Journal of Petrology Advance Access originally published online on March 28, 2006
Journal of Petrology 2006 47(7):1413-1437; doi:10.1093/petrology/egl016
© The Author 2006. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org
Experimental and Thermodynamic Constraints on the Sulphur Yield of Peralkaline and Metaluminous Silicic Flood Eruptions
BRUNO SCAILLET1,* and
RAY MACDONALD2
1 INSTITUT DES SCIENCES DE LA TERRE D'ORLÉANS, UMR 6113 CNRS-UO 1A RUE DE LA FÉROLLERIE, 45071 ORLÉANS, FRANCE
2 ENVIRONMENT CENTRE, LANCASTER UNIVERSITY LANCASTER LA1 4YQ, UK
RECEIVED
NOVEMBER 9, 2004;
ACCEPTED
MARCH 6, 2006
 |
ABSTRACT
|
|---|
Many basaltic flood provinces are characterized by the existence
of voluminous amounts of silicic magmas, yet the role of the
silicic component in sulphur emissions associated with trap
activity remains poorly known. We have performed experiments
and theoretical calculations to address this issue. The melt
sulphur content and fluid/melt partitioning at saturation with
either sulphide or sulphate or both have been experimentally
determined in three peralkaline rhyolites, which are a major
component of some flood provinces. Experiments were performed
at 150 MPa, 800–900°C,
fO
2 in the range NNO –
2 to NNO + 3 and under water-rich conditions. The sulphur content
is strongly dependent on the peralkalinity of the melt, in addition
to
fO
2, and reaches 1000 ppm at NNO + 1 in the most strongly
peralkaline composition at 800°C. At all values of
fO
2,
peralkaline melts can carry 5–20 times more sulphur than
their metaluminous equivalents. Mildly peralkaline compositions
show little variation in fluid/melt sulphur partitioning with
changing
fO
2 (
DS 
270). In the most peralkaline melt,
DS rises
sharply at
fO
2 > NNO + 1 to values of >500. The partition
coefficient increases steadily for S
bulk between 1 and 6 wt
% but remains about constant for S
bulk between 0·5 and
1 wt %. At bulk sulphur contents lower than 4 wt %, a temperature
increase from 800 to 900°C decreases
DS by
10%. These results,
along with (1) thermodynamic calculations on the behaviour of
sulphur during the crystallization of basalt and partial melting
of the crust and (2) recent experimental constraints on sulphur
solubility in metaluminous rhyolites, show that basalt fractionation
can produce rhyolitic magmas having much more sulphur than rhyolites
derived from crustal anatexis. In particular, hot and dry metaluminous
silicic magmas produced by melting of dehydrated lower crust
are virtually devoid of sulphur. In contrast, peralkaline rhyolites
formed by crystal fractionation of alkali basalt can concentrate
up to 90% of the original sulphur content of the parental magmas,
especially when the basalt is CO
2-rich. On this basis, we estimate
the amounts of sulphur potentially released to the atmosphere
by the silicic component of flood eruptive sequences. The peralkaline
Ethiopian and Deccan rhyolites could have produced
10
17 and
10
18 g of S, respectively, which are comparable amounts to published
estimates for the basaltic activity of each province. In contrast,
despite similar erupted volumes, the metaluminous Paraná–Etendeka
silicic eruptives could have injected only 4·6
x 10
15 g of S in the atmosphere. Peralkaline flood sequences may thus
have greater environmental effects than those of metaluminous
affinity, in agreement with evidence available from mass extinctions
and oceanic anoxic events.
KEY WORDS: silicic flood eruptions; sulphur; experiment; Ethiopia; Deccan
|
INTRODUCTION
|
|---|
A detailed understanding of the solubility of sulphur in silicate
melts is important for many geological processes (O'Neill &
Mavrogenes, 2002
): the origin of magmatic sulphide ores, the
geochemical behaviour of the chalcophile trace elements, including
the platinum group elements and the Re–Os isotopic system
and their use as tracers of core–mantle and crust–mantle
differentiation in the Earth and other planets or asteroids
(e.g. Kracher & Wasson, 1982
; Jana & Walker, 1997
; Chabot,
2004
). Sulphur solubility along with melt–vapour partitioning
also determines sulphur degassing during volcanic activity and
the potential effects on climate variability.
Most studies of sulphur solubility, cited by Carroll & Webster (1994), O'Neill & Mavrogenes (2002) and Jugo et al. (2005), have focused on mafic and intermediate melt compositions. Our understanding of S behaviour in silicic magmas is more restricted and comes mainly from work on Fe-poor, metaluminous compositions [mol. (CaO + Na2O + K2O) > Al2O3 > (Na2O + K2O)], of the type commonly associated with volcanic arcs (Carroll & Rutherford, 1985; Luhr, 1990; Scaillet et al., 1998; Keppler, 1999; Scaillet & Evans, 1999; Clemente et al., 2004; Costa et al., 2004). Such compositions are not directly applicable to the magmatism of continental extensional zones, where the rhyolites tend to be alkali- and Fe-rich and either metaluminous or peralkaline [mol. (Na2O + K2O) > Al2O3] (Macdonald et al., 1992; King et al., 1997). The alkali/alumina balance of silicate melts is known to profoundly affect their geochemical and physical behaviour (Mysen, 1988). Most notable among those effects are the increase in Fe3+/Fe2+ redox ratio (Gwinn & Hess, 1989; Moore et al., 1995; Gaillard et al., 2001) and water solubility (Dingwell et al., 1997), and the decrease of viscosity (Scarfe, 1977; Baker & Vaillancourt, 1995; Dingwell et al., 1998) and liquidus temperatures (Bailey & Schairer, 1966) as peralkalinity increases. These important differences imply that insight gained from the study of sulphur in metaluminous silicate melts is of little help in anticipating its behaviour in peralkaline rhyolites.
Although many flood basalt provinces seem to be dominated by basaltic lavas, several recent studies have emphasized that some have significant volumes of associated silicic rocks (Bellieni et al., 1986; Harris & Erlank, 1992; Ewart et al., 1998, 2004; Ayalew et al., 2002; Bryan et al., 2002; Peccerillo et al., 2003). The objective of the present study is to evaluate the sulphur yield potentially delivered to the atmosphere by silicic flood eruptions. Given that peralkaline rhyolites can largely dominate over metaluminous types in some silicic provinces (Ayalew et al., 2002), we currently lack the fundamental information on which to base any such evaluation. We have, therefore, performed melt solubility and fluid/melt partitioning experiments for sulphur in peralkaline rhyolites from the Kenya Rift Valley, which broadly typify the felsic end-member of bimodal associations in rift-related settings (Bellieni et al., 1986; Harris & Erlank, 1992; Ewart et al., 1998, 2004; Ayalew et al., 2002; Bryan et al., 2002; Peccerillo et al., 2003). We combine these data with thermodynamic and mass-balance calculations to evaluate the sulphur contents of potential sources of felsic magmas associated with flood basalts, considering two end-member cases for the origin of the silicic end-member: fractional crystallization of flood basalt and partial melting of dehydrated lower crust. We show that the attainment of peralkaline conditions dramatically increases the sulphur-carrying capacity of rhyolite magmas. We then estimate the sulphur yields of some silicic flood sequences. The results appear to be significant for the current debate on the volcanic origin of some of the main mass extinctions (e.g. Wignall, 2001; Morgan et al., 2004).
|
EXPERIMENTAL AND ANALYTICAL TECHNIQUES
|
|---|
We have performed sulphur solubility and fluid/melt partitioning
experiments on three well-characterized, peralkaline rhyolitic
obsidians from the Naivasha area of the Kenya Rift Valley (
Table 1).
