrations in submarine basaltic pillow rims from the
lation with H2O in comagmatic suites of glasses. Glasses Kerguelen and Ontong Java plateaus are relatively low
with the higher values (>2000 ppm) as well as many with (0.15e0.5 wt%), similar to values for midocean ridge and
values in the range from 200 to 2000 ppm also have high some ocean island basalts. Chlorine concentrations vary
Cl/K2O and therefore appear to have been affected by from low values similar to MORB (<50 ppm) to values as
assimilation of chlorine-rich altered oceanic crust at high as 1000e3000 ppm at Ontong Java, where basalts
shallow levels. By contrast, other glasses in the range from were likely affected by the assimilation of hydrothermally
80 to 2000 ppm with a very low Cl/K2O likely reflect altered wall rocks. Sulfur concentrations are similar
variably amounts of subduction-related enrichment from to or even lower than MORB values. Melt inclusions
the underlying slab into the mantle wedge. from several continental flood basalts have S concentra-
Some H2O-rich basaltic magmas erupted in backarc tions of about 800e2000 ppm and Cl concentrations of
basins or the submarine portions of island arcs have a high- 300e500 ppm. Much higher concentrations (as much as
enough water concentration that they vesiculate and degas 0.5 wt% S and 0.9 wt% Cl) are found in some alkalic
dissolved H2O even during eruption in very deep water. compositions or in magmas that have been extensively
These glasses are distinguished by their relatively high contaminated.
vesicularities (as much as 50% vesicles by volume). As The formation of many LIPs coincides in time with
expected, they also have anomalously low concentrations significant perturbations to the global carbon cycle and
of volatile components, such as S, that exsolve from mass extinctions, including the end-Permian Siberian
magma together with H2O. Traps, the Central Atlantic Magmatic Province at the Tri-
Measurements of H2O content in backarc basin glasses assiceJurassic boundary, and the Deccan Traps, erupted at
generally show negative correlations with TiO2, which, as the CretaceouseTertiary boundary. It has been recognized
an incompatible element during mantle melting, can be recently that the LIPs associated with the greatest impact on
used as a proxy for the extent of melting. The correlation the environment, such as the Siberian Traps, were those that
therefore shows that the degree of mantle melting is erupted through volatile-rich sediments, raising the possi-
controlled in part by how much H2O is added to the mantle bility that the volatile budget of such eruptions might
from the underlying subducted plate. The H2O concen- be augmented by the thermal decomposition of oil shales
trations estimated to be in the mantle beneath backarc and evaporites. This would have enhanced the emission
basins, based on these correlations, suggest that mantle of methane and organohalogens that could warm the
H2O contents increase toward the trench, as would be climate over long timescales and destroy the ozone layer,
predicted from the presence of a subducted slab that pro- respectively.
gressively dehydrates with depth. Backarc spreading seg- In addition to carbon-rich gases, the other climate-
ments that lie above the mantle with the highest estimated forcing gas of interest with regard to LIPs is sulfur diox-
H2O concentrations are at anomalously shallow water ide (SO2). The magmatic sulfur budget of LIP eruptions is
depths, probably because the high H2O increases mantle not well understood, owing to the fact that there are few
melt productivity and therefore results in a greater thick- data on preeruptive magmatic volatile abundances in these
ness of backarc basin crust. These interpretations have types of magmas. Some of the eruptions, however, have
been somewhat controversial, however, and an alternative yielded estimates of sulfur emissions. The eruption of the
possibility is that the negative correlations between H2O Siberian Traps is estimated to have released 6.8 x 107 Tg
and TiO2 are the result of mixing of magmas from H2O-- SO2 over less than 1 million years, and S isotope com-
poor mantle, similar to that beneath midocean ridges, positions of melt inclusions indicate that most of this
and subduction-modified mantle, like that which is beneath sulfur was mantle derived. Sulfur species in volcanic gases
island arcs. (SO2 and H2S) oxidize in the atmosphere to form sulfate
aerosol, which scatters, absorbs, and reflects incoming
5.5. Flood Basalts and Large Igneous solar radiation, causing short lived, but perhaps cata-
strophic, tropospheric cooling. The climate impact of
Provinces
these kinds of gas injections over such timescales are not
Huge outpourings of basaltic lava to form continental clear, but if the effects of recent, smaller eruptions can be
flood basalts and oceanic plateaus (known as large scaled up, it is likely that decades or longer of low
Chapter | 7 Volatiles in Magmas 177

temperatures occurred, which could have had a cata- Sulfur concentrations measured in melt inclusions in
strophic impact on life on Earth. andesites from subduction zone magmas are normally
1000 ppm, with typical values of <200 ppm in more silicic
melts. Sulfur has low solubility in Fe-poor oxidized rhyolitic
6. VOLATILE ABUNDANCES IN SILICIC melts, and as a consequence, most of the sulfur partitions into
MAGMAS hydrous vapor in the magma reservoir. Sulfur solubility is
strongly dependent on oxygen fugacity; for typical andesitic
6.1. Andesites magmas, this varies from slightly below the Ni-NiO (NNO)
Andesitic magmas typify convergent plate margins, and buffer reaction to about 2 log units more oxidized than NNO.
