Mineralium Deposita (2001) 36: 490±502

DOI 10.1007/s001260100185

A RT I C L E

Jacob B. Lowenstern

Carbon dioxide in magmas and implications

for hydrothermal systems

Received: 15 September 1999 / Accepted: 12 April 2001 / Published online: 6 July 2001
Ó Springer-Verlag 2001

Abstract This review focuses on the solubility, origin, presence of CO2 can induce immiscibility both within the
abundance, and degassing of carbon dioxide (CO2) in magmatic volatile phase and in hydrothermal systems.
magma±hydrothermal systems, with applications for Because some metals, including gold, can be more vol-
those workers interested in intrusion-related deposits of atile in vapor phases than coexisting liquids, the pres-
gold and other metals. The solubility of CO2 increases ence of CO2 may indirectly aid the process of
with pressure and magma alkalinity. Its solubility is low metallogenesis by inducing phase separation.
relative to that of H2O, so that ¯uids exsolved deep in
the crust tend to have high CO2/H2O compared with
¯uids evolved closer to the surface. Similarly, CO2/H2O
will typically decrease during progressive decompres- Introduction
sion- or crystallization-induced degassing. The temper-
ature dependence of solubility is a function of the After H2O, carbon dioxide is the second most common
speciation of CO2, which dissolves in molecular form in gas in volcanic exhalations (White and Waring 1963;
rhyolites (retrograde temperature solubility), but exists Symonds et al. 1994). In gases exsolved from basaltic
as dissolved carbonate groups in basalts (prograde). magmas, CO2 sometimes exceeds water in concentration
Magnesite and dolomite are stable under a relatively (e.g., Gerlach 1980; Giggenbach 1997), presumably due
wide range of mantle conditions, but melt just above the to its greater abundance in some mantle source regions.
solidus, thereby contributing CO2 to mantle magmas. The presence of CO2 in magmatic systems, combined
Graphite, diamond, and a free CO2-bearing ¯uid may be with its relatively low solubility, results in early exsolu-
the primary carbon-bearing phases in other mantle tion of vapors that are rich in CO2 (Holloway 1976).
source regions. Growing evidence suggests that most Once exsolved, this low density vapor phase has im-
CO2 is contributed to arc magmas via recycling of sub- portant implications for the ascent and eruption of CO2-
ducted oceanic crust and its overlying sediment blanket. bearing magmas (Papale and Polacci 1999). Many
Additional carbon can be added to magmas during workers have noted evidence for CO2-rich ¯uids in a
magma±wallrock interactions in the crust. Studies of variety of intrusion-related ore deposits, particularly
¯uid and melt inclusions from intrusive and extrusive those associated with Au and lithophile enrichment
igneous rocks yield ample evidence that many magmas (Newberry et al. 1995; Goldfarb et al. 1998; Thompson
are vapor saturated as deep as the mid crust (10±15 km) et al. 1999). In such systems, CO2 may be derived from
and that CO2 is an appreciable part of the exsolved the magma, and even from a mantle source, although
vapor. Such is the case in both basaltic and some silicic there are other means for CO2 to enter ore-forming
magmas. Under most conditions, the presence of a CO2- systems such as decarbonation reactions in wallrock and
bearing vapor does not hinder, and in fact may promote, dissolution of CO2 and carbonate minerals from crustal
the ascent and eruption of the host magma. Carbonic hydrothermal systems.
¯uids are poorly miscible with aqueous ¯uids, particu- This review focuses on CO2 in magmas, providing
larly at high temperature and low pressure, so that the an introduction to the behavior of CO2-bearing vapors
within magmas, and as they enter hydrothermal sys-
tems. First, it outlines current knowledge about the
solubility of carbon species in silicate melts, then con-
J.B. Lowenstern
U.S. Geological Survey, Mail Stop 910, tinues with a summary of carbon in the mantle and
345 Middle®eld Road, Menlo Park, CA 94025, USA mantle melts, and proceeds to discuss CO2 in inter-
E-mail: Jlwnstrn@[Link] mediate and silicic magmas. Next, it concentrates on
 491
2+
the composition of exsolved magmatic vapors and the gens). She found that the e€ect of Ca was most pro-
e€ect of CO2 on magma properties, ascent, eruption, nounced, followed by K+, Na+, Mg2+, and Fe2+ in
and intrusion. Finally, it provides a brief summary of descending order of importance. In contrast, CO2 does
CO2 in hydrothermal and geothermal systems, and not react with the melt as it dissolves in molecular form,
speculates on the importance of CO2 in some ore- so that there should be little variation in CO2 solubility
forming systems. This synthesis of current knowledge in felsic magmas. Moreover, on a molar basis, CO2
about CO2, and its role in magmas, attempts to provide solubility changes little with melt composition along the
economic geologists with a background that can be tholeiitic basalt±rhyolite join (Blank and Brooker 1994;
used when interpreting the role of CO2 in intrusion- see also Fig. 1).
related deposits.

Solubility as a function of pressure and temperature

The solubility of CO2 in silicate melts
In all studied magma types, the solubility of CO2 in-
CO2 dissolution creases with pressure at a nearly linear rate (Fig. 1).
Temperature e€ects are not so simple, as they are de-
To understand the systematics of CO2 behavior in ig- pendent on the mechanism of incorporation of the CO2
neous systems, one must ®rst consider how CO2 dis- in the melt (Stolper et al. 1987). In rhyolitic magmas, the
solves in magmas and the variables that in¯uence its solubility of CO2 clearly decreases as temperature in-
solubility (see Blank and Brooker 1994 for an excellent creases (see Fig. 4 from Fogel and Rutherford 1990).
review). Experimental studies in the 1960s and 1970s This is consistent with the typical inverse correlation
recognized that CO2 solubility in natural rock melts is between molecular gas solubility and temperature,
about an order of magnitude less than that of water whether in room-temperature liquids, hydrothermal
(e.g., Wyllie and Tuttle 1959; Kadik et al. 1972). In the ¯uids, or rock melts. Exceptions occur when the gases
late 1970s and 1980s, experimentalists were able to use dissolve by interacting with the melt to form charged
vibrational spectroscopy to determine the speciation of complexes (Fig. 2). For example, the solubility of sulfate
volatiles within silicate melts and obtain more quanti- (which exists only in oxidized magmas) increases with
tative solubility estimates. Mysen et al. (1975) used 14C temperature (Baker and Rutherford 1996), as does car-
beta-track mapping techniques to study the solubility bonate solubility in albite melts (Stolper et al. 1987) and
of CO2 in a variety of natural melt compositions and basaltic melts (Mysen et al. 1975). In general, though,
found that it dissolves both as carbonate groups and the temperature e€ect on solubility is very small com-
molecular CO2. Later, Fine and Stolper (1985) explored pared with the e€ects of melt composition and pressure
the incorporation of CO2 in sodium aluminosilicate (Blank and Brooker 1994).
glasses, and found that it dissolved both as molecular
CO2 and Na±CO3 ionic complexes, with the latter in-
creasing along with Na2O on the Na2O±SiO2 join. In
subsequent studies, researchers have found that CO2
dissolves solely as carbonate groups in basaltic melts
including ma®c alkaline magmas such as basanites,
leucitites, and nephelinites (Blank and Brooker 1994).
In contrast, solely molecular CO2 is found in rhyolites
(Fogel and Rutherford 1990). Intermediate melt com-
positions (andesites) have both species present (Mysen
et al. 1975; King et al. 1996), as is the case for evolved
silica-undersaturated magmas such as phonolites
(Blank and Brooker 1994).
Because CO2 reacts with the melt as it dissolves to
form carbonate groups in ma®c melts, its solubility is
highly dependent on the nature and abundance of vari-
ous cations; in particular, Ca, K, and Na. As a result,
carbonate is far more soluble in alkaline basalts than
tholeiites (Blank and Brooker 1994). Dixon (1997) de-
rived a parameter (P), which she used to model the re- Fig. 1 Solubility of total CO2 (molecular + carbonate) as a func-
lationship between carbonate solubility and melt tion of pressure for four melt compositions (ppmw parts per million
composition for tholeiitic to leucititic compositions. The by weight). Leucitite, basanite, and tholeiitic basalt calculated from
variable P, which correlates positively with carbonate compositions and parameters given in Holloway and Blank (1994)
at 1,200 °C. Rhyolite curve from experimental data of Fogel and
solubility, decreases as a function of silica and alumina Rutherford (1990) at 1,150 °C. The solubility of CO2 in rhyolite
concentrations. It also increases with alkalis and alkaline and tholeiitic basalt are very similar. The CO2 solubility increases
earth elements (and therefore with non-bridging oxy- both with pressure and with magma alkalinity
492