Two (ND, SMN) are mildly peralkaline comendites [(Na
2O + K
2O)/Al
2O
3 = 1·05 and 1·31, respectively from the Greater
Olkaria Volcanic Complex and were previously used in phase equilibrium
studies (Scaillet & Macdonald, 2001
, 2003
). The third, EBU,
is a pantelleritic obsidian from a welded fall deposit of the
Eburru Volcanic Complex, immediately north of Olkaria. It is
more strongly peralkaline than the comendites (NK/A = 1·88)
and represents a composition that could have been generated
by
50% crystallization of Olkaria-type comendites similar to
SMN (Scaillet & Macdonald, 2003
). These starting melt compositions
(
Table 1) encompass the entire peralkalinity range displayed
by rhyolites associated with flood basalts (Trua
et al., 1999
;
Ayalew
et al., 2002
; Peccerillo
et al., 2003
). We restricted
our investigations to low-pressure conditions (
150 MPa), thought
to be relevant to the production of alkali rhyolites (Mahood,
1984
; Scaillet & Macdonald, 2001
) and we focused on the
role of oxygen fugacity (
fO
2), which exerts a strong control
on sulphur behaviour in silicic magmas (e.g. Carroll & Webster,
1994
; Scaillet
et al., 1998
; Keppler, 1999
).
The experimental procedures are similar to those used in previous
studies performed at the ISTO experimental petrology laboratory,
such as that by Scaillet
et al. (1998)
and Clemente
et al. (2004)
.
For each composition, batches of starting glass + sulphur powders
with fixed sulphur contents were prepared by weighing 200 mg
of glass powder with appropriate amounts of elemental sulphur
(1–12 mg added sulphur). The resulting mixture was thoroughly
mixed in an agate mortar for several minutes. Most experiments
used ND, SMN and EBU batches with bulk sulphur contents of
c.
1 wt % (1·01, 1·19 and 1·08 wt %, respectively;
see
Table 1). However, three series with EBU composition were
performed with additional bulk sulphur contents of 0·5,
2·01, 3·85 and 6·14 wt % S, so as to explore
the effect of sulphur fugacity at fixed
P and
T and
fH
2. We
used Au capsules, which were loaded with the glass + sulphur
powder (
c. 20–30 mg,
Table 1) plus weighed amounts of
distilled water (
c. 10 wt %,
Table 1), and welded shut with
a graphite arc welder. Weighing before and after welding, as
well as after run completion, was used as a monitor for any
volatile loss that occurred at
P and
T. For all charges used
in this study, capsule weights remained constant to within 0·2
mg. We used either cold seal pressure vessels (CSPV) fitted
with an H
2 membrane (runs at 800°C) or an internally heated
pressure vessel (IHPV) with a drop quench setting (900°C).
Errors in quoted temperatures and pressures are ±8°C,
and 2 MPa, respectively. The
fO
2 was varied using various Ar–H
2 mixtures as the pressurizing gas. The H
2 fugacity was either
read with an H
2-membrane (800°C) or known from previous
experiments that used the same Ar–H
2 ratio and that were
run with an H
2 membrane (900°C). Errors on H
2 fugacities
are ±0·01 MPa (800°C), or ±0·5
MPa (900°C), the latter determined by repeat experiments
performed with the IHPV while fitted with a H
2 membrane. The
fO
2 of each charge was computed using the dissociation reaction
of water (Robie
et al., 1979
), the
fH
2 and the water fugacity
(
fH
2O). The
fH
2O of each charge (
Table 1) was computed assuming
ideal behaviour in the fluid [i.e.
aH
2O =
XH
2O
fl,
XH
2O
fl being
the mole fraction of H
2O in the fluid and
aH
2O the activity
of water defined as the ratio
fH
2O/
fH
2O°,
fH
2O° being
the fugacity of pure water at
P and
T as given by Burnham
et al. (1969)
] and using the calculated fluid composition as determined
from mass-balance constraints. The fluid composition was calculated
knowing the amounts of dissolved water and sulphur, the amount
of sulphur locked into solid phases (sulphide and sulphate)
and the bulk content of water and sulphur loaded into the capsules.
The water content of quenched melts (glasses) was determined
using the by-difference method employing appropriate sets of
hydrous glass standards of known H
2O contents (Scaillet &
Macdonald, 2001
). For charges with
c. 1 wt % bulk sulphur, the
calculated
XH
2O
fl of the equilibrium fluid (
Table 2) is close
to, or higher than, 0·9, with the result that the calculated
fO
2 is only marginally lower than that corresponding to pure
H
2O (the difference is <0·05 log units). For those
charges we estimate that the uncertainty on
fO
2 is <0·1
log units. In contrast, charges having bulk sulphur contents
in excess of 1 wt % have calculated fluid compositions that
are significantly poorer in H
2O, with
XH
2O
fl going down to 0·68
(
Table 2). As a result, the value of
fO
2 for the sulphur-rich
charges is lower than that corresponding to the pure H
2O case.
However, for a fixed set of
P–
T–
fH
2 conditions (i.e.
series EBU-9 in
Table 1), the difference in calculated
fO
2 between
the sulphur-poor and sulphur-rich charges does not exceed 0·3
log units (
Table 1). The main source of uncertainty in S-rich
charges comes from the estimate of the amount of H
2O dissolved
in the melt. An error of 1 wt % absolute (i.e. 4·5 wt
% instead of 5·5 wt %) produces a change in calculated
XH
2O
fl of about ±0·07, which in turn induces a
change in the calculated log
fO
2 of about ±0·07
units. Overall, for the sulphur-rich charges we estimate that
the
fO
2 is known to within 0·3 log units.
Because we use elemental sulphur as a source of sulphur, and
not either sulphide or anhydrite (e.g. Luhr, 1990
), the initial
redox state of the charges is grossly out of equilibrium, as
the sulphur in the fluid must be complexed either with O
2 to
form SO
2 or with H
2 to produce H
2S, which are the two dominant
fluid species in the fluid at
P and
T, depending on
fO
2 (see
Clemente
et al., 2004
). As long as H
2 from the vessel does not
diffuse across the capsule walls (i.e. below 600°C), the
redox state of the charges will be internally controlled to
some value that will depend on factors such as the ratio of
H
2O to S loaded, the amount of atmospheric O
2 present in the
capsule, and the rates of H
2O and S dissolution into glass/melt.
