JOURNAL OF
Journal of Petrology, 2015, Vol. 56, No. 7, 1407–1424
doi: 10.1093/petrology/egv041
Original Article
PETROLOGY
The Effect of FeO on the Sulfur Content at
Sulfide Saturation (SCSS) and the Selenium
Content at Selenide Saturation of Silicate Melts
Jeremy L. Wykes*, Hugh St. C. O’Neill and John A. Mavrogenes
Research School of Earth Sciences, Australian National University, Acton, ACT, 0200, Australia
*Corresponding author. Present address: Australian Synchrotron, 800 Blackburn Road, Clayton, VIC,
3168, Australia. Telephone: þ61 3 8540 4277. E-mail: [email protected]
Received September 26, 2014; Accepted July 8, 2015
ABSTRACT
The concentration of sulfur in basalt-like silicate melts as S2– is limited to the amount at which the
melt becomes saturated with a sulfide phase, such as an immiscible sulfide melt. The limiting solubility is called the ‘sulfur content at sulfide saturation’ (SCSS). Thermodynamic modelling shows
that the SCSS depends on the FeO content of the silicate melt from two terms, one with a negative
dependence that comes from the activity of FeO in the silicate melt, and the other with a positive
dependence that comes from the strong dependence of the sulfide capacity of the melt (CS) on FeO
content. The interaction between these two terms should yield a net SCSS that has an asymmetric
U-shaped dependence on the FeO content of the melt, if other variables are kept constant. We have
tested this thermodynamic model in a series of experiments at 1400 C and 15 GPa to determine
the sulfur contents at saturation with liquid FeS in melt compositions along the binary join between
a haplobasaltic composition and FeO. The SCSS is confirmed to have the asymmetric U-shaped dependence, with a minimum at 5 wt % FeO. The effect of FeO on the selenide content at selenide
saturation (SeCSeS) was investigated in an analogous fashion. SeCSeS shows a similar, though
not identical, U-shaped dependence, implying that the solubility mechanism of selenide in basaltlike silicate melts is similar to that of sulfide. The observation of increasing SCSS with decreasing
FeO in hydrous silicic melts was explored by inverse modelling of datasets from pyrrhotitesaturated hydrous silicic liquids, revealing that high SCSS at low FeO can be explained in terms of
the low-FeO limb of the ‘U’, rather than dissolution of sulfur as hydrous species such as H2S or
HS–. Recent measurements of the composition of the surface of Mercury prompted examination
of the high-SCSS, low-FeO limb of the ‘U’ as a potential explanation for the sulfur-rich but Fe-poor
surface of Mercury.
Key words: silicate melt; sulfur; selenium; sulfide capacity; selenide capacity; sulphide
INTRODUCTION
2–
The solubility limit of sulfur dissolved as sulfide (S )
in silicate melts is reached when the melt becomes
saturated with an immiscible sulfide phase, such as an
FeS-rich liquid, often found as globules in ocean floor
basaltic glasses (Mathez, 1976; Czamanske & Moore,
1977; Francis, 1990; McNeill et al., 2010; Patten et al.,
2012). The term ‘sulfur content at sulfide saturation’ or
SCSS is employed to explicitly represent the solubility
of sulfur, as sulfide, S2–, in sulfide-saturated silicate
melts (Shima & Naldrett, 1975). The thermodynamic
model of O’Neill & Mavrogenes (2002) predicts that, at
1 atm pressure, the value of SCSS in equilibrium with
stoichiometric FeS melt exhibits an asymmetric
U-shaped dependence on the FeO content of the silicate
melt, caused by the balance between two competing reactions: (1) the positive effect of FeO in enhancing sulfur
solubility, which predominates at high FeO, versus (2)
C The Author 2015. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: [email protected]
V
1407
1408
the negative effect of FeO in combining with S2– dissolved in the melt to form immiscible FeS (solid or liquid), which predominates at low FeO.