Fig. 2 Molecular CO2 vs. carbonate concentrations for CO2-

saturated albite melt (after Stolper et al. 1987). In albite, CO2
dissolves as both carbonate and molecular CO2. Bulk solubility
increases with pressure. The carbonate to molecular CO2 ratio Fig. 3 Solubility plot for system rhyolite±H2O±CO2 at 675 °C.
increases sympathetically with temperature Lines labeled with units of pressure represent isobars that display
solubility of H2O and CO2 as function of ¯uid composition;
intersections with X- and Y-axes represent H2O and CO2 solubility,
Solubility as a function of the presence respectively ± all other points being mixtures. Isopleths labeled
of other volatiles with units of XH2O display the composition of vapor in equilibrium
with the melt. Melt compositions (as deduced from silicate melt
Blank et al. (1993) showed that Henry's Law applies for inclusions, ¯uid inclusions, etc.) can be plotted to calculate the
minimum pressure at which the melt would be saturated with a
the H2O±CO2±rhyolite system at pressures <1 kbar and vapor phase (saturation pressure). The corresponding isopleth
magmatic temperatures. In a melt±vapor system where would indicate the composition of vapor that would exsolve
the vapor contains two components (in this case CO2
and H2O), addition of one gas dilutes the other gas,
decreasing its activity and consequently decreasing its ation of rifts with unusual magma compositions (car-
concentration (solubility) in the melt. In other words, bonatites and strongly silica-undersaturated magmas
addition of water to a melt±CO2 system will decrease the such as nephelinites and melilitites). Except when created
solubility of CO2. In such a case, Henry's Law holds for at unusually deep mantle depths where diamond is stable
the melt, and the activity of a constituent is proportional (Fig. 4; see Haggerty 1999), such magmas will most
to its concentration, which is decreased by addition of typically be created due to melting of carbonated mantle.
another component. This concept is illustrated in Fig. 3, Wyllie and Huang (1976) showed that the reaction:
where the solubility of CO2 at ®xed temperature and
pressure varies with vapor composition in the H2O± 2Mg2 SiO4 ‡ CaMgSi2 O6 ‡ 2CO2
CO2-melt system. There is some controversy whether , 4MgSiO3 ‡ CaMg…CO3 †2 …1†
this Henrian behavior in the melt continues at high
pressures (>5 kbar; e.g., Mysen et al. 1975) due to the is favored at mantle pressures (Fig. 4). At pressures
possible depolymerizing e€ect of water on the melt > 28 kbar (Olafsson and Eggler 1983), magnesite
structure (cf. Stolper et al. 1987). (MgCO3) is the stable crystalline carbonate phase, even
in iron-bearing compositions (Dalton and Wood 1993).
The carbonate-forming reactions destabilize the calci-
CO2 in the mantle um-rich (cpx) and silica-poor minerals (olivine) within
the mantle. Therefore, when lherzolite is melted, result-
The ®rst introduction of CO2 into ma®c magmas occurs ing liquids are rich in calcium and magnesium and are
during partial melting of the mantle. The presence of low in silica, as is observed in silica-undersaturated
mantle-derived carbon is most obviously associated with magmas (nephelinites, melilitites), carbonatites, and
continental and oceanic rifts where deep melting events kimberlites ± all magma types commonly associated with
create carbon-rich, silica-undersaturated magmas. Bailey high abundances of CO2 (Brey 1978). Later experimen-
and colleagues (e.g., Bailey 1987; Bailey and Macdonald tal studies have shown that solid carbonates should be
1987; Bailey and Hampton 1990) have repeatedly argued present in mantle peridotites at temperatures below
that tectonic rift zones are accompanied by unusually 1,200 °C and at oxidation states above QFM (the
carbon-rich magmas, as corroborated by gas-rich quartz±fayalite±magnetite bu€er) minus 0.5 (Dalton and
groundwaters, CO2-rich volcanic gases, and the associ- Wood 1995).
 493

idotites were held at 1,000 °C and decompressed at rates

equivalent to 90 km/h prior to rapid quenching; yet in
all cases the dolomite broke down completely. Berg
(1986) described calcite±brucite intergrowths, which
may also represent relict mantle carbonate phases that
did not survive their ascent to the surface. Given such
rapid breakdown, it is not surprising that mineralogical
evidence for the host phase for CO2 in the mantle is
quite rare. Typically, such evidence consists of either
remnant carbonate melts or carbonate inclusions within
rigid crystal hosts.
In mantle with lower oxidation states or higher
temperatures, carbon is likely to be present either in a
¯uid phase, as diamond (at >50 kbar), or as graphite.
Holloway (1998) presented a model for production of
mid-ocean-ridge basalt from a graphite-saturated mantle
under relatively reducing conditions. At low pressures,
carbonate minerals are not stable and will react to form
CO2 vapor. Roedder (1965) showed convincing evidence
for nearly pure-phase CO2 vapor exsolved in the deep
Fig. 4 Stability of carbonic phases in lherzolitic mantle (the CaO± crystallization environment of numerous basaltic mag-
MgO±SiO2±CO2 system). The CO2, dolomite, and magnesite ®elds mas. Of 72 occurrences of olivine-bearing nodules found
are after Dalton and Wood (1993), with peridotite±CO2 and in basalt lavas, 64 contained high-density CO2-domi-
peridotite±dry solidus from Wyllie (1978). Shield and oceanic
geotherms from Bailey (1985). Note that both the shield and
nated ¯uid inclusions that must have been trapped at
oceanic geotherms would pass through the ®eld of carbonated depths of 10±15 km. Many olivine phenocrysts from the
mantle provided that the mantle fO2 is suciently high to stabilize host lavas also contained similar inclusions, indicating
carbonate (i.e., >QFM minus 0.5; Dalton and Wood 1995) that the magmas were saturated at great depth with
CO2-rich ¯uids.

Although melting will tend to deplete the mantle with

respect to carbon, there are a number of processes that Estimates of CO2 concentrations in magmas
can lead to subsequent enrichment in volatiles. Studies
of mantle xenoliths, particularly those erupted with al- Evidence for the presence of CO2 in crustal magmas is
kali basalts, indicate that the mantle can undergo me- not limited to inclusions in mantle and cognate xeno-
tasomatism by trace element enriched ¯uids or melts liths. During the past 20 years, experimental and pet-
(e.g., Bailey 1987; Menzies and Hawkesworth 1987; rologic studies of igneous volatiles have led to a far
Stolz and Davies 1988). In many cases, the ¯uid re- greater understanding of the abundance of dissolved and
sponsible for these changes is interpreted to be CO2-rich, exsolved gases in magmatic systems. Although concen-
based on the presence of CO2-rich ¯uid inclusions and trations of CO2 in silicate glasses had been estimated by
reaction assemblages (Dautria et al. 1992). Carbonated techniques such as 14C beta-track counting and mano-
melts have been documented in peridotite nodules from metric techniques, it was not until Fourier transform
the Kerguelen Islands by Schiano et al. (1994), who infrared (FTIR) spectroscopy became widely used that
found carbonate-rich melt inclusions and concluded that quantitative techniques for measurement of CO2 con-
they represented metasomatic liquids. centrations were developed (Fine and Stolper 1985;
Evidence for carbonate mineral phases within mantle Stolper et al. 1987). Subsequently, studies of experi-
peridotite nodules was ®rst documented by McGetchin mental run products, quenched natural glasses and sili-
and Besancon (1973), who found small carbonate in- cate melt inclusions were begun (Dixon et al. 1988;
clusions within pyropes. Amundsen et al. (1987) later Newman et al. 1988; Anderson et al. 1989; Fogel and
described quenched carbonate-bearing melts within spi- Rutherford 1990). The latter are small blebs of silicate
nel peridotite xenoliths from Spitsbergen. Later studies melt (often quenched to glass) that are trapped in phe-
on these samples by Ionov et al. (1993) identi®ed mag- nocryst hosts during pre-eruptive crystallization. As-
nesium-rich dolomites, implying equilibrium with man- suming no post-entrapment processes have changed the
tle minerals. Most of the carbonate minerals in these melt inclusion composition (see Lowenstern 1995), these
rocks showed evidence for breakdown by reaction with features provide information on the melt composition
peridotite. Such breakdown should be common and the and dissolved volatiles at the time of trapping. In this
preservation of any relict mantle carbonate minerals is section, I review the recent literature on measured CO2
likely to be extremely rare. Canil (1990) found that Eq. concentrations in natural melts, and what has been
(1) is readily reversed at magmatic temperatures and learned about the nature and abundance of CO2-bearing
sub-mantle pressures. In his experiments, synthetic per- igneous ¯uids in the middle to upper crust.
494