At 800°C, however, the kinetics of H
2 diffusion even within
Au is so fast that osmotic equilibrium in H
2 is attained within
a few tens of seconds across the capsule walls (Scaillet
et al., 1992
), and maintained at a fixed value owing to the large
H
2-buffering capacity of the vessel volume (Schmidt
et al.,
1995
). Both the CSPV and IHPV attain the target temperature
in about 20 min, so that the period elapsed at low-temperature
conditions, during which transient redox state may occur, is
short relative to the total run duration (minimum of 96 h,
Table 1).
Redox equilibrium will be reached as soon as the dissolution
process of volatiles into melt is achieved. Using appropriate
diffusivities for H
2O and S [taken at 800°C as 10
–7 cm
2/s and 10
–9 cm
2/s, respectively (Watson, 1994
)], it
can be calculated that volatile dissolution occurs in a matter
of minutes (H
2O) to a few hours (S) owing to the finely powdered
and well-mixed nature of our starting material, which minimizes
the diffusion length scale of volatiles to a few tens of microns.
None of the charges produced in this study displayed textural
evidence of redox disequilibrium, such as sulphide rimmed by
sulphate or vice versa, suggesting that either the charge has
no time to record redox states radically different from that
imposed at run conditions, or any evidence of such variations
in
fO
2 during the heating-up period has been erased upon run
completion.
Run products were characterized first by optical inspection with a metallographic microscope and then by electron microprobe analysis (EMPA). Observation under reflected light prior to carbon-coating allowed the oxide to be distinguished easily from sulphide owing to their contrasted colours (grey for oxide vs yellowish for sulphide), as well as shape (rectangular to equant shape for oxide vs pentagonal to rounded shape for smaller individual sulphide grains). Investigation under transmitted light and crossed polars allowed straightforward identification of anhydrite in oxidized charges because of its high birefringence. In anhydrite-bearing charges, anhydrite was fully enclosed by glass, suggesting that it grew from the melt and is not the result of back-reaction of fluid upon cooling, in which case it should have partially filled open cavities representing former gas bubbles. In all cases, optical identification of minerals was confirmed by subsequent EMPA, although owing to their generally small size, most analyses of minerals were contaminated by glass. We used the following conditions to determine the composition of glasses. For major elements, the conditions were: accelerating voltage 15 kV, sample current 6 nA, counting time 10 s on peak for all elements, and a beam defocused to 5 µm. Na and K were analysed first and a ZAF correction procedure applied. Correction factors for Na loss were based on analyses of synthetic peralkaline rhyolitic compositions and a set of metaluminous rhyolitic and dacitic glasses, all of known water content as determined by Karl Fischer titration. Between six and 10 analyses were performed for each glass. The water content of quenched glasses varied between 5 and 6 wt %, except for the two charges with c. 6 wt % sulphur, which displayed slightly lower values (between 4 and 5 wt %). Owing to the large uncertainties of the by-difference technique in determining the H2O content of quenched glasses (e.g. Devine et al., 1995), in this work we use a fixed melt water content of 5·5 wt % for all charges. The Na+K/Al ratios of quenched glasses do not show any significant departure from that of the starting materials (Table 2). The concentration of total sulphur in glasses was determined by EMPA using three synthetic hydrous dacitic glasses containing 750, 1400 and 1900 ppm sulphur (determined by wet chemistry) as standards (Clemente et al., 2004). The EMPA was run with the following conditions: accelerating voltage 15 kV, sample current 50 nA, beam diameter 10 µm, and a counting time of 60 s. The background was determined by analysing a dry rhyolitic glass without sulphur, using the above analytical procedure. The detection limit under these analytical conditions is about 80 ppm. The sulphur data given in Table 2 are averages of 5–16 analyses.
Apart from oxide and sulphur-bearing phases, no other mineral was detected. Two S-bearing phases crystallized: pyrrhotite at NNO <1 and anhydrite at NNO >1. At high fO2, an oxide co-precipitated with anhydrite. Knowing the amount of sulphur-bearing phases crystallized, and the amount of sulphur dissolved in the melt, the sulphur content of the coexisting fluid is calculated by difference with the known bulk sulphur content, using stoichiometric FeS and CaSO4 as sulphide or sulphate minerals (e.g. Scaillet et al., 1998). Because, depending on fO2, sulphide or sulphate + oxides were the sole phases crystallizing (Table 1), the maximum proportions of sulphur-bearing minerals can be determined by the changes in either FeOtot (fO2 < NNO) or CaO (fO2 > NNO + 1) of the quenched glass. This obviously assumes that neither iron or calcium is transported into the fluid (or lost to the capsule in the case of iron) but, under our experimental conditions, sulphur was never detected in Au capsules. If either of these elements is partitioned into the fluid, the amounts of sulphide or sulphate calculated from variations in FeO and CaO abundances in glass will be overestimated, with the consequence that there will be too much sulphur locked in solid phases. As a result, the calculated partition coefficient will be lower than its real value. Conversely, as stated above, we assume that pyrrhotite is end-member FeS. Pyrrhotite departure from FeS stoichiometry can be up to Fe5S6. An Fe5S6 stoichiometry would decrease our calculated partition coefficients, as it maximizes the amount of sulphur tied up with iron in pyrrhotite and this sulphur is, therefore, no longer available to the fluid. For charges saturated in anhydrite, a problem also arises from the low bulk CaO content of the starting rocks. Generally those charges have melts with CaO contents close to the detection limit. The fact that this extreme depletion in CaO goes along with a decrease in melt Cl content relative to the starting value (see Table 2) suggests that not all the Ca complexes with sulphur to form anhydrite, but that some goes into the fluid, possibly as CaCl2 species. The concentration of CaO in charges run at low fO2 remains close to the starting value, although some depletion does occur, possibly also as a result of Ca complexing with Cl. We recognize that partition coefficients of sulphur between fluid and melt determined in this way (by default) can be affected by a number of errors, as illustrated below. However, there is, unfortunately, no straightforward way of assessing this parameter in hydrothermal experiments. In particular, measuring the H–O–S fluid compositions upon quench is unlikely to retrieve the correct numbers, as back-reactions within fluid can alter both its speciation and composition (L. Baker, personal communication, 1996 20XX).