their volatiles play major roles in planetary degassing and Over this range, sulfur speciation changes from dominantly
differentiation. Most andesite magmas are complex hybrids sulfide to dominantly sulfate (Figure 7.4).
arising from mid to upper crustal differentiation of mafic There remains limited data on CO2 abundances in an-
magma, long-timescale residence of crystal-rich mush in desites, although it seems likely that there is a pervasive
upper crustal magma reservoirs, and repeated injections of flux of deeply derived CO2-rich vapor released from the
mafic magma into such reservoirs. Most andesites have volatile-rich mafic melts that pond and fractionate in the
30e45 vol% phenocrysts coexisting with a rhyolitic inter- lower arc crust. Melt inclusions in andesites are typically
stitial melt. The volatile budget is similarly complex: the poor in CO2, with abundances of <500 ppm. Volcanic gas
magma reservoirs likely coexist with a significant fraction measurements at arc andesitic volcanoes, however, suggest
of vapor, rich in H2O and S. It is likely that the recharging that CO2 fluxes are several times the flux of SO2 by mass,
mafic magmas supply volatiles during mixing and mingling which lends support to the idea that these gases are sourced
with the resident andesites, either by second boiling during in the lower to middle arc crust.
quenching and crystallization at the mingling interface, or Melts in andesites are typically rich in halogens, which
by changes in volatile solubility arising from differences in are derived ultimately from the subducting slab. Andesitic
temperature and oxygen fugacity between the two magmas. melt inclusions from the Izu volcanic arc contain
Standard methods for assessing volatile budgets are >800 ppm fluorine and up to 0.9 wt% chlorine and are
problematic for andesites: melt inclusions in plagioclase or similarly rich in other elements that behave in a volatile
pyroxene are often subject to the loss of volatiles through way during subduction such as lithium and boron. Rhyolitic
leakage and, if they are pristine, it remains unclear how melt inclusions in andesites from Augustine and Soufriere
they relate to the bulk magma and to the recharging Hills volcanoes contain up to 0.6 wt% Cl, which is close to
magmas. All the observations thus far, however, suggest the solubility limit for this melt composition. At Colima
that the melts are vapor saturated and H2O-rich at mid to Volcano, Mexico, melt inclusions contain up to 0.24 wt%
upper crustal depths. Rhyolitic melt inclusions in plagio- Cl. For particularly chlorine-rich melts, an immiscible
clase in andesite from Soufriere Hills contain up to hydrosaline melt (brine) phase may coexist with silicate
approximately 6 wt% H2O, suggesting that the melts are melt and vapor, which can sometimes be recognized as an
H2O-rich and that the magma resides in the midcrust additional phase in melt inclusions. Chlorine is relatively
(perhaps at depths >8 km, once the CO2 content of the melt soluble and remains largely dissolved at magma reservoir
is taken into account) prior to eruption. Andesitic to pressures, but it partitions into hydrous vapor at low pres-
rhyolitic melt inclusions hosted by plagioclase and pyrox- sure and is thus partially lost by degassing during magma
ene contain up to 3e4 wt% H2O in andesites from Colima ascent and eruption.
and Popocatepetl volcanoes, Mexico. Rhyolitic melt in-
clusions from Mt Hood in the Cascades contain up
6.2. Dacites and Rhyolites
to 5.5 wt% H2O.
Experimental volatile studies are difficult to design for a Dacitic and rhyolitic eruptions span a wide range of vol-
system characterized by disequilibrium on all scales. Some ume, volcano type (e.g., stratovolcanoes, dome complexes,
andesites contain volatile-bearing mineral phases, such as and large calderas), and tectonic association, including both
hornblende and apatite, which lend further constraints on arc and intraplate settings. Preeruptive H2O concentrations
the volatile budget and magma pathways through the crust, are often relatively high providing the energy for the
and some mineralemelt systems may be used as hygrom- powerful explosive eruptions that typify silicic magmas.
eters. From phase equilibrium studies, a minimum of Glassy melt inclusions are common in rapidly quenched
approximately 3 wt% dissolved H2O is necessary to stabi- phenocrysts from explosive silicic eruptions, particularly in
lize hornblende, a common phenocryst mineral in andes- quartz and sanidine, and analyses of the inclusions
ites, although hornblende stability is also strongly affected commonly yield values of 4e6 wt% H2O (Figure 7.12). In
by other factors, such as the Na2O content of the coexisting this way, the range of H2O contents in dacites and rhyolites
silicate melt. is remarkably similar to that in basalts and andesites.