Alkalic basalts magmas from Mt. Erebus, Antarctica, which ranged in

composition from basanite through evolved phonolites.
Given the apparent presence of solid carbonates, CO2- They found very high dissolved CO2 concentrations in
bearing ¯uid, and graphite in di€erent parts of the melt inclusions trapped within olivine phenocrysts; as
mantle, magma generation will result in CO2-bearing much as 0.7 wt%, corresponding to saturation pressures
liquids. For example, carbonate is the ®rst mineral phase of 3±4 kbar given the measured H2O concentrations of
to melt when the peridotite solidus is exceeded at 1.15±1.75 wt%. As in the Vesuvius magmas, evolved
40 kbar (Canil and Scarfe 1990) and this would also be magmas related to the deep ma®c melts are depleted in
expected at lower pressures (Fig. 4). The CO2 concen- volatiles, re¯ecting crystallization and chemical evolu-
trations in partial melts from carbonate-bearing mantle tion in a shallow magma chamber where CO2, H2O, and
peridotite will be dependent on its bulk carbonate con- S are lost to the atmosphere and/or overlying hydro-
tent and the extent of melting. In many cases, initial thermal system. Eschenbacher et al. (1998) conclude that
partial melts might be extremely carbon rich, but would volatile saturation and degassing occurs throughout the
be diluted as melting proceeded. Such a trend was re- continuous fractionation from basanite through phon-
cently observed by Dixon et al. (1997), who looked at olite.
dissolved CO2, both measured and calculated, for a
range of basalt compositions from the submarine North
Arch volcanic ®eld, north of Oahu (Hawaii). They argue MORB and ocean island basalts
that a homogeneous mantle source was variably melted
(1.6±9.0% batch melting), producing liquids ranging A large amount of work has been done on volatile
with progressive melting from nephelinite to basanite to concentrations in MORB glasses dredged from the
alkaline olivine basalt (AOB). Initial volatile contents ocean ¯oor. Johnson et al. (1994) review some of the
were assumed to correlate with P2O5 and were calculated pertinent literature and describe the analytical diculties
to range from 6.9 wt% CO2+H2O in the nephelinites to of volatile analysis in bulk glass samples. In general,
1.8 wt% in the AOBs (with CO2/H2O by wt=2.5 in MORB glasses are H2O-undersaturated, but contain
both). The mantle source region was inferred to be ¯uid vesicles, requiring saturation with a CO2-rich gas at the
absent, although the host phase for carbon was not depth of emplacement. Concentrations of CO2 are
stated and batch melting was assumed. The volatile higher for bulk manometric analyses than for spot an-
content of the unmelted source mantle was estimated to alyses (e.g., FTIR spectroscopy), presumably because
be 525‹75 ppmw (ppm by weight) H2O and they are a€ected by the presence of CO2 in vesicles
1,300‹800 ppmw CO2. Upon ascent, the magmas (Moore et al. 1977). In a study of lavas with numerous
would not have begun degassing (i.e., reached their CO2-rich bubbles (``popping rocks''), Pineau and Javoy
saturation pressures) until reaching depths of 25±40 km, (1983) showed that the di€erence between the carbon
two to three times shallower than their source region. isotopic composition of CO2 from vesicles and CO2
Matrix glass in lavas dredged from the sea¯oor at dissolved in MORB glass could be attributed to isotopic
4 km depth had degassed, but the bulk rocks still fractionation due to outgassing during ascent and that
contained several wt% total volatiles due to the abun- the magnitude of fractionation implied CO2 concentra-
dance of CO2-rich vesicles, consistent with the high tions of 0.2 to 1.0 wt% in the mantle source region. In
calculated volatile contents. Similar results were found general, H2O concentrations in MORBs correlate with
by Moore et al. (1977), who discovered that vesicles in a K2O, and are interpreted to track the degree of melting
variety of submarine basalts contained more than 95% in the source region. The CO2 does not correlate with
CO2, requiring saturation with CO2 at water depths of major or trace elements, presumably due to the e€ect of
5 km and greater. degassing (Johnson et al. 1994).
The highest dissolved CO2 contents measured in Estimates of initial dissolved CO2 concentrations in
natural glasses come from silicate melt (glass) inclusions parental Kilauean ocean island basalts (Hawaii) range
in olivine from ma®c basalts. Because the inclusions are from a minimum of 0.2 wt% (Anderson and Brown
often trapped at great pressure (several kbar), these 1993) to 0.3 wt% (Greenland et al. (1985) to 0.6 wt%
melts have degassed far less than basalts emplaced at (Gerlach and Graeber 1985). Anderson and Brown
depths of 3 to 4 km below sea level (0.3±0.4 kbar). (1993) estimated saturation pressures for melts in Ki-
Marianelli et al. (1999) analyzed glass inclusions in po- lauean glass inclusions that ranged up to 5 kbar, al-
tassium-rich tephrites from Vesuvius, Italy, that have as though most crystals appeared to grow at much lower
much as 3 wt% dissolved H2O and 0.3 wt% dissolved pressures. The 5 kbar estimate is consistent with the data
CO2, requiring saturation with a volatile phase at pres- of Roedder (1965), who found CO2-dominated ¯uid
sures up to and over 5 kbar. These magmas fed a shal- inclusions in both olivine phenocrysts and cumulate
low magma chamber at a depth of 1±2 km, where they xenoliths, requiring saturation with a CO2-rich vapor
degassed and crystallized to form a potassium-rich phase at depth. At Reunion Island, Bureau et al. (1998)
phonotephrite with very low dissolved CO2 interpreted ¯uid and melt inclusion compositions to
(<500 ppmw) and less than 1 wt% dissolved H2O. imply that magmas were volatile-saturated with a CO2-
Similarly, Eschenbacher et al. (1998) studied alkalic rich ¯uid at pressures of at least 5 kbar.
 495

Arc basalts and andesites found evidence for appreciable CO2. Using infrared
(FTIR) spectroscopy, Newman et al. (1988) undertook
Recently, several studies have explored the dissolved CO2 the ®rst study to quantify the abundances of both CO2
concentrations within ma®c arc magmas. Sisson and and H2O dissolved in natural silicic glasses. Analyzing
Bronto (1998) worked on high-magnesium basalts from quenched obsidian fragments ejected from the throat of
the 1983 eruption of Galunggung, on the volcanic front the Mono Craters (California) vent, they found evidence
in Java, Indonesia. They found low H2O concentrations for a degassing trend whereby glasses with successively
(0.21 to 0.38 wt%) in olivine-hosted melt inclusions, but lower CO2 and H2O were created as the magma rose
elevated CO2 (as much as 750 ppmw) and S (as much as through its conduit. The trend could be modeled by
2,200 ppmw), indicating that the inclusions had not closed-system degassing from a magma that started out
leaked and were instead trapped during ascent through with 1.2 wt% CO2 and 5 wt% H2O in the lower crust.
the crust at depths of 3 to 6 km. Presumably, the magmas Lower initial CO2 concentrations would have resulted in
were saturated with a CO2-dominated vapor during signi®cantly di€erent trajectories on a plot of H2O vs.
much of their ascent through the crust, although the in- CO2.
clusions were trapped at low pressure, after much de- Following the lead of Newman et al. (1988), other
gassing occurred. Sisson and Bronto (1998) argue that researchers soon began using FTIR spectroscopy to
the high-magnesium basalt was created by upwelling and measure the H2O and CO2 concentrations in silicate melt
pressure-release melting of the mantle wedge, without inclusions. Anderson et al. (1989) provided the ®rst re-
signi®cant input from ¯uids and magmas derived from liable CO2 concentrations in silicate melt inclusions from
the subducted slab. This accounts for the extremely wa- any magma. Studying the 760 ka eruption of the rhyo-
ter-depleted nature of the magma, which would be un- litic Bishop Tu€, they found evidence for hundreds of
expected for basalts incorporating slab-derived ¯uids. ppmw CO2 dissolved in melt inclusions as well as higher
Such a slab-in¯uenced melt was studied by Roggen- water concentrations than were generally accepted for
sack et al. (1997), who found that basaltic melt inclusions silicic melts (>6 wt%). Anderson et al. (1989) calculated
from Cerro Negro (Nicaragua) had elevated concentra- saturation pressures above 2 kbar. They concluded that
tions of dissolved H2O (as much as 6 wt%) and CO2 the host quartz crystals were forming in a vapor-satu-
(some >1,000 ppmw), yielding estimated saturation rated melt with vapors containing as much as 45 mol%
pressures as high as 6 kbar (i.e., 21 km). Studied bas- CO2. Later work by Wallace et al. (1995) found that
alts that erupted explosively in 1992 are interpreted to dissolved CO2 concentrations were negatively correlated
have stagnated and crystallized at greater depths than with incompatible elements such as uranium, and that
those that erupted e€usively three years later. The high the slope of the correlation could be explained by the
CO2 (and H2O) ¯ux from arc magmas with a signi®cant mass fraction of gas present in various parts of the
slab component is consistent with the conclusions of magma chamber. To date, the study of Wallace et al.
Sano and Williams (1996), who estimated that only (1995) is the only one that has used melt inclusions (and
about 10% of the CO2 degassed from subduction-zone CO2 concentrations) to estimate the amount of gas
volcanoes was derived from the mantle and that a major present in a magma chamber prior to eruption.
part of the CO2 ¯ux was recycled from the subducted Lowenstern (1994) studied volcanic equivalents to the
oceanic crust and its overlying sediments. Because sub- 23 Ma Pine Grove molybdenum porphyry deposit (Utah)
duction zones produce almost two thirds of the volca- and found quartz-hosted melt inclusions with 6±8 wt%
nogenic carbon ¯ux, crustal recycling represents a major H2O and as much as 960 ppmw CO2. Calculated satura-
source of carbon to the atmosphere (Sano and Williams tion pressures ranged as high as 4.3 kbar, requiring that
1996; Alt and Teagle 1999). Similarly, Giggenbach (1996) volatile exsolution began at depths of 16 km during as-
asserted that the volatile budget of arcs follows a steady cent to the shallow regions (2±3 km) where the porphyries
state, whereby a major part of the CO2 and N2 ¯ux is were emplaced. Other studies showed appreciable CO2
recycled via breakdown of sediments from the subducted dissolved in rhyolitic magmas, including the work of
slab and these gases return to the surface through arc Wallace and Gerlach (1994) on the 1991 eruption of
volcanism. Presumably, much of the recycled carbon Pinatubo (Philippines), and that of Gansecki and
comes from carbonaceous sediments and ooze, and ma- Lowenstern (1995) on the 0.6 Ma Lava Creek Tu€
rine limestone (Wyllie and Huang 1976; Plank and (Yellowstone, Wyoming). Lowenstern et al. (1997) found
Langmuir 1993; Alt and Teagle 1999), given that car- relatively low CO2 (40±60 ppmw) and H2O (2 to 3.5 wt%)
bonate phases subducted with basalt are likely to break concentrations in an A-type rhyolite from the Afar rift of
down at low pressures (Canil and Scarfe 1990), before Eritrea. They concluded that the low concentrations were
they can reach the depth of magma generation. due to shallow entrapment pressures (0.4±0.9 kbar),
which was consistent with geologic constraints.
A number of studies of silicic eruptions have found
Silicic magmas no evidence for CO2 in phenocryst-hosted melt inclu-
sions. Such eruptions include the 1912 eruption at the
Although rhyolitic and dacitic magmas are often as- Valley of Ten Thousand Smokes (Lowenstern 1993), the
sumed to contain only H2O-rich gas, many studies have climactic 6.8 ka Crater Lake eruption (Bacon et al.
496