As an example we consider charge ND1 whose partition coefficient, DS, calculated using the above assumptions, is 307 (Table 1). This charge consisted of 20·2 mg of silicate powder with 1·01 wt % S (0·2 mg), and 1·9 mg of H2O. The bulk FeO is 1·81 wt % (0·37 mg) and after the run the hydrous glass (5·5 wt % or 1·11 mg H2O) has 0·39 wt % FeOtot (0·08 mg). The difference in FeO content in glass before and after the run (0·2867 mg) implies that 0·00399 millimoles of Fe is sequestered in pyrrhotite (0·2867/71·85). Assuming stoichiometric pyrrhotite this implies in turn that 0·128 mg (32 x 0·004) of sulphur is locked up in pyrrhotite. Knowing that the glass has 268 ppm dissolved sulphur, corresponding to 0·0054 mg sulphur, this leaves 0·07 mg sulphur for the fluid (or 0·074 mg H2S). The amount of water in the fluid is 0·79 mg (1·9–1·11), which implies that the mole fraction of H2O in the fluid, XH2O, is 0·95, and that the amount of fluid at P and T is 0·864 mg (0·79 + 0·074), which corresponds to 3·9 wt % fluid in the system [0·864/(20·2 + 1·9)], the fluid having 8·2 wt % sulphur. The amount of pyrrhotite is 1·65 wt %, calculated on the basis of condensed phases only (hydrous glass + pyrrhotite).
We now explore the individual effect of the main parameters that affect the calculated DS, namely glass iron content, pyrrhotite stoichiometry, iron loss toward the capsule, and glass water content. If, instead of 0·39 wt % FeO, the glass contains 0·49 wt % (i.e. the amount of iron is allowed to increase by 1
of EMPA), this increases DS from 307 to 342. Similarly, if instead of FeS a stoichiometry of Fe5S6 is taken (1·2 mole of S for 1 mole of Fe), this means that 0·15 g of sulphur is locked into pyrrhotite, which decreases DS to 202, or by 30%. Alternatively, if we assume that the Au capsule has dissolved 100 ppm Fe [corresponding to a loss of 8 wt % of iron relative to bulk content, as observed in supra-liquidus charges by Scaillet & Macdonald (2004)], then this will increase DS to 394. Finally, if the amount of H2O dissolved is 4 wt % instead of 5·5 wt % (which would correspond to the possible melt H2O content of the charges with 6 wt % sulphur), DS decreases to 227. There is clearly a considerable uncertainty on our fluid/melt partition coefficients, yet each parameter taken in isolation affects DS by less than 40% when allowed to vary within a reasonable range. Our assumptions of stoichiometric pyrrhotite and fixed melt water content, if incorrect, lead to an overestimation of the calculated DS. Conversely, other assumptions (in particular no Fe or Ca loss toward the fluid), if properly evaluated, would yield an underestimation of partition coefficients (i.e. real values are higher than the values listed in Table 2). This is, in fact, the main source of uncertainty in the present work, as we have no control on the amount of dissolved fluid species other than H2O and S. However, the work of Scaillet et al. (1998), using the same procedure, yielded partition coefficients for sulphur in silicic arc magmas that agree within a factor of two with those derived independently from remote sensing of volcanic plumes. We note in passing that this study explored the effect of CO2 and concluded that this volatile species has no detectable effect on the partition behaviour of sulphur between fluid and melt in silicic magmas in the fO2 range explored. We, thus, conclude that the partition coefficients reported here are accurate to within 50%, and the proportions of sulphide/sulphate are known to within 15%.
|
RESULTS
|
|---|
The experimental conditions together with run products and phase
proportions are listed in
Table 1. The melt compositions together
with the calculated partition coefficients are listed in
Table 2.
Variations in melt sulphur content (S
melt) with
fO
2 (here expressed
in log units notation relative to the NiNiO solid buffer, such
that NNO – 1 means one log unit below NNO) are shown in
Fig. 1a. Also plotted is the S
melt of synthetic metaluminous
silicic melts held under similar
P and
T and bulk S content
(S
bulk) conditions (Luhr, 1990
; Scaillet
et al., 1998
). The
sulphur solubility of rhyolite melts is strongly dependent on
their NK/A ratio, in addition to
fO
2. At 800°C, under reduced
conditions (<NNO), S
melt rises from
c. 50 ppm in metaluminous
rhyolite to nearly 1000 ppm in strongly peralkaline rhyolite
(
Fig. 1a). In the two less peralkaline compositions (ND and
SMN), an increase in
fO
2 produces a smooth increase in S
melt,
which broadly conforms with observed behaviour in other silicate
melt compositions (e.g. Carroll & Webster, 1994
). In contrast,
in the most peralkaline composition EBU, the increase in S
melt peaks at around NNO + 1 and then sharply decreases, so that
at higher
fO
2 the relative order of sulphur enrichment for the
three peralkaline rhyolites is opposite to that observed below
NNO. The enhanced sulphur solubility of peralkaline rhyolites
at low
fO
2 is in part related to the higher iron content (
Table 2),
which complexes with S (Carroll & Webster, 1994
), but may
also be due to the greater proportion of oxygen anions in peralkaline
melts relative to metaluminous varieties, as oxygen anions are
believed to substitute for S (Carroll & Webster, 1994
) via
the following reaction:
This
reaction makes no assumption as to the nature of the element
complexed to sulphur in the melt. It simply states that, at
fixed
fS
2, an increase in the activity of free oxygen will increase
the amount of sulphur dissolved. Given that peralkaline melts
have more free oxygen than metaluminous ones (e.g. Mysen, 1988
),
this could be one explanation for their enhanced sulphur solubility
at low
fO
2. At high
fO
2, the opposite, and thus peculiar, trend
displayed by EBU warrants further discussion. The above reaction
predicts that an increase in
fO
2 decreases the sulphur solubility,
if again it is assumed that
fS
2 does not change across the interval
of
fO
2 shown in
Fig. 1. However, rigorously evaluating the reason
for such a trend would require us to know the exact values of
intensive parameters other than
fO
2; namely,
fS
2 and the activities
of oxygen and sulphur anions in the melt, all of which is unknown
in our experiments (and in almost all experimental work so far
done at high
P on aluminosilicate melts of geological interest).
One possibility would be that our oxidized experiments with
the EBU composition did not reach anhydrite saturation, and,
thus, that the anhydrite in these charges represents a quench
phase and that the drop in solubility observed at high
fO
2 simply
reflects the fact that sulphur is being increasingly partitioned
toward the fluid as
fO
2 increases. We do not agree with such
an interpretation because of the textural evidence given above.
In addition, the experimental conditions adopted in our study
are similar to those of other experimental work aimed at defining
the stability of anhydrite in silicate melts, in particular
hydrous melts (e.g. Carroll & Rutherford, 1985
; Luhr, 1990
;
Scaillet
et al., 1998
; Scaillet & Evans, 1999
; Clemente
et al., 2004
), as are the criteria used to define anhydrite
stability domains in magmas. In other words, if anhydrite is
a quench product in our experiments, so it is in other experimental
studies, in which case conclusions derived from those experiments
about the stability field of anhydrite in magmas should be revised.