178 PART | I Origin and Transport of Magma

500 pressure equals the pressure of crystallization. For

2 kbar 3 kbar
1 kbar magmas lacking vapor (vapor undersaturated), the actual
pressure of crystallization is greater than the vapor-
400 0.5 kbar
saturation pressure indicated by melt inclusions. Vapor-
N. Chile saturation pressures based on melt inclusion H2O and
C O2 (ppm)

300 CO2 for rhyolites vary from about 1 to 4 kbar, implying a

Pinatubo
range of upper crustal depths through which silicic volcanic
magmas differentiate and reside before eruption.
200 Detailed experimental phase equilibrium studies have
Satsuma- been performed on the 1912 rhyolite from Katmai, Alaska,
Iwojima Augustine
Toba the 1980 Mount St Helens dacite (Figure 7.7), the 1991 Mt
100
Pinatubo dacite, and the 2008 and 2011 rhyolitic eruptions
Katmai Krakatau of Chaitén and Cordón Caulle, respectively, in Chile. The
0 results indicate crystallization pressures of 0.5e1.0 kbar
0 1 2 3 4 5 6 7 (Katmai, Cordón Caulle), 1.2e2.0 kbar (Chaitén), and
Mazama
H2O (wt%) approximately 2.2 kbar (St Helens, Pinatubo) at tempera-
tures ranging from 780 to 920  C. The experimental data
FIGURE 7.12 Dissolved H2O and CO2 in melt inclusions from dacites
and rhyolites from subduction-related arcs. Diagonal lines show the vapor
require dissolved H2O contents for Pinatubo, Katmai and St
saturation pressures for rhyolitic melt from 0.5 to 3 kbar pressure. (From Helens that agree with melt inclusion studies. For the
Wallace, 2005). Mount St Helens dacite, the experimentally synthesized
mineral assemblage and mineral compositions match the
Complementary to the melt inclusion data for silicic natural assemblage when there is about 4.6 wt% H2O dis-
magmas are experimental phase equilibrium studies that solved in the melt. The experimental data for Mt Pinatubo
provide information about temperatures and depths of dacite indicates a higher H2O concentration of approxi-
equilibration before eruption and dissolved H2O concen- mately 6 wt%. The data for Katmai indicate lower H2O
trations. The analytical (melt inclusion) and experimental contents (approximately 2.5e4.5 wt%) given their lower
results give broadly consistent H2O values. equilibration pressures. The preeruptive magmas at Kat-
Water-rich melt inclusions are typical of low tempera- mai, Mount St Helens, and Pinatubo all appear to have
ture (700e800  C) rhyolites containing hydrous pheno- contained a vapor phase during crystallization, because the
crysts of biotite and amphibole from both arc and vapor-saturation pressures calculated from melt inclusion
intracontinental settings. Phase equilibrium data for hotter data and pressures based on experimental phase equilibria
(900  C) rhyolites on the Snake River Plain, which are are equal.
associated with the Yellowstone hotspot, require much One of the best studied rhyolitic eruptions in terms of
lower values (1.5 wt%), consistent with their lack of magmatic volatile abundances is the 0.76 Ma explosive,
hydrous phenocrysts. Dacitic and rhyolitic magmas from caldera-forming eruption that formed the >600-km3 Bishop
the catastrophic 1883 eruption of Krakatau, Indonesia, Tuff. Melt inclusions in Bishop quartz phenocrysts reveal
demonstrate, however, a higher amount of H2O in the melt systematic, preeruptive gradients in dissolved H2O and
(4 wt%) at similar temperatures. Interestingly, melt in- CO2 during crystallization. Melt inclusions from early
clusions in fayalitic olivines from Hekla volcano in Iceland erupted pumice fall deposits contain 5.3  0.4 wt% H2O
contain relatively high H2O (5e6 wt%), despite their and 60  40 ppm CO2, whereas those from the middle of the
occurrence in a hotspot setting. Peralkaline rhyolites eruption contain higher H2O (5.7  0.2 wt%) and CO2
(pantellerites) also contain high H2O (4e5 wt%), which as (120  60 ppm). Thus, early erupted magma contained less
with Hekla, are likely to be caused by the large amounts of dissolved H2O and CO2 than did magma from the middle of
fractional crystallization needed to produce rhyolite from the eruption. Compared with melt inclusions from early and
parental basaltic magma. middle Bishop Tuff samples, inclusions from late-erupted
Wide variations in dissolved CO2 are found in rhyolitic magma have much lower H2O (4.1  0.3 wt%), and
melt inclusions, from <30 ppm to about 1200 ppm, but higher and variable CO2 (150e1100 ppm), and the most
most values are <200 ppm. Based on the pressure- CO2-rich inclusions were trapped in the rims of crystals
dependent solubilities of H2O and CO2 in rhyolitic melts, shortly before eruption, following an episode of magma
the melt inclusion data can be used to compute the pressure mixing and partial quartz dissolution. The estimated pres-
at which the melt would be saturated with a vapor phase sures of crystallization (vapor-saturation pressures) are
(the vapor-saturation pressure) and thereby constrain lower for inclusions from the earliest erupted samples than
pressures of crystallization (inclusion formation). For in the samples from the middle of the eruption and highest
vapor-saturated magmas, the calculated vapor-saturation for the rims of late erupted quartz. This pattern is consistent
Chapter | 7 Volatiles in Magmas 179

with sequential tapping of a large zoned body of vapor- flows, or culminate with quiescent dome extrusion. Gradi-
saturated magma, because the first material erupted would ents in dissolved H2O in silicic magma bodies have more
likely come from the shallowest (lowest pressure) portion of recently been confirmed using ion probe and infrared
the magma body. spectroscopic techniques on melt inclusions in phenocrysts.