SCSS is dependent on the Fe:S ratio of the saturating
sulfide liquid, the oxygen content of the sulfide liquid,
and substitution of other chalcophile elements such as
Ni and Cu into the sulfide liquid. When the activity of
FeS in the sulfide liquid is constant, then SCSS is independent of f O2 and f S2. Furthermore, the dependence
of SCSS on the FeO content of the silicate liquid requires an unchanging Fe2þ/Fe3þ ratio for SCSS to be independent of f O2. The O’Neill & Mavrogenes (2002)
model was developed for silicate melts in which iron is
approximated as entirely Fe2þ, and thus is valid only at
reducing f O2 conditions. O’Neill & Mavrogenes (2002)
did not demonstrate the low-FeO limb of the U-shaped
dependence experimentally, because the high f S2
required to achieve FeS saturation at low FeO cannot be
accessed in 1 atm experiments as partial pressures of
S2 greater than 1 bar are necessary. Limited experimental demonstration [an exception is the study by
Tsujimura & Kitakaze (2005)] may help explain why the
thermodynamic basis of SCSS has been ignored in favour of empirical parameterization of the available experimental data (e.g. Li & Ripley, 2005, 2009; Liu et al.,
2007). An inherent limitation of empirical parameterizations is that they may provide inaccurate results when
applied outside the range of experimental conditions on
which they are calibrated (Baker & Moretti, 2011).
The results of experiments conducted over a sufficiently large range of FeO contents to demonstrate
unambiguously the asymmetric U-shaped dependence
of SCSS on the FeO content of a basaltic silicate melt
are reported in this contribution. The experimental approach developed for sulfide was then applied to determining the ‘selenium content at selenide saturation’ or
SeCSeS.
Recent work has suggested that selenium does not
degas from some magmas during eruption (Jenner
et al., 2010) under conditions at which sulfur is lost (e.g.
Lesne et al., 2011), thus allowing the original undegassed S contents to be inferred from assumed S–Se
ratios. This may be related to the much lower stability
of SeO2 (gas) relative to Se2– dissolved in silicate melt,
versus the stability of SO2 (gas) relative to S2– in silicate
melt (Jenner et al., 2010). At present, there are no experimental data on the behaviour of reduced selenium
(selenide, Se2–) in silicate melts of geologically relevant
compositions. The study of selenium in silicate melts
has been restricted to compositions relevant to glass
making, where it has a role in colouring (Weyl, 1959;
Rüssel, 2000; Müller-Simon et al., 2001); fayalitic and
calcium ferrite slag compositions involved in smelting
of copper anode slimes (Nagamori & Mackey, 1977;
Fang & Lynch, 1987; Johnston et al., 2007, 2010); and
borosilicate glasses for high-level radioactive waste immobilization, where 79Se is a long-life fission product
(Schreiber et al., 1988; Schreiber & Schreiber, 1993;
Bingham et al., 2011).
Journal of Petrology, 2015, Vol. 56, No. 7
Selenium solubility in silicate melts cannot be easily
investigated by conventional gas-mixing methods such
as those used by O’Neill & Mavrogenes (2002) owing to
(1) the unavailability of SeO2 as a commercial gas—
SeO2 is a solid at room temperature, which sublimes at
350 C, and (2) the intractable prospect of using highly
toxic H2Se gas in a laboratory gas-mixing furnace.
Fortunately, there is no experimental impediment to
measuring the SeCSeS at high pressure, in an analogous manner to SCSS experiments. Here, we also show
that SeCSeS exhibits a similar asymmetric U-shaped
dependence on FeO content, with some subtle differences arising from the different thermochemical properties of Se and S.
Thermodynamic background
At oxygen fugacities sufficiently reducing that sulfur
occurs in silicate melts as S2– and iron as Fe2þ, sulfur
has been demonstrated to dissolve in silicate melts by
directly replacing oxide anions (O2–) on the anion sublattice, as described by the reaction
05 S2ðgasÞ þO2– ðsilicate
liquidÞ $ S
2–
ðsilicate liquidÞ þ05
O2ðgasÞ
(1)
(Fincham & Richardson, 1954; O’Neill & Mavrogenes,
2002). The pseudo-equilibrium constant for equation (1)
can be expressed
liquid
ln Kð1Þ
¼ ln aSsilicate
þ 0:5 ln fOgas
0:5 ln fSgas
2
2
2
silicate liquid
ln aO
2
(2)
Equation (2) is referred to as a pseudo-equilibrium
constant as it includes charged species. The abundance
of O2– ions in silicate melts is vastly greater than that of
liquid)
S2– (i.e. ln a(silicate
0), such that the term may be
O2–
omitted [see Ariskin et al. (2013) for an alternative viewpoint]. The expression can be thus simplified:
lnCS ¼ ln ½S þ 0:5 ln f O2 =f S2 :
(3)
This pseudo-reaction constant (CS) is referred to as
the ‘sulfide capacity’ of a silicate melt, and [S] is the sulfur content of the silicate melt, given in parts per million
by weight. This relationship was first proposed by
Fincham & Richardson (1954) after investigation of Fefree CaO–Al2O3–SiO2 melts at 1 bar, and was tested by
the experimental study of O’Neill & Mavrogenes (2002),
who confirmed the Fincham–Richardson relationship.