Rather, we suggest that the trend shown by the EBU melt composition
is a real feature that reflects the fundamental structural changes
that occur in strongly peralkaline silicic liquids relative
to metaluminous types, as noted in the Introduction, and the
control that
fO
2 exerts upon them. By analogy with what we observe
at low
fO
2, we can only speculate that the increase in
fO
2 dampens
the role of oxygen anions in peralkaline melts, which results
in a decreased S
melt (at fixed fS
2).
More generally, if sulphur solubility in hydrous silicic melts
is a simple linear function of their iron content (at low
fO
2),
as conventional wisdom would predict, then there would be no
reason to have the huge difference between metaluminous and
peralkaline compositions that we document here (
Fig. 1a). The
very fact that
Smelt increases by a factor of
c. 5 between standard
metaluminous rhyolitic and slightly peralkaline melts with similar
iron contents indicates that the FeO content is not the sole
parameter affecting sulphur behaviour in the studied compositions,
unlike in dry basaltic systems (O'Neill & Mavrogenes, 2002
).
This is also shown by the three series of experiments carried
out with various amounts of sulphur, in which the melt iron
content decreases whilst the amount of sulphur dissolved in
the melt increases (
Table 2, series EBU 7, 8 and 9). For instance,
at 900°C, the melt with 969 ppm dissolved sulphur has an
FeO content of 6·32 wt %, whereas that with 2254 ppm
of sulphur has an FeO content of 1·52 wt %. This trend
is clearly due to our procedure of adding elemental sulphur,
which forces the system to crystallize FeS, which in turn removes
FeO from the melt. It shows, however, that the FeO content and
dissolved sulphur in hydrous silicic melts are not simply correlated.
In this respect, the experimental study of Clemente
et al. (2004)
has clearly shown that in metaluminous rhyolites, the iron control
on sulphur solubility is not straightforward, and that variations
in
fS
2,
fH
2S and
fSO
2 affect predominantly S
melt. This study
has shown that in hydrous silicate melts, the S
melt can be modelled
as the result of two dissolution reactions, one involving H
2S
and the other SO
2, as proposed by Burnham (1979)
. Given the
hydrous character of our experiments, it can be anticipated
that the fugacities of sulphur-bearing species will exert a
role, possibly a predominant one, on S
melt in peralkaline rhyolites
as well. Direct comparison of our results with experimental
studies bearing on the sulphur solubility of anydrous melts
is, therefore, to a large extent meaningless (or of any anhydrous
vs hydrous studies), because in the latter the dissolution reactions
of S-bearing species cannot account for the role of H-bearing
ones (i.e. H
2S), leaving aside the profound structural changes
produced by the incorporation of water into silicate melts.
In particular, the presence of hydrogen opens the possibility
that Fe is not the element with which S preferentially complexes
in silicate melts, but instead that H plays that role, as proposed
by Burnham (1979)
, based on theoretical and experimental considerations,
a hypothesis that is in agreement with the findings of Clemente
et al. (2004)
. In other words, because both the nature and abundances
of dissolving species (H
2S vs S
2 or SO
2) and the sulphur complexes
in the melt (H
2S vs FeS) are likely to differ between hydrous
and anhydrous melts, there is no reason to expect that results
gathered from the study of sulphur in anhydrous melts can be
directly applicable to hydrous melts. If this statement could
be rigorously demonstrated, there would be no reasons for performing
the present experimental work.
Whatever the mechanism of sulphur dissolution, our results show that at all values of fO2, the sulphur-carrying capacity of peralkaline rhyolites is 5–20 times greater than their metaluminous equivalents. Increasing the temperature to 900°C increases the Smelt by a factor of two, other parameters being kept equal (Fig. 2). Similarly, varying Sbulk from 0·5 to 6 wt % increases Smelt by 2–2·5 times, as a result of increasing sulphur fugacity (Fig. 2). The increase may also reflect the fact that the melt composition is changing when sulphur is added, as it removes part of the iron in solution to produce sulphide.
View larger version (13K):
[in this window]
[in a new window]
|
Fig. 2. Variation in the melt sulphur content of EBU rhyolite with the bulk S content for two temperatures, both at an fO2 of c. NNO – 1·6.
|
|
Variations with
fO
2 in the sulphur content of fluid (S
fluid)
over S
melt, or the partition coefficient
DS, are shown in
Fig. 1b for rhyolite melts having 1 wt % of S
bulk. Here again, the two
less peralkaline compositions (ND, SMN) show little variation
in
DS with
fO
2,
DS being
270, except at around NNO + 1 when
anhydrite and pyrrhotite crystallize together, such as in charge
ND5, which has a slightly lower
DS (
Table 1). In the most peralkaline
melt (EBU), however, the partition coefficient rises abruptly
at
fO
2 > NNO + 1, behaviour similar to that found in silicic
magmas from volcanic arcs (Scaillet
et al., 1998
). The reason
for the limited variation in
DS in the two less peralkaline
melts is that the increase in sulphur solubility with
fO
2 is
compensated by crystallization of lower proportions of anhydrite
at high
fO
2 relative to pyrrhotite at low
fO
2 (
Table 1). In
contrast, the increase in
DS in the most peralkaline melt is
because it crystallizes slightly higher proportions of sulphide
at low
fO
2, whereas at high
fO
2 this composition does not crystallize
anhydrite in higher modal amounts than ND and SMN. Because its
solubility decreases beyond NNO + 1, this results in an overall
increase in the partition coefficient as
fO
2 rises.
A comparison with experimental results obtained on a natural metaluminous dacite (Scaillet et al., 1998) shows that, under the same T–fO2 conditions, a peralkaline rhyolite (SMN) with bulk iron and sulphur contents similar to those in the dacite crystallizes 30–40 % less sulphide (Fig. 1c). Even the most peralkaline rhyolite, with 7 wt % FeOtot (EBU, Table 1), crystallizes less sulphide than the metaluminous dacite, which has 4·4 wt % FeOtot. We interpret this as resulting from the fact that, when held at the same fO2, the Fe2+/Fe3+ ratio is lower in peralkaline than in metaluminous melts (Gwinn & Hess, 1989; Gaillard et al., 2001), such that there is less Fe2+ available for sulphur complexation in the melt and consequently for sulphide crystallization. The lower modal proportion of sulphide in peralkaline rhyolites results in elevated fluid/melt partition coefficients, in particular for moderately peralkaline melts, even at low fO2, differing in this respect from metaluminous melts (Scaillet et al., 1998). The partition coefficient steadily increases for Sbulk between 1 and 6 wt %, but remains broadly constant for Sbulk between 0·5 and 1 wt % (Fig. 3). As for the solubility trend, this reflects an increase in fS2 perhaps coupled to a change in melt chemistry (which becomes iron-poor as the bulk sulphur content increases). At bulk sulphur contents lower than 4 wt %, a rise in temperature from 800 to 900°C decreases DS by only 10%, as a consequence of the modest increase of sulphur solubility with temperature (Fig. 3). Our results show, therefore, that any petrogenetic process favouring the formation of peralkaline over metaluminous silicic melts will minimize the possibility of sulphur loss via sulphide fractionation.