Dissolved S concentrations in silicic magmas tend to be In some cases, as at Mount St Helens, changes in eruptive
relatively low (200 ppm), and are frequently below the style from Plinian to subplinian and Vulcanian are reflected
minimum detection limit of common analytical techniques. in decreases in melt inclusion H2O concentrations. In
These low concentrations occur despite the fact that most others, like the Bishop Tuff, Plinian and pyroclastic flow
silicic magmas are saturated with a sulfide (pyrrhotite) and/ deposits have similar melt inclusion H2O values, probably
or sulfate phase (anhydrite). The low S solubility is caused because the different types of deposits were coeval rather
by a combination of low melt FeO content and the low than sequential. Even effusive (nonexplosive) eruptions of
temperature of silicic magmas. The former is an important silicic magma appear to involve magma with initially high
factor at a relatively low magmatic oxygen fugacity where dissolved H2O concentrations. The fact that such magmas
dissolved S is dominantly sulfide, whereas decreasing are erupted nonexplosively requires that water be lost prior
temperature significantly decreases both sulfide and sulfate to extrusion. This may occur when magma decompression
solubilities (Figure 7.5). induces sufficient vesicularity to form a magmatic foam
Fluorine and chlorine abundances are quite variable in that is highly permeable to gas loss or when magma rises so
metaluminous dacites and rhyolites (F, 200e1500 ppm; slowly that bubbles can segregate efficiently.
Cl, 600e2700 ppm). In the Bishop Tuff melt inclusions, F It has commonly been thought that magmas of inter-
varies from 160 to 460 ppm and Cl varies from 550 to mediate to silicic composition only become vapor saturated
approximately 800 ppm. Both elements are more abundant during shallow ascent and emplacement, during eruptive
in the most highly differentiated, early erupted Bishop decompression, or during advanced (pegmatitic) stages of
Tuff, whereas lower concentrations occur in the least plutonic crystallization. However, studies of eruptive dy-
differentiated magma that erupted later. Melt inclusions namics, CO2 and SO2 emissions, and melt compositional
from the 1991 Mt Pinatubo dacite contain approximately features all suggest that many magmas in subvolcanic
1000 ppm Cl. In contrast, Cl concentrations in melt in- reservoirs are saturated with a multicomponent vapor
clusions from Taupo Volcanic Field rhyolites (New Zea- phase. Volcanoes could be viewed primarily as gas vents. In
land) and Augustine volcano dacites (Alaska) range as particular, the small but finite solubility of CO2 in silicic
high as 2700 ppm, sufficient to cause the preeruptive magmas at crustal pressures and the relatively high CO2
magma to be saturated with an immiscible hydrosaline contents of mafic magmas that provide heat and mass to
melt in addition to a vapor phase. Compared to halogen create crustal magmatic systems could result in saturation
abundances in metaluminous magmas, values for per- with H2OeCO2 vapor during silicic magma formation as
alkaline rhyolites (pantellerites) are much higher, with as well as during ascent, emplacement, and crystallization.
much as 1.3 wt% Cl and 1.5 wt% F. These high concen- High-temperature fluid inclusions in phenocrysts from
trations reflect their derivation from parental mafic some silicic volcanic rocks provide direct evidence of
magmas created by small degrees of mantle melting, the preeruptive vapor saturation. Gradients in alkali and trace
large amount of fractional crystallization necessary to element abundances might result in part from upward flux
form the rhyolitic magmas (which causes Cl and F to in- of exsolved vapor in a large magma body.
crease like other incompatible trace elements), and to the Correlations between CO2 and trace elements in melt
effects of alkalis in increasing Cl and F solubilities in inclusions from the Bishop tuff indicate that the magma was
silicate melts. Peraluminous tin and topaz rhyolites of the vapor saturated during preeruptive crystallization. Quanti-
western United States and central Mexico also contain tative models of vapor exsolution during crystallization of
high halogen concentrations, with Cl reaching 0.5 wt% the Bishop magma are consistent with exsolved gas contents
and F as great as 5 wt%. varying from about 1 wt% in the deeper regions of the
magma body to nearly 6 wt% near the top. Thus, the data
6.3. Gradients in Volatile Concentration and models suggest that there was a gradient in exsolved
gas, with mass fractions of gas increasing upward toward
and Vapor Saturation in Magmas the roof of the magma body. Because the exsolved gas phase
Gradients in dissolved volatiles, especially H2O, within would be H2O-rich, such large exsolved gas contents imply
individual silicic magma bodies have been inferred from that the early erupted Bishop magma had a bulk (dis-
(1) variations in phenocryst assemblages and compositions; solved þ exsolved) H2O content of approximately 10 wt%.