O’Neill & Mavrogenes (2002) used the reciprocal solid
solution formalism for mixing of cations and anions on
different sublattices to develop a model for CS with the
form
lnCS ¼ A0 þ R XM AM
M
(4)
where CS is the sulfide capacity, XM terms represent the
mole fraction of cation M, AM coefficients represent the
preference of a metal for sulfur as a neighbour over
oxygen, and A0 is a constant combining the conversion
factor between weight per cent and mole fraction with
Journal of Petrology, 2015, Vol. 56, No. 7
1409
the activity coefficient for the sulfur species in the melt.
The model was generated from over 150 experiments
at 1 atm involving silicate melt compositions in the
system CMAS 6 Ti 6 Fe 6 Na 6 K. The O’Neill & Mavrogenes (2002) parameterization of ln CS revealed that the
FeO content of the silicate melt was by far the most
dominant control on sulfur solubility, with the coefficient for Fe (26) at least three times larger than the
next most important cation, Ca (75), at 1400 C.
The sulfur content of a silicate melt in equilibrium
with an iron sulfide phase (solid or liquid) can be
described by
FeOðsilicate liquidÞ þ 0 5 S2ðgasÞ $ FeSðsulfide liquidÞ þ 0 5 O2ðgasÞ
(5)
for which the equilibrium constant can be expressed as
DGð5Þ
RT
sulfide liquid
silicate liquid
ln aFeO
þ 0:5 ln
¼ ln aFeS
fO2
: (6)
fS2
When expression (3) above is combined with (5), the
f O2 and f S2 terms are eliminated to produce
ln ½SSCSS ¼
DGð5Þ
RT
sulfide liquid
silicate liquid
ln aFeO
þ ln CS þ ln aFeS
(7)
where [S]SCSS represents the ‘sulfur content at sulfide
saturation’. Expression (7) can be further extended by
including the pressure term from Mavrogenes & O’Neill
(1999) to produce
ln ½SSCSS ¼
DGð5Þ
RT
sulfide liquid
þ ln CS þ ln aFeS
silicate liquid
ln aFeO
þ
C P
T
(8)
where C is a constant (for a given melt composition)
representing the bulk PDV component of the terms
of (7):
C ¼ ð–DVCs þ DV ð 5Þ þ D V FeO Þ=R
determined from the SCSS experiments of this study.
The DG (5) value (8474 T ln T J mol–1) is taken from
O’Neill & Mavrogenes (2002), and the C parameter of –
0047 K bar–1 determined by Mavrogenes & O’Neill
(1999) for a basaltic melt. It should be noted that
whereas DG (5) is expressed in terms of T, the AM coefficients have been determined only for 1400 C, preventing application of the model to other temperatures. The
model presented in Fig. 1 shows that at constant P, T
and FeO-free silicate melt composition (herein referred
to as ‘matrix’; Doyle & Naldrett, 1987), the negative ln
(silicate melt)
aFeO
term dominates [S]SCSS at low FeO contents, and the ln CS term dominates at high FeO content,
producing a characteristic asymmetric U-shaped curve
for the dependence of [S]SCSS on silicate melt FeO content. For the range of FeO contents commonly seen in
terrestrial basalts, (i.e. 4 wt % FeO) the model predicts
increasing SCSS with increasing FeO.
(9)
P is pressure in bars, and T is temperature in Kelvin.
Consideration of (8) reveals that SCSS is independent of f O2 and f S2 (whenever all dissolved S is S2–, Fe is
dominantly Fe2þ, and a(sulfide)
remains 1), and is deFeS
pendent on the FeO content of the silicate melt in two
terms, namely the XFeAFe term from (4) and the ln
(silicate melt)
aFeO
term from (6). The XFeAFe term (and thus ln
CS) increases with increasing FeO content of the melt,
(silicate melt)
whereas –ln aFeO
has a negative, concave-upwards slope versus the FeO content of the melt.