1992), and the 74 ka eruption of the Toba Tu€ (Chesner Crustal sources of carbon
1998). The absence of detectable CO2 in the melt in-
clusions does not mean that CO2 is absent from the Granites and silicic volcanic rocks are more likely to have
magma, given the ready partitioning of CO2 into a vapor an origin through crustal melting than their ma®c coun-
phase. Lowenstern and Mahood (1991) found no de- terparts, and CO2 concentrations in rhyolites and dacites
tectable (<50 ppmw) CO2 in silicate melt inclusions may be augmented by partial melting and assimilation of
from peralkaline rhyolites at Pantelleria, Italy. Never- crustal materials. The formation of skarns often includes
theless, CO2 was detected in vapor bubbles found within breakdown of carbonate-bearing rocks by in®ltration of
melt inclusions (Lowenstern et al. 1991). Because the water-rich magmatic ¯uids (Einaudi et al. 1981; Meinert
magma chamber was located at shallow depths (2± et al. 1997). Such a process is certain to add CO2 to the
3 km), little CO2 was present in the melt phase and magma, while creating decarbonated calc-silicate assem-
nearly all had partitioned into the vapor. Lowenstern blages and carbonate-rich melts (Lentz 1999). Graywac-
et al. (1991) concluded that the CO2-bearing bubbles kes, calcareous sandstones, and many hydrothermally
were present in inclusions where both melt and a coex- altered rocks may contain sucient carbon to provide a
isting vapor phase had been trapped simultaneously. source of CO2 to crustal melts and ascending ma®c
In some systems, CO2 may be virtually absent due to magmas. One diculty in identifying this source of car-
long-term, open-system degassing of the magma cham- bon in silicic magmas is the lack of igneous phenocrysts
ber and consequent loss of the least soluble gases. Evi- that incorporate carbon. With the exception of a small
dence for such open-system degassing has been inferred carbonate component in ¯uorapatite, CO2 is partitioned
by the behavior of sulfur and sulfur isotopes as mea- almost entirely into outgassed ¯uids and has a very low
sured in melt inclusions, matrix glass, and coexisting abundance in analyses of volcanic rocks and igneous in-
sul®de minerals (Mandeville et al. 1998). As such, CO2 trusions (White et al. 1999). Where present, such carbon
could be distilled out of magmas by long-term loss of may be sub-solidus in nature (White et al. 1999). As a
vapor, combined with little input of new CO2 from ei- result, only direct sampling of magmatic gases is likely to
ther wallrock or associated magmas. provide insight into the source of CO2 in magmas. Un-
fortunately, at this point, most isotopic data for carbon in
volcanic gas is from ma®c systems.
Evidence for carbonic ¯uids in granites

Fluid inclusion studies of unaltered granites have pro- Can vapor-saturated magmas ascend
vided similar evidence of vapor-saturated silicic magmas through the crust and erupt?
in the intermediate to shallow crust. Frost and Touret
(1989) studied coeval saline and CO2-rich ¯uid inclu- Magma ascent
sions in the Laramie anorthosite complex and concluded
that silicate melt, nearly pure CO2, and hydrosaline melt As discussed above, many recent studies provide evi-
coexisted at a late magmatic stage of this complex. dence for CO2-rich vapors in magmas within the middle
Calculated isochores allowed them to infer that the in- to upper crust. It is a commonly held belief that vapor-
clusions were trapped at 3 kbar and 950 °C. The coex- saturated magmas cannot ascend through the crust
istence of immiscible silicate melt, CO2-rich vapor, and without freezing, and that addition of CO2 to a magma
hydrosaline melt was ®rst shown by Roedder and Coo- should also cause it to crystallize. However, the presence
mbs (1967) for granitic blocks ejected during eruptions of vapor is not always likely to impede the movement of
at Ascension Island and has since been studied experi- magmas through the crust.
mentally by Shinohara et al. (1989; see also Shinohara Tuttle and Bowen (1958) stated that H2O-saturated
1994). Thomas and Spooner (1992) studied CO2-bearing anatectic magmas that reside at the haplogranite ternary
¯uid inclusions in quartz crystals from the Tanco gran- minimum cannot rise through the crust because any loss
itic pegmatite and found that the high-temperature of H2O via degassing would move the solidus to higher
``magmatic'' inclusions contained dominantly H2O temperatures, resulting in crystallization and solidi®ca-
(95%) and CO2, and that this ¯uid unmixed at low tion. This statement has been corroborated by numerous
temperatures to form a vapor phase and a dense brine subsequent studies, but it overlooks the probability that
phase. Frezzotti et al. (1994) found evidence that the most anatectic melts are created at depths where water
leucogranites from the Deep Freeze Range of Antarctica saturation would require >10 wt% dissolved H2O. If
were vapor saturated at a pressure of 3 kbar at 750 °C magmas instead contain less dissolved water, as has
with a ¯uid having aH2O=0.5, with the remainder con- typically been inferred in many of the studies listed
sisting mostly of CO2. Nabelek and Ternes (1997) in- above, they should be able to ascend through the crust
terpreted CO2-rich ¯uid inclusions from the Harney without impediment, even if they become saturated with
Peak Granite (Black Hills, South Dakota) to be primary a multicomponent vapor containing other gases in ad-
magmatic vapors trapped at 3.5 kbar in equilibrium dition to H2O.
with a granitic melt containing 3.5 wt% dissolved H2O Because CO2 dissolves in its molecular form in silicic
and 1,500 ppmw dissolved CO2. melts at crustal pressures (Blank et al. 1993), it has little
 497