(2) variations in whole-rock abundances of F and Cl; and However, these interpretations assume that gas remained in
(3) the observation that many volcanic eruptions evolve the magma during most of its evolution. The presence of
with time from high-energy Plinian columns to pyroclastic significant mass fractions of exsolved gas in Bishop magma
180 PART | I Origin and Transport of Magma

is consistent with H2O and CO2 dissolved in variably between CO2 and SO2 in the volcanic gas, typically using
degassed rhyolitic obsidians at Mono Craters, California. infrared spectroscopy and electrochemical sensors, and
Similar mass fractions of preeruptive exsolved gas have combining it with the SO2 flux measured using UV spec-
been inferred for the magma bodies of the 1883 Krakatau, troscopy as described above, to yield an estimate of CO2
1982 El Chichón and 1991 Mount Pinatubo eruptions, and flux. It has been shown that the sustained flux of CO2 from
many other smaller silicic eruptions, by assuming that all volcanoes such as Kilauea and Mount Etna is very large, at
SO2 released (measured by remote sensing techniques) was times >20 kilotons per day (kt/day), but more generally
stored in the erupted volume of magma. <7 kt/day. The explosive eruption of Eyjafjallajökull in
Iceland in 2010 was associated with the emission of
6.4. Volcanic Gases: Total Volatile Budgets approximately 150 kt/day CO2, sustained for several days.
These amounts of CO2 are still very small when compared to
and Climate Impacts of Eruptions
the time-averaged CO2 emissions from anthropogenic
An important constraint on the volatile budgets of volcanic sources (which is around 100e200 times larger). For
eruptions comes from measurements of gas fluxes vented example, the integrated daily output of CO2 from automo-
directly to the atmosphere, which represent the integrated biles in a single metropolitan area dWashington DCdis
sum of degassing over the magma column beneath the vol- approximately equal to the daily global CO2 output from all
cano. Gas flux measurements are also important for evalu- volcanoes worldwide.
ating the climate and environmental impacts of eruptions. The magnitude of these CO2 emissions, when the
For magmas that are vapor saturated at or below the magma supply rate is known, places some constraints on
depths at which crystallization occurs, melt inclusions in the CO2 contents of primary melts in these settings, which
phenocrysts record only a fraction of the volatiles that were probably lie within the range 0.5e1.0 wt%. For andesitic
originally present in the melt before vapor saturation. arc volcanoes, where volatiles are sourced from complex
Volcanic gas flux measurements therefore provide impor- hybrid magmas that are likely continually recharged at
tant additional information about the total volatile budgets depths, the volcanic CO2 flux at the surface cannot be easily
of volcanic eruptions that is complementary to data from related back to a primary melt, but the overall magnitude of
melt inclusion studies. For intermediate and silicic CO2 emission suggests that parental arc magmas may also
magmas, and for relatively oxidized basaltic magmas in arc have CO2 contents in the 0.5e1.0 wt% range.
settings, the concentrations of S in melt inclusions are far As described in the section above for eruptions involved
too low to account for the total amount of S degassed in the formation of LIPs, volcanic gases have environ-
during eruptions, implying that much of the preeruptive S mental and climate impacts. Sulfur-rich gases are converted
budget is partitioned into a magmatic vapor phase rather to sulfate aerosol rapidly in the atmosphere, where they act
than dissolved in silicate melt. In order to characterize S in to scatter, absorb, and reflect incoming solar radiation.
volcanic systems, measurements of the fluxes of SO2 and Satellite-based spectroscopic measurements are the pri-
H2S gases at the surface are therefore required. For volatile mary means of observing this conversion and quantifying
components with a very low solubility, such as CO2, the it. Large aerosol clouds arising from eruptions in the tropics
problem of evaluating the degassing budget is even more can circumnavigate the Earth very rapidly, within days,
severe: magmas may become saturated with a CO2-rich causing global climatic effects. Eruptions that occur at high
vapor phase at depths of tens of kilometers in the crust, and latitudes, however, have only local or regional effects. This
perhaps even in the upper mantle in some cases. is mainly due to the circulation pattern of air in the atmo-
Gas flux measurements from volcanoes are made using sphere, in which air is drawn from the equator and then
spectroscopic techniques that collect ultraviolet (UV) ra- flows out to high latitudes. Additionally, there is relatively
diation, utilizing the strong absorption signature of SO2 in low solar flux at high latitudes, which means that the effect
this region of the electromagnetic spectrum. For larger of scattering by aerosol is limited.