View larger version (14K):
[in this window]
[in a new window]
|
Fig. 3. Variation in the Sfluid/Smelt partition coefficient with increasing bulk S content for the EBU rhyolite at 800 and 900°C at three fO2 conditions. Two error bars are shown, one for partition coefficients lower than 100 (bottom part) and one for partition coefficients at around 300 (middle part). Each error bar represents an uncertainty of c. 50%. For clarity, the error bar corresponding to data at c. 400 is not reported.
|
|
The increased solubility of sulphur in peralkaline melts is
consistent with the apparent scarcity of modal sulphides in
peralkaline rhyolites. We are aware of three occurrences. Crisp
& Spera (1987)
recorded scarce pyrrhotite microphenocrysts
in comendites and pantellerites of the Tejeda volcano, Gran
Canaria. Mungall & Martin (1996)
noted rare pyrite phenocrysts
in pantellerites of the Pico Alto volcano on Terceira Island,
Azores. Lowenstern
et al. (1993)
found phenocrysts of pyrrhotite
and molybdenite in pantellerites and pantelleritic trachytes
from Pantelleria, Italy, the occurrence of molybdenite being
apparently unique in a silicate magma.
|
SULPHUR CONTENT OF CRUSTALLY AND MANTLE-DERIVED RHYOLITES
|
|---|
General
Before using the above experimental results to constrain the
atmospheric sulphur yields of rhyolites associated with flood
basalts, we need to fix possible limits on the bulk sulphur
contents of rhyolites, which will depend on their source region.
Two end-member petrogenetic models have been proposed bearing
on sulphur enrichment in magmas: crystal fractionation of basaltic
magmas, e.g. the Ethiopian ignimbrites (Ayalew
et al., 2002
),
and partial melting of the lower–middle continental crust
(albeit with variable mantle input), e.g. the Paraná–Etendeka
deposits (Bellieni
et al., 1986
; Garland
et al., 1995
; Harris
& Milner, 1997
; Ewart
et al., 1998
, 2004
). The Ethiopian
rhyolites are peralkaline, whereas the felsic magmas of the
Paraná–Etendeka province are predominantly metaluminous,
being in addition slightly less silicic and more K
2O-rich than
those in Ethiopia. Both types of magma have high iron contents,
however, which is usually attributed to low
fO
2 during magma
genesis (Ewart
et al., 1998
, 2004
). Whatever their mode of origin,
the sulphur content of silicic magmas is controlled by the source
composition during partial melting and by subsequent evolutionary
processes. Sulphur in a magma can reside in fluid, melt and
solids, the last being mostly sulphides under the low
fO
2 thought
to prevail in basalts and in the lower crust. In the following
discussion we first consider a mantle origin (i.e. basalt fractionation)
and then a crustal origin for rhyolites associated with trap
sequences.
Mantle origin
Method
In basalts, sulphide stability largely dictates the possibility of generating a sulphur-rich silicic derivative because, given their high density, sulphides can easily settle out from the host magma. Thus, to produce a sulphur-rich rhyolite supposes that during fractionation the basalt does not crystallize significant amounts of sulphide. Generation of peralkaline rhyolites by fractionation of alkali basalt requires 80–90% crystallization (e.g. Barberi et al., 1975; Ayalew et al., 2002). Alkali basalts typically have bulk H2O contents of at least 1 wt %, CO2 contents sometimes higher than 1 wt %, and sulphur contents that can exceed 1000 ppm (Clocchiatti et al., 1992; Dixon et al., 1997; Bureau et al., 1999; Wallace, 2002). They evolve at fO2 
NNO (Dixon et al., 1997). We have, therefore, calculated the conditions under which alkali basalts become saturated in sulphide during crystallization, for bulk H2O–CO2 contents of 1–2 wt % as inferred from studies of modern analogues (Dixon et al., 1997; Gerlach et al., 2002; Lange, 2002), using a Sbulk of 0·1 wt %, and over a range of plausible redox conditions from NNO – 2 to NNO + 1 (Fig. 4). We also have calculated the corresponding proportions of bulk sulphur partitioned into sulphide and fluid phases (Fig. 5).
View larger version (21K):
[in this window]
[in a new window]
|
Fig. 4. Variations in sulphide activity [in log scale, log(aFeS)] in the melt during the crystallization of a basalt magma for different fO2 values. (a) With 1 wt % H2O and 1000 ppm sulphur. The decrease of aFeS observed at high melt fractions is due to the fact that intermediate compositions are Fe-rich compared with the parent magma, and thus have a higher sulphur solubility (which consequently induces a decrease in aFeS) than the totally molten starting magma. (b) With 1 wt % H2O, 1 wt % CO2 and 1000 ppm sulphur. It should be noted that, at NNO – 2, the magma does not reach sulphide saturation even after 80% crystallization. In all calculations we assume that, once stable, the sulphide melt is made of pure FeS and we thus neglect the role of Ni.
|
|
View larger version (23K):
[in this window]
[in a new window]
|
Fig. 5. Relative amounts of sulphur hosted in the sulphide and in the fluid in a crystallizing basalt with 1000 ppm sulphur. (a) Proportion of bulk sulphur locked in sulphide during the crystallization of a hydrous basalt (1 wt % H2O), calculated for four fO2 values. Under these conditions (i.e. no CO2 present) there is no fluid phase so the remaining sulphur is in the residual melt. (b) Proportion of bulk sulphur dissolved in the fluid during the crystallization of a H2O–CO2-bearing basalt calculated for various fO2 values. The presence of CO2 promotes early fluid saturation and, as a result, a significant part of the sulphur present in the system goes into the fluid phase. The curve labelled 2 wt % H2O + CO2 corresponds to the case where the bulk H2O and CO2 are each set at 2 wt %. All other curves are for 1 wt % H2O and 1 wt % CO2.
|
|
Melt compositions used. We used the basalt–rhyolite sequence
of the Boina centre in the Afar rift (Barberi
et al., 1975
)
as a proxy for the evolution of liquid compositions in a fractionating,
mildly alkaline basalt series. Although we could have used a
thermodynamic model to simulate this evolution, available models
still fail to reproduce the transition from metaluminous to
peralkaline melts, a feature that is critical for the understanding
of sulphur behaviour, as shown below. The petrological and geochemical
study of the Boina volcano concluded that the silicic magmas
were derived from fractional crystallization of transitional
basalt (Barberi
et al., 1975
). Five representative compositions
were used here: G485, S52, D237, D210, D224B (Barberi
et al.,
1975
) (see
Table 2). The degree of crystallization was determined
using the K
2O content, assuming perfectly incompatible K
2O behaviour
during crystallization, a reasonable assumption except for composition
D224B in which K-feldspar crystallizes, and thus for which the
extent of crystallization relative to the parent basalt G485
is a minimum. The following proportions of residual liquids
were calculated: S52, 73%; D237, 37%; D210, 24%; D224B, 18%.