e€ect on melt structure and controls phase equilibria ¯uids. Just as the presence of a vapor phase does not
primarily through dilution of coexisting vapors and necessarily hasten magma crystallization, it also does not
consequent reduction of H2O activity. Holloway (1976) imply immediate eruption. Shinohara et al. (1995) pre-
showed that closed-system equilibration of silicic magma sented a model whereby bubble-rich magmas undergo
with CO2-bearing vapor can bu€er the H2O concentra- open-system degassing at shallow depths (2±3 km) to
tion of the melt, causing PH2O to stay constant, or in- create Climax-type molybdenum porphyry deposits. In
crease, during magma ascent. This process is illustrated such a scenario, the magma is suciently volatile-rich
in Fig. 5 (diagram similar to those in Anderson et al. that it can outgas as a porous open system without fur-
1989 and Lowenstern 1994). The trends for decom- ther bubble growth and expansion, allowing magma that
pression degassing show that ascent will have minimal loses its bubbles to sink relative to the volatile- and
e€ect on dissolved H2O concentration and, therefore, on molybdenum-bearing magma rising from depth. This
melt structure and phase relations. In other words, if convective model would work best for volatile-rich
degassing does not cause the dissolved H2O content to magmas; they would be less eruptible because they can
change appreciably, then there will be no impetus to- become permeable (see Eichelberger et al. 1986) and lose
wards crystallization. In fact, slow adiabatic (isenthal- exsolved gas at greater depths than less gas-rich magmas.
pic) ascent of the magma alone will be likely to cause a As discussed above, thermodynamics and phase dia-
small temperature increase (Waldbaum 1971). Thus, grams predict that rising hydrous magmas should crys-
during decompression, the vapor-saturated magma will tallize before reaching the surface if ascent took place
not necessarily crystallize and can rise into the upper under equilibrium conditions. If the vapor-saturated
crust (4±6 km depth). Once the magma rises to shallow magma represented in Fig. 5 were to rise to depths where
regions, the 1±4 km depths associated with epithermal pressures are less than 2 kbar, then the vapor phase
and porphyry ore deposits, subsequent ascent will likely would become increasingly H2O-rich as water is drawn
force degassing of H2O-rich (but still CO2-bearing) ¯u- out of the melt. At low pressures, degassing has a pro-
ids. This process will inevitably begin to a€ect the soli- nounced e€ect on dissolved H2O concentration and
dus and to induce crystallization. should inhibit magma ascent by inducing crystallization.
Even so, we know from modern volcanologic studies that
vapor- (and H2O-) saturated magmas can and do rise and
Magma eruption erupt from relatively shallow magma chambers (<11 km
depths; e.g., Pinatubo, Bishop Tu€, Cerro Negro, and all
Often, magma does not erupt but instead ponds in the the other volcanic events summarized above). The
crust to crystallize as intrusions. The creation of a lode rhyolitic melts that form glassy lavas, such as obsidian
gold deposit from a magmatic source requires that at domes, although clearly H2O-saturated during much of
least some magma remains behind in the crust and un- their ®nal ascent to the surface (Eichelberger et al. 1986;
dergoes devolatilization to release the metal-bearing Newman et al. 1988; Taylor 1988), are able to rise and
erupt without signi®cant crystallization and can ¯ow for
kilometers over the Earth's surface as supercooled liq-
uids. Their ability to do so is a result of the sluggish
crystallization kinetics of silicic melts and the decrease in
viscosity caused by the presence of bubbles in magmas
(Bagdassarov and Dingwell 1992).
The most likely situation where the presence of CO2
would impede magma ascent would be where a CO2-
poor magma was created in the mantle or deep crust at a
relatively low temperature. As it rose through the crust,
if there were a mechanism to add large amounts of CO2
to the magma (e.g., by assimilation/melting of carbo-
naceous sediments or limestone; Lentz 1999), the re-
sulting CO2-rich vapor phase would draw water out of
Fig. 5 Solubility plot for the system rhyolite±H2O±CO2 at 750 °C. the melt, potentially causing the magma to cross its
Lines labeled with units of pressure represent isobars that display solidus and crystallize (Fig. 6).
solubility of H2O and CO2 as function of ¯uid composition (see
Fig. 3). Isopleths not shown (see Fig. 3). Trend parallel to isobars
shows e€ect of crystallization on vapor-saturated melt at 4.6 kbar
starting with 2,000 ppmw CO2 and 5.0 wt% H2O. Decompression- The degassing process
induced degassing trends show paths of isothermal ascent, in
equilibrium with the vapor (closed system) or allowing created Given the evidence for CO2-bearing vapors in a wide
vapor to escape (open). Numbers along paths list mol% H2O in range of crustal magmas, it becomes important to
vapor (remainder being CO2). The decompressing magma should
degas without signi®cant crystallization until it reaches virtual
understand the general chemical behavior and char-
H2O-saturation (X-axis). At that point, magma ascends as a acteristics of such vapors. One generality is that early-
supercooled liquid, as in all silicic eruptions exsolved vapors will always be more CO2-rich than those
498

CH4/CO2 ratios may be appreciable at the temperatures

expected for hydrothermal mineralization in intrusion-
hosted gold deposits (300±450 °C).
Another important concept to consider is the e€ect of
CO2 on unmixing of the volatile phase, particularly in
silicic magmas where temperatures may not exceed
800 °C. Under such conditions, halides and some sul-
fates have extremely low solubility in CO2-bearing va-
pors. For example, in the H2O±NaCl system at 700 °C,
¯uids with more than 5 wt% NaCl will unmix to form
immiscible vapor and brine phases at any pressure lower
than about 1.1 kbar (Souririjan and Kennedy 1962).
However, addition of 20 wt% or more CO2 will cause
such a ¯uid to unmix at any pressure lower than about
5 kbar (Bodnar and Sterner 1987; Joyce and Holloway
1993). Increasing temperature will only increase the
likelihood of unmixing (as long as the temperature is
below the melting point of halite; 800 °C at 1 bar). This
Fig. 6 Pressure versus temperature diagram for the granite solidus means that the conditions where silicic magmas are
(thin lines) and the decarbonation of calcite (thick lines: ®gure after saturated with both a vapor and brine (hydrosaline melt)
Holloway 1976). A vapor-saturated magma near its solidus and will be common in the middle to upper crust. Given the
with aCO2 in the vapor equal to 0.1 would be supercooled (relative di€erent partitioning behavior of metals between melt,
to its solidus) by about 35 °C if the magma were ¯ooded with
enough CO2 to move the vapor-phase activity of CO2 to 0.25 vapor, and brine (Candela and Piccoli 1995), the pres-
(arrow). This would be possible, for example, by heating siliceous ence of CO2 will increase the likelihood of fractionation
limestones above 600 °C, causing calcite (cc) + quartz (qtz) to of metals among ¯uids with di€erent densities and
form wollastonite (wo). This reaction would eventually slow down transport properties.
due to the production of CO2

exsolved later. As the magma rises and/or cools, the low CO2 in hydrothermal ¯uids and surface manifestations
solubility of CO2 relative to H2O will cause it to parti-
tion preferentially into any exsolving vapor phase. The Although this review focuses primarily on CO2 in
earliest-formed vapors may also be (relatively) rich in magmas, it is appropriate to discuss some general con-
other magma-insoluble gases such as the noble gases and cepts that describe the evolution and behavior of car-
N2 (Giggenbach 1997). Both closed- and open-system bonic ¯uid in hydrothermal systems. As magmatic
degassing of any natural silicate melt will result in pro- vapors, brines, and supercritical ¯uids leave the magma
gressively more H2O-rich vapors with time. This will be and enter hydrothermal systems, they can potentially act
the case whether degassing is caused by depressurization as mineralizing ¯uids. Just as in the magma, the presence
or crystallization (``second boiling''). If bubble forma- of CO2 will promote immiscibility and creation of a
tion is delayed until the magma reaches shallow depths, vapor phase that will incorporate sulfur gases and noble
partitioning into the vapor may be dominated by H2O, gases as well as CO2 and steam.
swamping any CO2 in the system. Magmas can also Pressure decreases will tend to cause boiling or e€er-
contain little CO2 if their di€erentiation is protracted vescence of CO2- and H2S-bearing vapors. Such boiling
and magma-insoluble gases are distilled away over time. events may be most dramatic in the high-temperature
Evidence for open-system degassing has been noted by (300±450 °C) environments envisioned for intrusion-
Mandeville et al. (1998) and is discussed above in the hosted Au deposits. At high temperatures, rock loses
section on silicic magmas. strength and becomes ductile, which spurs annealing and
CO2 is not the only carbon-bearing species that will reduction of rock permeability. Pressure can build until
be present in magmatic emanations. Concentrations of rapid ¯uid-release events force brittle fracturing, some-
as much as 0.4 wt% CO occur in some hot, low fO2 times causing pressure to decline from lithostatic to hy-
basaltic gases (Table 3 of Symonds et al. 1994), about drostatic gradients (Fournier 1999). Such catastrophic
5% of the abundance of CO2. More typically, CO is two depressurization will force boiling of hydrothermal so-
to four orders of magnitude less abundant than CO2, lutions and release of CO2-rich vapors. Sheeted hydro-
particularly in andesitic and dacitic gases and those with thermal veins may result from repetition of these
fO2>QFM. Methane (CH4) is extremely rare in volcanic pressure ¯uctuations with their associated boiling, min-
gas and is usually attributed to contamination with eral precipitation reactions, and subsequent annealing.
geothermal and low-temperature crustal gases (Symonds At temperatures above 300 °C, CO2 strongly parti-
et al. 1994) or initial oxidation states well below QFM. tions into the vapor phase and has a relatively low sol-
However, as magmatic ¯uids re-equilibrate at lower ubility in aqueous ¯uids (Rimstidt 1997). The vapor can
temperatures, methane concentrations will increase, and be relatively unreactive because carbonic acid weakly
 499