eruptions, the gas emissions might be very large and
injected to high altitudes in the stratosphere. In this case,
satellite-based spectrometers are more appropriate for 7. ORIGIN OF VOLATILES, EARTH
capturing the full mass of SO2 in the volcanic gas cloud, DEGASSING, AND VOLATILE
again utilizing absorption in the UV (e.g., the Ozone
RECYCLING BY SUBDUCTION
Mapping Instrument), and also in the infrared (e.g., the
Infrared Atmospheric Sounder Interferometer). Studies of inert gases such as helium, argon, and neon are
Determination of carbon dioxide (CO2) fluxes at persis- consistent with the theory that early in its history, the Earth
tently active volcanoes has become more commonplace, and was covered by a deep magma ocean, caused by conversion
this is of importance for quantifying the carbon output from into heat of some of the gravitational potential energy of
volcanoes. The method involves measuring the ratio accreting material. Inert volatiles would have become
Chapter | 7 Volatiles in Magmas 181

predominantly degassed and concentrated in the atmo- reservoirs mainly reflect the geological processes that
sphere during the Earth’s formation. Such an early atmo- partition water into these reservoirs. Water initially
sphere could have been largely lost from the Earth as a released from magmas at igneous temperatures is eventu-
result of solar activity or violent impacts. Significant ally redistributed by formation of new low-temperature
amounts of reactive volatiles such as H2O could have dis- minerals in the reservoirs of altered rocks, seawater, fresh
solved in the magma ocean. Subsequent crystallization of water, and ice. Isotopic fractionation during this redistri-
the magma ocean and formation of minerals poor in H2O bution causes the oxygen and hydrogen isotopic composi-
would promote degassing by concentrating the reactive tion of seawater to differ from that dissolved in parental
volatiles in a diminishing residue of liquid, thereby leading basaltic magmas; the balance is made up mainly by the
to the formation, rise, and escape of bubbles of gas. isotopically light water bound in altered rocks. There is no
It is conceptually important that present degassing rates isotopic evidence for a significant source of surface water
of primordial, nonradiogenic inert gas from midocean ridge beyond that which degasses from basaltic magmas. How-
magmatism (based on the 3He rate and ratios of other inert ever, the D/H ratio of seawater entering subduction zones is
gases to 3He), when multiplied by the age of the earth, yield considerably higher, by approximately 7%, compared with
much less total nonradiogenic neon and argon than is pre- upper mantle water. If the H2O content and D/H ratio of the
sent in the earth’s atmosphere. Thus, it is inferred that the mantle are both at steady state, as suggested by an
rate of degassing from the earth either was much greater in approximately constant ocean mass through geologic time,
the past or that significant inert gas has been separately this requires that H2O delivered to the mantle sources of arc
added to the outermost earth by meteoritic infall. The iso- magmas must be preferentially enriched in deuterium
topic composition of nonradiogenic inert gases that is lost compared with MORB. This is generally supported by the
from the earth is noticeably different from that in air, and D/H ratios measured in arc melt inclusions.
intermediate between that of inert gases from the sun and The isotopic composition of carbon in parental basaltic
from meteorites. The discrepancy can therefore be recon- magmas and various terrestrial reservoirs is not certain.
ciled with either a mixture of sources for the earth’s inert There may be a significant isotopic shift upon degassing,
gases or preferential loss from the earth’s atmosphere of and this makes it difficult to assess the isotopic composition
less massive isotopes. of carbon in parental basaltic magmas. The predominant
Subduction provides a mechanism by which atmo- recognized reservoir of terrestrial carbon is in sedimentary
spheric volatiles, contained in sediments and altered limestones and dolomites, and it differs isotopically from
oceanic crust, are recycled back into the mantle. The sed- that which degasses from basaltic magmas at igneous
iments and altered crust are relatively rich in the two major temperatures. Photosynthesis yields reduced carbon in
reactive volatiles: carbon dioxide stored as carbonates and biological tissues that is relatively poor in 13C, and this
organic carbon, and water, both as pore water and as water affects the isotopic composition of carbon in surficial res-
bound in hydrous minerals such as clays. Seawater-derived ervoirs. Existing knowledge permits the amount and iso-
chlorine is also added to the oceanic crust during its topic composition of buried organic matter (mainly reduced
alteration prior to subduction. The pore water is buoyant, C in shales and coals) to balance the amount and isotopic
and increasing pressure during subduction collapses pores composition of carbon in carbonate rocks yielding a sedi-
and forces water out of deeply buried sediments and rocks. mentary average that is consistent with igneous gas.