This natural liquid line of descent shows that peralkaline rhyolites
can be produced from basalt crystallization after
c. 80 wt %
crystallization (
Table 3, composition D224B). Temperatures were
calculated assuming a linear relationship between melt fraction
and
T, and using a liquidus of 1200°C for basalt G485 and
890°C for the rhyodacite D210 (
Table 3). Clearly, such an
approximation is unlikely to be valid in detail. We are, however,
interested in fixing the general behaviour of C–H–O–S
volatiles in a crystallizing basalt, for which we need to know
the approximate amount of residual melt available at each temperature
in order to calculate the partitioning of each volatile species
between melt and fluid using mass-balance constraints and solubility
laws (see below). Any departure from a linear trend will either
promote or delay the attainment of volatile and sulphide saturation.
This will not affect, however, the general conclusion derived
from our analysis, which concerns the importance of the role
of CO
2. In addition, we note that our assumption of linearity
predicts a temperature of 861°C for the peralkaline rhyolite
D224B, which is in reasonable agreement with available estimates
for such compositions (Bizouard
et al., 1980
; Ayalew
et al.,
2002
; Nekvasil
et al., 2004
). All calculations descibed below
were performed for a pressure of 150 MPa, as our experimental
results were collected at this pressure.
Solubility and activity models. For a given initial volatile
content and residual melt fraction, the amount of volatiles
dissolved in the melt was calculated using the solubility models
for H
2O and CO
2 of Dixon
et al. (1995)
. In the absence of appropriate
solution models for both H
2O and CO
2 in intermediate melt compositions,
those for basalt were used. For sulphur we use the solubility
model of Scaillet & Pichavant (2005)
. The latter is an empirical
model that allows us to relate the melt sulphur content to the
sulphur fugacity, for a variety of melt compositions, including
hydrous basalts. This model was derived in an attempt to evaluate
the behaviour of sulphur in hydrous mafic magmas, for which
there are few experimental data [apart from those of Luhr (1990)
].
The model has the form
where
S is the total sulphur concentration in ppm,
P the pressure
in bars,
T the temperature in °C,
NNO and
FFS are the referenced
fO
2 and
fS
2 against the Ni–NiO and Fe–FeS solid
buffers respectively,
Wi represents the weight % of oxide
i,
and
a,
b,
c,
d,
e,
f and
gi are fitted parameters (
Table 4)
that were obtained by linear regression of the experimental
databases of Luhr (1990)
, O'Neill & Mavrogenes (2002)
and
Clemente
et al. (2004)
. The summation is carried out over all
major oxides, including FeO, Fe
2O
3, OH
– and H
2O. This
third-order polynomial function is necessary to reproduce the
inverted bell-shaped pattern of sulphur solubility in silicate
melts (e.g. Carroll & Webster, 1994
; Clemente
et al., 2004
),
and the crossed
fO
2–
fS
2 term is needed to take into account
the effect of varying
fS
2 on the relationship between
fO
2 and
S in melt (Clemente
et al., 2004
). This model encompasses a
large SiO
2 range (35–80 wt %), and reproduces measured
fS
2 within an average of 0·65 log unit, over more than
15 log units when normalized to the FFS solid buffer. Finally,
it should be noted that the model of Scaillet & Pichavant
(2005)
is calibrated only for metaluminous compositions, which
implies that sulphur solubilities of intermediate or silicic
peralkaline compositions corresponding to any given
fS
2 are
underestimated by this model and thus that sulphide activities
calculated here for those melts are overestimated (see below).
For any
fO
2 and
fS
2, the activity of FeS in the melt was calculated
from the following equilibrium, using thermodynamic data from
O'Neill & Mavrogenes (2002)
:
The activity of FeO in the melt was calculated by
determining the Fe
2+/Fe
3+ ratio of the melt using the Kress
& Carmichael (1991)
method and an activity coefficient for
FeO of 1·4 (O'Neill & Mavrogenes, (2002
). For any
melt in which the calculated FeS activity was found to exceed
unity, the melt sulphur content was fixed to that corresponding
to an
aFeS = 1, and the excess sulphur was converted to sulphide
(we thus neglect the role of additional elements such as Ni
or Cu that could promote early saturation in sulphide).
The fluid phase composition was calculated using a Modified Redlich–Kwong type equation of state (MRK), using as input parameters fH2O, fCO2 and fS2. It must be stressed that, regardless of our current knowledge or body of experimental constraints on the sulphur content of fluids in magmas, the sulphur content of a fluid coexisting with a silicate melt can be calculated using a thermodynamic approach, provided that appropriate solubility laws exist for the main volatile species. This stems from the fact that in the C–O–H–S system, once fH2O, fCO2 and fS2 are fixed (at P and T), the fugacities of all other volatile species are fixed as well (Holloway, 1987; Scaillet & Pichavant, 2003, 2004). In other words, the composition of the fluid is uniquely defined, including its sulphur content. This allows us to compute the partition coefficients of sulphur between fluid and melt, as the melt sulphur content is fixed by fS2 (and fO2). The accuracy of such an approach relies, among other things, on our knowledge of the thermodynamic properties of C–O–H–S fluids, which appear to be reasonably well known (e.g. Shi, 1992), at least in the low-pressure range (<1000 MPa). Using this approach, Scaillet & Pichavant (2003) showed that there was a good overall agreement between calculated and measured fluid compositions for silicic arc magmas. The same method has been applied to active basaltic volcanoes (Scaillet & Pichavant, 2005), for which remote sensing of volcanic gases can be used to constrain the gas chemistry at depth. In this case too, generally good agreement is observed (Scaillet & Pichavant, 2005). Therefore, although we recognize that the current experimental database on the sulphur content of fluids of mafic magmas is almost non-existent, this gap can be partly circumvented by using a thermodynamic approach, which, when compared with independent estimates, appears to retrieve the correct order of magnitude in terms of the sulphur content of magmatic fluids.
Procedure. For any given initial H2O, CO2 and S contents, the equilibrium distribution of those volatiles between melt, fluid and sulphide was calculated via an iterative procedure by finding the fugacities that satisfy the following two sets of conditions: first, the condition of chemical equilibrium:
second, the mass-balance constraints:
The mass-balance equations also take
into account the amounts of C, S and H in other species such
as H
2S, SO
2, CO and CH
4. In the H
2O-only case, with a bulk H
2O
content of 1 wt %, we considered that the crystallizing basalt
does not reach fluid-saturated conditions, as the H
2O content
of the residual melt reached after 82% crystallization is 5·65
wt % (
Table 3), which is on the verge of H
2O saturation of rhyolitic
melts at 150 MPa (e.g. Zhang, 1999
). In this case, the sulphur
was partitioned between melt and sulphide only.