dissociates at temperatures above 300 °C (Bischo€ and tance of carbonic ¯uids in transporting ore-forming
Rosenbauer 1996; Giggenbach 1997). As temperatures metals. Perusing a recent compilation on melt/vapor/
fall, however, carbonic acid (H2CO3) more readily dis- brine partition coecients (Candela and Piccoli 1995), it
sociates, allowing it to react with rock to leach cations becomes apparent that little is known about the role of
and form bicarbonate-rich geothermal ¯uids (Bischo€ CO2 as a ligand, and little ®rm evidence exists that CO2
and Rosenbauer 1996). Even at low temperatures, CO2 plays a direct role in ore formation. Seward and Barnes
has only limited solubility in geothermal waters. For (1997) demonstrate that based on Lewis acid±base be-
example, at 200 °C, a 1 mol% solution of CO2 in pure havior of metals and ligands, one would not predict
H2O has a partial pressure of 66 bar (data from Chapter metal±carbonate complexes to be particularly stable.
4 of Henley et al. 1984), whereas that for water is 16 bar. Where experimental studies have explored the parti-
Thus, at or below 82 bar, a CO2-dominated vapor will tioning of metals into CO2-rich ¯uids, most have con-
exsolve from this solution as bubbles. Addition of other cluded that increasing concentrations of CO2 tend to
gases or the presence of NaCl in solution will increase decrease metal vapor/melt partition coecients (e.g.,
the vapor pressure and thus the pressure at which bub- Webster et al. 1989).
bles will form. At the Broadlands geothermal area, New A few studies, however, have documented apparent
Zealand, evidence for CO2-saturated geothermal ¯uids mobilization of metals by CO2-rich ¯uids, primarily in
extends to depths of 2,500 m (p. 54, Henley et al. 1984). the deep crust. Wendlandt and Harrison (1979) dem-
As a result, CO2 gas often accumulates in the shallow onstrated the ability of carbonate-rich melts to transport
crust and can result in hydrothermal and geothermal LREE and discussed the importance of LREE enrich-
well-®eld explosions (e.g., Hedenquist and Henley 1985). ments in fenites and metasomatized mantle. Keppler and
Volcanic regions have long been known as subject to Wyllie (1990) found limited, but signi®cant, uranium
CO2 buildup. The 1986 disaster at Lake Nyos, Camer- mobility in high-temperature magmatic ¯uids at mod-
oon, evidently occurred due to deep volcanic degassing erate crustal pressures (2 kbar). They noted that in-
and resultant accumulation of CO2-saturated water in creasing CO2 concentrations in the ¯uid were
the deep lake, followed by catastrophic overturn and accompanied by higher uranium concentrations. In a
release of a large, CO2 gas cloud, killing over 1,700 petrographic study of tungsten-mineralized rocks, Hig-
people (Kusakabe 1996). At Mammoth Mountain, Cal- gins (1980) noted a correlation between CO2-rich ¯uid
ifornia, ongoing di€use degassing of 530 t/day CO2 has inclusions and tungsten concentrations. However, ex-
caused about 30 ha of forest to die and is a continuing perimental studies have yet to document this relation-
source of concern (Farrar et al. 1995, 1999). The gas ship.
appears to emanate from a vapor-rich gas reservoir lo- We are, therefore, left with a preponderance of
cated in seismically active zones within 3 km of the sur- studies that show evidence for CO2-bearing, ¯uid in-
face (Julian et al. 1998). The isotopic composition of the clusions in gold-bearing intrusions and the apparent lack
gas is consistent with derivation from magma and the of experimental evidence that CO2 acts as a ligand for
tree-kill began after a 1989 seismic swarm that was ap- most ore-associated metals. One potential explanation
parently related to intrusion of a basaltic dike (Farrar for the apparent discrepancy would be simply that CO2
et al. 1995). On Mt. Etna, Italy, residents have long been is a very common gas in igneous and hydrothermal
aware of di€use percolation of CO2 through the volcanic systems and should be ubiquitous regardless of whether
edi®ce, resulting in CO2-charged groundwater and dif- it has a role in the ore-forming process. Appreciable
fuse degassing that constitutes a signi®cant percentage of CO2/H2O ratios (e.g., >0.1) could indicate that the
the total CO2 released by the volcano (Allard et al. 1997). magma did not undergo a protracted period of di€er-
Clearly, CO2 behaves as a highly volatile species, both entiation and degassing, which would have caused loss
within magmas and in superjacent hydrothermal sys- of the more magma-insoluble components such as CO2.
tems. In magmas, there are no crystalline components Deep degassing, and input of volatiles from underlying
that incorporate CO2 or carbonate groups (except for a magmas could also result in the presence of appreciable
very minor amount in apatite) and CO2 solubility is CO2 in ¯uids derived from the ore-hosting magma. It is
generally low in the melt. In hydrothermal systems, CO2 also likely that CO2 acts as a fugitive agent in the hy-
is not highly soluble under most conditions and precip- drothermal ¯uid, readily causing unmixing of the ¯uid to
itates as carbonate minerals primarily at lower temper- create a separate vapor phase. Heinrich (1995) summa-
atures. As a result, CO2 can exsolve from hydrothermal rizes evidence for vapor-phase transport of copper in a
waters and groundwaters to escape to the surface. number of intrusion-hosted deposits, and especially in
tin deposits. Audetat et al. (1998) give an even more
striking example of vapor-phase transport of copper and
Metal partitioning into CO2-bearing ¯uids boron, presumably by sul®de complexes. They stress
that gold should also be complexed by sul®de in the
Given the evidence for CO2-bearing ¯uids in magmatic vapor phase. If this is the case, the presence of CO2 will
systems and, in particular, associated with gold-bearing hasten the creation of a vapor phase causing H2S, and
intrusions (Newberry et al. 1995; Thompson et al. 1999 possibly HCl, to move into the vapor and act as ligands
and references therein), one must ascertain the impor- for some metals. At the same time, e€ervescence of CO2
500

from the ¯uid will result in a decrease in PCO2 and an Brey G (1978) Origin of olivine melilitites± chemical and experi-
increase of pH in the remaining liquid, causing impor- mental constraints. J Volcanol Geotherm Res 3:61±73
Bureau H, Pineau F, MeÂtrich N, Semet MP, Javoy M (1998) A melt
tant mineral precipitation reactions (e.g., Simmons and and ¯uid inclusion study of the gas phase at Piton de la
Christenson 1994) and destabilization of ligands such as Fournaise volcano (ReÂunion Island). Chem Geol 147:115±130
gold bisul®de (Seward 1973). Therefore, although the Candela PA, Piccoli PM (1995) Model ore±metal partitioning from
e€ect of CO2 on mineralization may not be direct, this melts into vapor and vapor±brine mixtures. In: Thompson JFH
(ed) Magmas, ¯uids and ore deposits. Mineral Assoc Canada
common gas plays a key role in the creation and evo- Short Course 23, pp 101±127
lution of metal-bearing vapors in ore-forming systems. Canil D (1990) Experimental study bearing on the absence of
carbonate in mantle-derived xenoliths. Geology 18:1011±1013
Acknowledgements Sally Newman provided code adapted to cal- Canil D, Scarfe CM (1990) Phase relations in peridotite+CO2
culate Figs. 3 and 5. I appreciate helpful reviews by Terry Gerlach, systems to 12 GPa: implications for the origin of kimberlite and
Je€ Hedenquist, Larry Meinert, Stuart Simmons, and Tom Sisson. carbonate stability in the Earth's upper mantle. J Geophys Res
95:15805±15816
Chesner CA (1998) Petrogenesis of the Toba Tu€s, Sumatra, In-
donesia. J Petrol 39:397±438
References Dalton JA, Wood BJ (1993) The partitioning of Fe and Mg be-
tween olivine and carbonate and the stability of carbonate
Allard P, Jean-Baptiste P, D'Alessandro W, Parello F, Parisi B, under mantle conditions. Contrib Mineral Petrol 114:501±509
Flehoc C (1997) Mantle-derived helium and carbon in Dalton JA, Wood BJ (1995) The stability of carbonate under up-
groundwaters and gases of Mount Etna, Italy. Earth Planet Sci per-mantle conditions as a function of temperature and oxygen
Lett 148:501±516 fugacity. Eur J Mineral 7:883±891
Alt JC, Teagle DAH (1999) The uptake of carbon during alteration Dautria JM, Dupuy C, Takherist D, Dostal J (1992) Carbonate
of ocean crust. Geochim Cosmochim Acta 63:1527±1535 metasomatism in the lithospheric mantle: peridotitic xenoliths
Amundsen HEF, Grin WL, O'Reilly SY (1987) The lower crust from a melilitic district of the Sahara basin. Contrib Mineral
and upper mantle beneath northwestern Spitsbergen; evidence Petrol 111:37±52
from xenoliths and geophysics. Tectonophysics 139:169±185 Dixon JE (1997) Degassing of alkalic basalts. Am Mineral 82:368±
Anderson AT Jr, Brown GG (1993) CO2 and formation pressures 378
of some Kilauean melt inclusions. Am Mineral 78:794±803 Dixon JE, Stolper E, Delaney JR (1988) Infrared spectroscopic
Anderson AT Jr, Newman S, Williams SN, Druitt TH, Skirius C, measurements of CO2 and H2O in Juan de Fuca Ridge basaltic
Stolper E (1989) H2O, CO2, Cl and gas in Plinian and ash-¯ow glasses. Earth Planet Sci Lett 90:87±104
Bishop rhyolite. Geology 17:221±225 Dixon JE, Clague DA, Wallace P, Poreda R (1997) Volatiles in
Audetat A, GuÈnther D, Heinrich CA (1998) Formation of a alkalic basalts from the North Arch Volcanic Field, Hawaii:
magmatic±hydrothermal ore deposit: insights with LA±ICP± extensive degassing of deep submarine-erupted alkalic series
MS analysis of ¯uid inclusions. Science 279:2091±2094 lavas. J Petrol 38:911±939
Bacon CR, Newman S, Stolper E (1992) Water, CO2, Cl, and F in Eichelberger JC, Carrigan CR, Westrich HR, Price RH (1986)
melt inclusions in phenocrysts from three Holocene explosive Non-explosive silicic volcanism. Nature 323:598±602
eruptions, Crater Lake, Oregon. Am Mineral 77:1021±1030 Einaudi MT, Meinert LD, Newberry RJ (1981) Skarn deposits.
Bagdassarov NS, Dingwell DB (1992) A rheological investigation Econ Geol 75th Anniv Vol, pp 317±391
of vesicular rhyolite. J Volcanol Geotherm Res 50:307±322 Eschenbacher AJ, Kyle PR, Lowenstern JB, Dunbar NW (1998)
Bailey DK (1985) Fluids, melt, ¯owage and styles of eruption in Melt inclusion investigation of the volatile behavior of frac-
alkaline ultrama®c magmatism. Trans Geol Soc S Afr 88:449± tionating alkaline magma, Mount Erebus, Ross Island. Ant-
457 arctica. In: Vanko DA, Cline JS (eds) PACROFI ± Pan
Bailey DK (1987) Mantle metasomatism: perspective and prospect. American Conference on Research on Fluid Inclusions, Pro-
In: Fitton JG, Upton BGJ (eds) Alkaline igneous rocks. Geol gram and Abstracts 7, p 30
Soc Spec Publ 30, pp 1±13 Farrar CD, Sorey ML, Evans WC, Howle JF, Kerr BD, Kennedy
Bailey DK, Hampton CM (1990) Volatiles in alkaline magmatism. BM, King C-Y, Southon, JR (1995) Forest-killing di€use CO2
Lithos 26:257±165 emission at Mammoth Mountain as a sign of magmatic unrest.
Bailey DK, Macdonald R (1987) Dry peralkaline felsic liquids and Nature 376:675±678
carbon dioxide ¯ux through the Kenya rift zone. In: Mysen BO Farrar CD, Neil JM, Howle JF (1999) Magmatic carbon dioxide
(ed) Magmatic processes: physicochemical processes. Geochem emissions at Mammoth Mountain, California. US Geol Surv
Soc Spec Publ 1, pp 91±105 Water Resour Invest Rep 98±4217
Baker LL, Rutherford MJ (1996) Crystallization of anhydrite- Fine G, Stolper E (1985) The speciation of carbon dioxide in
bearing magmas. Trans R Soc Edinburgh 87:243±250 sodium aluminosilicate glass. Contrib Mineral Petrol 91:105±
Berg GW (1986) Evidence for carbonate in the mantle. Nature 121
324:50±51 Fogel RA, Rutherford MJ (1990) The solubility of carbon dioxide
Bischo€ JL, Rosenbauer RJ (1996) The alteration of rhyolite in in rhyolitic melts: a quantitative FTIR study. Am Mineral
CO2 charged water at 200 and 350 °C: the unreactivity of CO2 75:1311±1326
at higher temperatures. Geochim Cosmochim Acta 60:3859± Fournier RO (1999) Hydrothermal processes related to movement
3867 of ¯uid from plastic into brittle rock in the magmatic±epither-
Blank JG, Brooker RA (1994) Experimental studies of carbon di- mal environment. Econ Geol 94:1193±1211
oxide in silicate melts: solubility, speciation, and stable carbon Frezzotti ML, Di Vicenzo G, Ghezzo C, Burke EAJ (1994) Evi-
isotope behavior. In: Carroll MR, Holloway JR (eds) Volatiles dence of magmatic CO2-rich ¯uids in peraluminous graphite-
in magmas. Volatiles in magmas. Rev Mineral vol 30, pp 157± bearing leucogranites from Deep Freeze Range (northern Vic-
186 toria Land, Antarctica). Contrib Mineral Petrol 117:111±123
Blank JG, Stolper EM, Carroll MR (1993) Solubilities of carbon Frost BR, Touret JLR (1989) Magmatic CO2 and saline melts from
dioxide and water in rhyolitic melt at 850 °C and 750 bar. Earth the Sybille Monzosyenite, Laramie Anorthosite complex, Wy-
Planet Sci Lett 119:27±36 oming. Contrib Mineral Petrol 103:178±186
Bodnar RJ, Sterner SM (1987) Synthetic ¯uid inclusions. In: Ulmer Gansecki CA, Lowenstern JB (1995) Pre-eruptive volatile compo-
GC, Barnes HL (eds) Hydrothermal experimental techniques. sitions of the Lava Creek Tu€ magma, Yellowstone Plateau
Wiley Interscience, New York, pp 423±457 Volcanic Field. Trans Am Geophys Union, Eos 76:F665
 501