Most of the hydrous minerals break down to anhydrous However, significant additions of isotopically distinctive
minerals and release water vapor, which is also buoyant, as carbon to the Earth’s early atmosphere during the final
the subducting rock heats up. Some deep storage of water is stages of accretion, as well as storage of carbon in Earth’s
possible, however, in view of experimental studies that core, are possible. Furthermore, it is possible that the iso-
show that some dense hydrous magnesian silicates are topic composition of carbon now released from magmas
stable to very high pressures and temperatures. Also, small reflects a recycling steady state, and it may have differed in
concentrations of water occur as structurally bound OH in the remote past, reflecting a smaller proportion of recycled
nominally water-free minerals such as olivine and pyrox- carbon. The sum of these processes has led to an H/C ratio
ene. Subduction-related basaltic magmas are notably more in the Earth’s exosphere that is higher than that in its
H2O-rich than midocean ridge and ocean island basaltic interior, and significantly higher than that of chondritic
magmas, and such magmas may return most or all the meteorites.
subducted water to the surface. The downgoing view of carbon dioxide is illuminated
The isotopic composition of water is unaffected by by the fact that at fairly shallow mantle depths the car-
radioactive decay (except for bomb tritium) and has prob- bonate mineral dolomite becomes stable in a typical
ably been negligibly affected by preferential escape from assemblage of mantle silicates. Thus, a deep residence of
Earth of hydrogen compared to more massive deuterium. mineralogically bound CO2 is feasible. Under the condi-
Ratios of deuterium to hydrogen (D/H) in various water tions of pressure and temperature where dolomite is stable
182 PART | I Origin and Transport of Magma

in the mantle, CO2 is quite soluble in silicate melts and can eruption processes. In this context, petrologic techniques
flux partial melting at comparatively low temperatures. The for estimating the bulk (dissolved þ exsolved) volatiles in
upgoing view of ascending carbonate-rich magmas is magma bodies need to be further developed and refined to
marked by the expected release of CO2 gas at shallow upper better understand volcanic gas fluxes, low pressure
mantle pressures. Plausibly this pressure-sensitive release degassing, and the relationship between volatiles and
of CO2 gas is partially responsible for the explosive nature eruption dynamics. The extent to which volatile species
of rare magmas bringing diamonds and fragments of the leak from partially enclosed inclusions or diffusively
mantle to the surface. reequilibrate through crystals is potentially among the
Despite the measured differences in the isotopic fastest chronometers in geology, clocking the minutes of
composition of H2O and CO2 in various terrestrial reser- magma ascent prior to eruption. Although the interpretation
voirs, the isotopic composition of hydrogen, carbon, and of melt inclusion data is frequently complex due to effects
chlorinedas well as nitrogendare all similar to the range of magma mixing, crystallization, and volatile leakage, the
of isotopic compositions measured in carbonaceous chon- ability to sample melt inclusions from different strati-
drites. This striking isotopic similarity is convincing evi- graphic levels of an eruptive deposit makes it possible to
dence that the terrestrial budgets of these volatile elements get an unprecedented look at the workings of magma
were sourced from the late addition, after magma ocean bodies.
solidification, of a mass of carbonaceous chondrite equiv- It is increasingly being recognized that magmatic vol-
alent to approximately 2% of the present mass of the Earth. atiles, including volatile trace metals, make important
Yearly amounts of reactive volatiles H2O and Cl that are contributions to magmatic-hydrothermal ore deposits.
subducted are in the range of rough estimates of what is Future analytical developments that make it possible to
returned to the surface by midocean ridge and arc mag- measure the stable and radiogenic isotope ratios of melt
matism. If there were no surficial return of subducted H2O inclusions will significantly add to our knowledge of
and Cl, then the formation, alteration, and subduction of magmatic inputs to ore-forming systems. Another impor-
oceanic crust would comprise a net drain on the water and tant area of future research will be the relationship between
salt in the oceans. The geological record and understanding volatile abundances in volcanic rocks and their plutonic
of tectonism reveal that continental crust and ocean water equivalents. The goal of such work will be to gain a better
have existed for most of earth’s history; subduction has not understanding of the relationship between plutonic and
caused the oceans to dry up. Thus, water lost to altered volcanic rocks, and in particular, whether volcanic erup-
oceanic crust is probably largely returned to the oceans via tions tap the volatile-enriched upper portions of larger
dewatering of subducted crust and subduction-related crustal magma reservoirs, leaving behind less volatile-rich,
magmatism. uneruptable material that eventually solidifies as a pluton.

8. FUTURE DIRECTIONS FURTHER READING

The study of magmatic volatiles lies at the interface of Baker, D.R., Alletti, M., 2012. Fluid saturation and volatile partitioning
between melts and hydrous fluids in crustal magmatic systems: the
igneous petrology and volcanology. Much of our present
contribution of experimental measurements and solubility models.
state of knowledge has been derived relatively recently due
Earth-Science Reviews 114, 298e324.
to major advances in techniques of microanalysis.