Let us consider, as an example of the calculation procedure, the case of a basalt carrying only H2O and sulphur crystallizing at an fO2 of NNO – 2. At 890°C, the basalt has 24·46 wt % residual liquid with a composition of D210 (Table 3). If sulphur had a perfectly compatible behaviour (bulk S of 1000 ppm), then the melt composition D210 would have 4088 ppm of dissolved sulphur (1000/0·2446). Using the model of Scaillet & Pichavant (2005), this amount of dissolved sulphur would correspond to an fS2 of 0·035 MPa, or to an aFeS = 18 when calculated using equilibrium (2) and the assumptions given above. Clearly, there is too much sulphur in solution and some of it must be withdrawn to decrease the aFeS to unity, because by definition the activity of FeS cannot increase beyond unity (for appropriate standard states for solids and liquids at P and T). The next step is thus to remove the excess sulphur in solution in the melt until the calculated fS2 corresponds to an aFeS = 1, and convert that excess sulphur into immiscible/solid sulphide [considering also the constraints set out in equations (3)–(8)]. For that specific case, using equation (1), we calculate that saturation of melt composition D210 with sulphide (aFeS = 1, because we perform an equilibrium calculation, aFeS in the silicate liquid is equal to that in the sulphide) is achieved when fS2 = 10–4 MPa at 890°C and NNO – 2, which corresponds to a melt sulphur content of 850 ppm. In other words, this means that of the 1000 ppm sulphur dissolved in the molten basalt at 1200°C, 208 ppm (850 x 0·2446) are dissolved in the residual melt at 890°C, the remainder (792 ppm = 1000 – 208) being locked up in sulphide, as under these conditions no fluid is present [Table 3; the melt water content is below the saturation value of an andesitic melt at 150 MPa (Burnham, 1979)].The introduction of CO2 into the system promotes early fluid saturation and thus the sulphur content of melt at 890°C (i.e. when there is only 24·46 wt % residual melt) must be lower than when only H2O is present. For instance, implementing the example above with addition of CO2 (i.e. a basalt with 1 wt % H2O and 1 wt % CO2 at NNO – 2), we calculate that at 890°C the sulphur content of the residual melt is 635 ppm. Finally, as noted above, peralkaline melts dissolve more sulphur than metaluminous types, so our calculated sulphide proportions should be considered maxima for the most fractionated melt (D224B, Table 3). Assuming, on the basis of our experimental results, that the peralkaline melt D224B dissolves five times more sulphur than that calculated by the model of Scaillet & Pichavant (2004), or 4250 ppm, then it follows that such compositions would be barely saturated at NNO – 2, the proportion of sulphide being low. This implies that the amounts of CO2 needed to scavenge the sulphur toward the fluid, which under our P–T conditions is calculated to be 2 wt % (see below), must be considered maximum values where peralkaline derivatives are produced.
Results
The results of our calculations are shown on Figs 4–6. We have considered three cases corresponding to different bulk volatile contents: case (1), 1 wt % H2O and 1000 ppm sulphur; case (2), 1 wt % H2O, 1 wt % CO2 and 1000 ppm sulphur; case (3), 2 wt % H2O, 2 wt % CO2 and 1000 ppm sulphur. Figure 4a shows the evolution of aFeS of a basalt with degree of crystallization, calculated at four fO2 values. An H2O-bearing (1 wt %) but CO2-free alkali basalt crystallizing at or above NNO – 1 is sulphide-saturated after c. 60% of crystallization. At NNO – 2, saturation in sulphide is slightly delayed, to 70% crystallization. Figure 5a shows the evolution of the proportion of bulk sulphur sequestered in sulphide with degree of crystallization, corresponding to the calculations shown in Fig. 4a. By the time the residual melt is peralkaline (20–30% liquid; see Table 2), it can be seen that 90 wt % of Sbulk is locked up in sulphide (Fig. 5a). Further crystallization results in massive sulphide precipitation; after 82% crystallization more than 80 wt % of Sbulk is locked up in sulphide. The calculations corresponding to case (2) illustrate the role of CO2. When this volatile is introduced, the sulphur behaviour is dramatically altered because fluid saturation occurs at an early stage, owing to the low solubility of CO2 in silicate melts (e.g. Dixon et al., 1995) and chemical equilibrium demands that sulphur is also partitioned into the fluid, thus lowering Smelt and the activity of sulphide. With a bulk CO2 content of 1 wt % and 1 wt % H2O [case (2)], similar to the bulk content inferred for Kilauean or Etnean basalts (Clocchiatti et al., 1992; Gerlach et al., 2002), the calculations show that sulphide saturation is slightly delayed compared with the H2O-only case (Fig. 4). However, the main difference relative to the CO2-free situation [case (1)] is that after 80% crystallization, at least 60% of Sbulk is in the fluid phase, whatever the prevailing redox conditions, rising to 90% for fO2 = NNO + 1 (Fig. 5b). Finally, case (3) shows that at fO2 = NNO – 2, for initial CO2 and H2O contents of 2 wt % each, a crystallizing alkali basalt remains below sulphide saturation, even after 80% crystallization when derivative liquids are rhyolitic (Fig. 6). Under these conditions 95% of Sbulk is hosted by the fluid, even after 80 wt % of crystallization (Fig. 5b).
View larger version (17K):
[in this window]
[in a new window]
|
Fig. 6. Effect of bulk H2O and CO2 contents on the sulphide activity (log scale) of a crystallizing basalt at an fO2 of NNO – 2. It should be noted that when there is a large amount of CO2 in the magma (i.e. the 2 wt % case), the coexisting fluid can scavenge a significant proportion of the bulk sulphur such that the system does not reach saturation in sulphide even after 80% of crystallization.
|
|
We stress that such elevated bulk volatile contents have been
inferred for some alkali basalts (Dixon
et al., 1997
), including
those in flood sequences (Lange, 2002
). In addition, recent
experimental data on the generation of silica-saturated alkalic
suites from alkali basalts suggest that the latter contain at
least 0·5 wt % H
2O, and possibly up to 2 wt % (Nekvasil
et al., 2004
), which reinforces the view that trap basalts may
be considerably richer in fluids than those at mid-ocean ridges.
The above calculations show that, provided they are CO
2- and
H
2O-rich, basaltic magmas can readily produce sulphur-rich derivatives,
regardless of the imposed redox conditions. The amount of CO
2 appears to be the key parameter, as it triggers early volatile
exsolution, which prevents sulphide precipitation. Without CO
2,
fO
2 exerts a prime control on the amount of sulphide crystallization
in basalt with a certain initial sulphur content. A reduced
CO
2-poor basaltic magma will yield silicic derivatives that
are sulphur-poor owing to extensive sulphide precipitation,
whereas a basalt crystallizing at or above NNO + 1, as in arc
settings, may give rise to sulphur-rich silicic magmas (Scaillet
& Pichavant, 2003
; Scaillet
et al., 2003
). The
TOP
ABSTRACT
INTRODUCTION