Gerlach TM (1980) Chemical characteristics of the volcanic gases Lentz DR (1999) Carbonatite genesis: a reexamination of the role
from Nyiragongo lava lake and the generation of CH4-rich ¯uid of intrusion-related pneumatolytic skarn processes in limestone
inclusions in alkaline rocks. J Volcanol Geotherm Res 8:177±189 melting. Geology 27:335±338
Gerlach TM, Graeber EJ (1985) The volatile budget of Kilauea, Lowenstern JB (1993) Evidence for a copper-bearing ¯uid in
Hawaii. Nature 313:273±277 magma erupted at the Valley of Ten Thousand Smokes, Alaska.
Giggenbach WF (1996) Chemical composition of volcanic gases. Contrib Mineral Petrol 114:409±421
In: Tilling RI, Scarpa R (eds) Monitoring and mitigation of Lowenstern JB (1994) Dissolved volatile concentrations in an ore-
volcanic hazards. Springer, Berlin Heidelberg New York, forming magma. Geology 22:893±896
pp 221±256 Lowenstern JB (1995) Applications of silicate melt inclusions to the
Giggenbach WF (1997) The origin and evolution of ¯uids in study of magmatic volatiles. In: Thompson JFH (ed) Magmas,
magmatic±hydrothermal systems. In: Barnes HL (ed) Geo- ¯uids and ore deposits. Mineral Assoc Canada Short Course 23,
chemistry of hydrothermal ore deposits, 3rd edn. Wiley, New pp 71±99
York, pp 737±796 Lowenstern JB, Mahood GA (1991) New data on magmatic H2O
Goldfarb RJ, Phillips GN, Nokleberg WJ (1998) Tectonic setting contents of pantellerites, with implications for petrogenesis and
of synorogenic gold deposits of the Paci®c Rim. In: Ramsay eruptive dynamics at Pantelleria. Bull Volcanol 54:78±83
WRH, Bierlein FP, Arne DC (eds) Mesothermal gold miner- Lowenstern JB, Mahood GA, Rivers ML, Sutton SR (1991) Evi-
alization in space and time. Ore Geol Rev 13, pp 185±218 dence for extreme partitioning of copper into a magmatic vapor
Greenland LP, Rose WI, Stokes JB (1985) As estimate of gas phase. Science 252:1405±1409
emissions and magmatic gas content from Kilauea Volcano. Lowenstern JB, Clynne MA, Bullen TD (1997) Comagmatic A-
Geochim Cosmochim Acta 49:125±129 type granophyre and rhyolite from the Alid volcanic center,
Haggerty SE (1999) A diamond trilogy: superplumes, superconti- Eritrea, northeast Africa. J Petrol 38:1707±1721
nents and supernovae. Science 285:851±860 Mandeville CW, Sasaki A, Saito G, Faure K, King R, Hauri E
Hedenquist JW, Henley RW (1985) Hydrothermal eruptions in the (1998) Open-system degassing of sulfur from Krakatau 1883
Waiotapu geothermal system, New Zealand: their origin, as- magma. Earth Planet Sci Lett 160:709±722
sociated breccias, and relation to precious metal mineralization. Marianelli P, Metrich N, Sbrana A (1999) Shallow and deep res-
Econ Geol 80:1640±1668 ervoirs involved in magma supply of the 1944 eruption of
Heinrich CA (1995) Geochemical evolution and hydrothermal Vesuvius. Bull Volcanol 61:48±63
mineral deposition in Sn (±W ± base metal) and other granite- McGetchin TR, Besancon JR (1973) Carbonate inclusions in
related ore systems: some conclusions from Australian exam- mantle-derived pyropes. Earth Planet Sci Lett 18:408±410
ples. In: Thompson JFH (ed) Magmas, ¯uids and ore deposits. Meinert LD, Hefton KK, Mayes D, Tasiran I (1997) Geology,
Mineral Assoc Canada Short Course 23, pp 203±220 zonation and ¯uid evolution of the Big Gossan Cu±Au skarn
Henley RW, Truesdell AH, Barton PB Jr, Whitney JA (1984) deposit, Ertsberg District, Irian Jaya. Econ Geol 92:509±534
Fluid±mineral equilibria in hydrothermal systems. Rev Econ Menzies MA, Hawkesworth (eds) (1987) Mantle metasomatism.
Geol 1 Academic Press, London
Higgins NC (1980) Fluid inclusion evidence for the transport of Moore JG, Batchelder JN, Cunningham CG (1977) CO2-®lled ves-
carbonate complexes in hydrothermal solutions. Can J Earth icles in mid-ocean basalt. J Volcanol Geotherm Res 2:309±327
Sci 17:823±830 Mysen BO, Arculus RJ, Eggler DH (1975) Solubility of carbon
Holloway JR (1976) Fluids in the evolution of granitic magmas: dioxide in natural nephelinite, tholeiite and andesite melts to
consequences of ®nite CO2 solubility. Geol Soc Am Bull 30 kbar pressure. Contrib Mineral Petrol 53:227±239
87:1513±1518 Nabelek PI, Ternes K (1997) Fluid inclusions in the Harney Peak
Holloway JR (1998) Graphite±melt equilibria during mantle Granite, Black Hills, South Dakota, USA: implications for
melting: constraints on CO2 in MORB magmas and the carbon solubility and evolution of magmatic volatiles and crystalliza-
content of the mantle. Chem Geol 147:89±97 tion of leucogranite magmas. Geochim Cosmochim Acta
Holloway JR, Blank JG (1994) Application of experimental results 61:1447±1465
to C±O±H species in natural melts. In: Carroll MR, Holloway Newberry RJ, McCoy DT, Brew DA (1995) Plutonic-hosted gold
JR (eds) Volatiles in magmas. Rev Mineral vol 30, pp 187±230 ores in Alaska: igneous vs. metamorphic origins. In: Ishihara S,
Ionov DA, Dupuy C, O'Reilly YO, Kopylova MG, Genshaft YS Czamanske GK (eds) Proceedings of the Sapporo international
(1993) Carbonated peridotite xenoliths from Spitsbergen: im- conference on mineral resources of the NW Paci®c Rim. Japan,
plications for trace element signature of mantle carbonate me- 4±8 October 1994. Soc Resour Geol Japan, Spec Issue vol 18,
tasomatism. Earth Planet Sci Lett 119:283±297 pp 57±100
Johnson MC, Anderson AT Jr, Rutherford MJ (1994) Pre-eruptive Newman S, Epstein S, Stolper E (1988) Water, carbon dioxide and
volatile content of magmas. In: Carroll MR, Holloway JR (eds) hydrogen isotopes in glasses from the ca. 1340 A.D. eruptions
Volatiles in magmas. Rev Mineral vol 30, pp 281±330 of the Mono Craters, California: constraints on degassing
Joyce DB, Holloway JR (1993) An experimental determination of phenomena and initial volatile content. J Volcanol Geotherm
the thermodynamic properties of H2O±CO2±NaCl ¯uids at high Res 35:75±96
pressures and temperatures. Geochim Cosmochim Acta 75:733± Olafsson M, Eggler DH (1983) Phase relations of amphibole, am-
746 phibole±carbonate and phlogopite±carbonate peridotite: pet-
Julian BR, Pitt AM, Foulger GR (1998) Seismic image of a CO2 rologic constraints on the asthenosphere. Earth Planet Sci Lett
reservoir beneath a seismically active volcano. Geophys J Int 64:305±315
133:F7±F10 Papale P, Polacci M (1999) Role of carbon dioxide in the dynamics
Kadik AA, Lukanin OA, Lebedev YB, Kolovushkina EY (1972) of magma ascent in explosive eruptions. Bull Volcanol
Solubility of H2O and CO2 in granite and basalt melts at high 60:583±594
pressures. Geochem Int 9:1041±1050 Pineau F, Javoy M (1983) Carbon isotopes and concentrations in
Keppler H, Wyllie PJ (1990) Role of ¯uids in transport and frac- mid-ocean ridge basalts. Earth Planet Sci Lett 62:239±257
tionation of uranium and thorium in magmatic processes. Na- Plank T, Langmuir CH (1993) Tracing trace elements from sedi-
ture 348:531±533 ment input to volcanic output at subduction zones. Nature
King PL, Forneris J, Holloway JR, Hervig RL (1996) C±H±O 362:739±742
speciation in andesite glasses prepared at high pressure and Rimstidt JD (1997) Gangue mineral transport and deposition. In:
temperature. Trans Am Geophys Union, Eos 77:F802 Barnes HL (ed) Geochemistry of hydrothermal ore deposits,
Kusakabe M (1996) Hazardous crater lakes. In: Tilling RI, Scarpa 3rd edn. Wiley, New York, pp 487±515
R (eds) Monitoring and mitigation of volcanic hazards. Roedder E (1965) Liquid CO2 inclusions in olivine-bearing nodules
Springer, Berlin Heidelberg New York, pp 573±598 and phenocrysts from basalts. Am Mineral 50:1746±1782
502