Baker, D.R., Moretti, R., 2011. Modeling the solubility of sulfur in
Increasingly, it is possible to measure the isotopic compo- magmas: a 50-year old geochemical challenge. Reviews in Mineralogy
sition of volatile and nonvolatile species in silicate melt and Geochemistry, Mineralogical Society of America 73, 167e213.
inclusions, as well as trace metals in trapped fluid or Dixon, J.E., Clague, D.A., Wallace, P., Poreda, R., 1997. Volatiles in al-
vapor þ melt inclusions in volcanic phenocrysts. Combined kali basalts from the North Arch Volcanic field, Hawaii: extensive
analyses of volatile species, trace elements, and isotope degassing of deep submarine-erupted alkalic series lavas. Journal of
ratios (e.g., H, O, S, C, He, and Sr) will contribute to a Petrology 38, 911e939.
better understanding of volatile recycling in the Earth, Dixon, J.E., Clague, D.A., 2001. Volatiles in basaltic glasses from Loihi
melting processes, magmatic differentiation, and magma Seamount, Hawaii: evidence for a relatively dry plume component.
degassing. Journal of Petrology 42, 627e654.
Edmonds M, Sides I, Maclennan J., 2015. The role of volatiles in magma
The pressure-dependent solubilities of the major vola-
transport and eruption at Kilauea Volcano, Hawaii. In: Hawaiian Vol-
tiles H2O and CO2 in silicate melts make it possible to use
canoes: From Source to Surface, Rebecca Carey, Valerie Cayol,
melt inclusions as geobarometers. Combined with major, Michael Poland, Dominique Weis (Eds.), AGU Monograph, 323e350.
trace, and isotopic data, and a detailed knowledge of Klimm, K., Kohn, S.C., Botcharnikov, R.E., 2012. The dissolution
timeevolumeecompositional variations in eruptive de- mechanism of sulphur in hydrous silicate melts. II: solubility and
posits, melt inclusions can be used to place constraints on speciation of sulphur in hydrous silicate melts as a function of fo2.
magma chamber configurations, and magmatic and Chemical Geology 322e323, 250e267.
Chapter | 7 Volatiles in Magmas 183

Le Roux, P.J., Shirey, S.B., Hauri, E.H., Perfit, M.R., Bender, J.F., 2006. Metrich, N., Wallace, P., 2008. Volatile abundances in basaltic
The effects of variable sources, processes and contaminants on the magmas and their degassing paths tracked by melt inclusions.
composition of northern EPR MORB (8e10 N and 12e14 N): evi- Minerals, Inclusions, and Volcanic Processes Reviews in Miner-
dence for volatiles (H2O, CO2, S) and halogens (F, Cl). Earth and alogy and Geochemistry, Mineralogical Society of America 69,
Planetary Science Letters 251, 209e231. 363e402.
Luhr, J.F., 1990. Experimental phase relations of water- and sulfur- Rutherford, M.J., Sigurdsson, H., Carey, S., Davis, A., 1985. The May 18,
saturated arc magmas and the 1982 eruptions of El Chichón Vol- 1980, eruption of Mount St Helens: melt composition and experi-
cano. Journal of Petrology 31, 1071e1114. mental phase equilibria. Journal of Geophysical Research 90,
Iacono-Marziano, G., Morizet, Y., Le Trong, E., Gaillard, F., 2012. New 2929e2947.
experimental data and semi-empirical parameterization of H2OeCO2 Taylor, B., Martinez, F., 2003. Back-arc basin basalt systematics. Earth
solubility in mafic melts. Geochimica et Cosmochimica Acta 97, 1e23. and Planetary Science Letters 210, 481e497.
Newman, S., Lowenstern, J.B., 2002. VolatileCalc: a silicate Wallace, P., 2005. Volatiles in subduction zone magmas: concentrations
melteH2OeCO2 solution model written in Visual Basic for excel. and fluxes based on melt inclusion and volcanic gas data. Journal of
Computers and Geosciences 28, 597e604. Volcanology and Geothermal Research 140, 217e240.
Perfit, M.R., Fornari, D.J., Malahoff, A., Embley, R., 1983. Geochemical Wallace, P.J., Edmonds, M., 2011. The sulfur budget in magmas: evidence
studies of abyssal lavas recovered by DSRV Alvin from eastern from melt inclusions, submarine glasses, and volcanic gas emissions.
Galapagos Rift, Inca Transform, and Ecuador Rift 3. Trace element Reviews in Mineralogy and Geochemistry, Mineralogical Society of
abundances and petrogenesis. Journal of Geophysical Research 88, America 73, 215e246.
10551e10572. Workman, R.K., Hauri, E., Hart, S.R., Wang, J., Blusztajn, J., 2006.
Plank, T., Kelley, K., Zimmer, M., Hauri, E.H., Wallace, P.J., 2013. Why Volatile and trace elements in basaltic glasses from Samoa: impli-
do mafic arc magmas contain w4 wt% water on average? Earth and cations for water distribution in the mantle. Earth and Planetary
Planetary Science Letters 364, 168e179. Science Letters 241, 932e951.