Roedder E, Coombs DS (1967) Immiscibility in granitic melts, Holloway JR (eds) Volatiles in magmas. Rev Mineral vol 30,
indicated by ¯uid inclusions in ejected granitic blocks from pp 1±66
Ascension Island. J Petrol 8:417±451 Taylor BE (1988) Degassing of rhyolitic magmas: hydrogen isotope
Roggensack K, Hervig RL, McKnight SB, Williams SN (1997) evidence and implications for magmatic±hydrothermal ore de-
Explosive basaltic volcanism from Cerro Negro volcano: posits. In: Taylor RP, Strong DF (eds) Recent advances in the
in¯uence of volatiles on eruptive style. Science 277:1639±1642 geology of granite-related mineral deposits. Can Inst Min
Sano Y, Williams SN (1996) Fluxes of mantle and subducted car- Metall Spec Vol 39, pp 33±49
bon along convergent plate boundaries. Geophys Res Lett Thomas AV, Spooner ETC (1992) The volatile geochemistry of
23:2749±2752 magmatic H2O±CO2 ¯uid inclusions from the Tanco zoned
Schiano P, Clocchiatti R, Shimizu N, Weis D, Mattielli N (1994) granitic pegmatite, southeastern Manitoba, Canada. Geochim
Cogenetic silica-rich and carbonate-rich melts trapped in Cosmochim Acta 56:49±65
mantle minerals in Kerguelen ultrama®c xenoliths: implications Thompson JFH, Sillitoe RH, Baker T, Lang JR, Mortensen JK
for metasomatism in the oceanic upper mantle. Earth Planet Sci (1999) Intrusion-related gold deposits associated with tungsten±
Lett 123:167±178 tin provinces. Miner Deposita 34:323±334
Seward TM (1973) Thiocomplexes of gold and the transport of Tuttle OF, Bowen NL (1958) Origin of granite in the light of
gold in hydrothermal ore solutions. Geochim Cosmochim Acta experimental studies in the system NaAlSi3O8±KAlSi3O8±SiO2±
37:379±399 H2O. Geol Soc Am Mem 74
Seward TM, Barnes HL (1997) Metal transport by hydrothermal Waldbaum DR (1971) Temperature changes associated with adia-
ore ¯uids. In: Barnes HL (ed) Geochemistry of hydrothermal batic decompression in geological processes. Nature
ore deposits, 3rd edn. Wiley, New York, pp 435±486 232:545±547
Simmons SF, Christenson BW (1994) Origins of calcite in a boiling Wallace PJ, Gerlach TM (1994) Magmatic vapor source for sulfur
geothermal system. Am J Sci 294:361±400 dioxide released during volcanic eruptions: evidence from
Sisson TW, Bronto S (1998) Evidence for pressure-release melting Mount Pinatubo. Science 265:497±499
beneath magmatic arcs from basalt at Galunggung, Indonesia. Wallace PJ, Anderson AT Jr, Davis AM (1995) Quanti®cation of
Nature 391:883±886 pre-eruptive exsolved gas contents in silicic magmas. Nature
Shinohara H (1994) Exsolution of immiscible vapor and liquid 377:612±616
phases from a crystallizing silicate melt: implications for chlo- Webster JD, Holloway JR, Hervig RL (1989) Partitioning of
rine and metal transport. Geochim Cosmochim Acta lithophile trace elements between H2O and H2O+CO2 ¯uids
58:5215±5221 and topaz rhyolite melt. Econ Geol 84:116±134
Shinohara H, Iiyama JT, Matsuo S (1989) Partition of chlorine Wendlandt RF, Harrison WJ (1979) Rare earth partitioning be-
compounds between silicate melt and hydrothermal solutions I. tween immiscible carbonate and silicate liquids and CO2 vapor:
Partition of NaCl±KCl. Geochim Cosmochim Acta results and implications for the formation of light rare earth-
53:2617±2630 enriched rocks. Contrib Mineral Petrol 69:409±419
Shinohara H, Kazahaya K, Lowenstern JB (1995) Volatile trans- White AF, Bullen TD, Vivit DV, Schulz MS, Clow DW (1999) The
port in a convecting magma column: implications for porphyry role of disseminated calcite in the chemical weathering of
Mo mineralization. Geology 23:1091±1094 granitoid rocks. Geochim Cosmochim Acta 63:1939±1953
Sourirajan S, Kennedy GC (1962) The system H2O±NaCl at White DE, Waring GA (1963) Data of geochemistry, 6th edn.
elevated temperatures and pressures. Am J Sci 260:115±141 Chapter K, Volcanic emanations. US Geol Surv Prof Pap 44-K
Stolper E, Fine G, Johnson T, Newman S (1987) Solubility of Wyllie PJ (1978) Mantle ¯uid compositions bu€ered in peridotite-
carbon dioxide in albitic melt. Am Mineral 72:1071±1085 CO2±H2O carbonates, amphibole and phlogopite. J Geol
Stolz AJ, Davies GR (1988) Chemical and isotopic evidence from 86:687±713
spinel lherzolite xenoliths for episodic metasomatism of the Wyllie PJ, Huang WL (1976) Carbonation and melting reactions in
upper mantle beneath Southeast Australia. In: Menzies MA, the system CaO±MgO±SiO2±CO2 at mantle pressures with
Cox KG (eds) Oceanic and continental lithosphere: similarities geophysical and petrological applications. Contrib Mineral
and di€erences. Oxford University Press, Oxford. J Petrol Spec Petrol 54:79±107
Issue pp 303±330 Wyllie PJ, Tuttle OF (1959) E€ect of carbon dioxide on the melting
Symonds RB, Rose WI, Bluth GJS, Gerlach TM (1994) Volcanic- of granite and feldspars. Am J Sci 257:648±655
gas studies: methods, results and applications. In: Carroll MR,