In Fig. 1, ln [S]SCSS and its component terms are plotted versus the FeO content (weight per cent) of a silicate
melt at 1400 C and 15 GPa. In the calculations, the activity of FeS is assumed to be unity, the activity coefficient for FeO(silicate) is 13 [the average of all
determinations from O’Neill & Eggins (2002) for CMAS
and CMASTi melts], and the silicate melt composition
is the average, normalized, FeO-free composition
METHODS
Starting material and encapsulation
The silicate starting material was a sintered oxide mix
of the following nominal composition: 530 wt % SiO2,
220 wt % Al2O3, 105 wt % CaO, 105 wt % MgO, 15 wt
% TiO2 and 25 wt % Na2O. Reagent or analytical grade
oxides and carbonates were ground under acetone in
an agate mortar and pestle before pressing into pellets
and firing at 1100 C in air for 16–24 h. FeO was subsequently added as synthetic wüstite (nominally FeO, but
almost certainly nonstoichiometric), which was produced by firing synthetic hematite at 1300 C under a
30% CO2–70% CO atmosphere. Mixes for Se experiments were prepared using synthetic fayalite as a
source of FeO (produced by firing reagent Fe2O3 and
SiO2 at 1300 C under a 50% CO2–50% CO atmosphere).
Fayalite was added to an aliquot of SiO2-deficient, FeOfree base mix to create an 18 wt % FeO mix.
Intermediate FeO contents were achieved by mixing the
FeO-free and 18 wt % FeO mixes.
The majority of SCSS experiments were contained in
graphite-lined 35 mm diameter welded Pt capsules,
whereas a few experiments were conducted using Re
capsules (see below), as detailed in Table 1. Médard
et al. (2008) suggested that the graphite-in-Pt technique
produces f O2 conditions equivalent to graphite–CO
(CCO) – 07 [iron wüstite (IW) þ 15 or fayalite–magnetite–quartz (FMQ) – 22] at 1360 C and 15 GPa. At such
conditions, Fe2þ will dominate over Fe3þ (Schreiber,
1987; Kress & Carmichael, 1991; Berry et al., 2003; Wilke
et al., 2004); all sulfur (Fleet et al., 2005; Métrich et al.,
2009; Jugo et al., 2010; Klimm et al., 2012) and all selenium (Wykes, 2014) occur in the 2 valence state. The
maximum f O2 attainable with such a capsule assembly
is CCO, where the silicate melt is saturated with CO2 vapour, pure except for the necessary dissociation to CO
and O2. Under such conditions f O2, fCO and fCO2 are all
at their maxima for equilibrium with graphite (Ulmer &
Luth, 1991; Jakobsson & Oskarsson, 1994). Thus, if Fe
1410
Journal of Petrology, 2015, Vol. 56, No. 7
Fig. 1. Component terms of the O’Neill & Mavrogenes (2002) model calculated using the average of normalized FeO-free compositions determined from the SCSS experiments of this study.
Table 1: EPMA analyses of silicate glasses saturated with FeS liquid (1400 C, 15 GPa)
Run ID
Na2O
MgO
Al2O3
SiO2
CaO
TiO2
FeO
S
Total
n
capsule material
D0094
C1473
D0093
D0052
D0065
D0064
D0066
D0055
D1471
D0067
D0068
D0095
D0057
D0058
D0063
273(48)
261(34)
341(6)
262(45)
240(42)
239(35)
241(45)
242(47)
380(49)
233(38)
258(70)
252(48)
253(67)
222(14)
195(12)
1039(15)
1040(29)
1055(6)
1038(5)
992(12)
999(13)
984(14)
969(9)
953(3)
954(8)
927(9)
922(5)
899(11)
854(15)
736(13)
2068(53)
2076(84)
2102(20)
2101(21)
2013(27)
2000(25)
1975(33)
1951(14)
1999(19)
1923(11)
1886(21)
1868(22)
1837(25)
1667(21)
1414(18)
5305(59)
5347(89)
5369(39)
5360(48)
5134(61)
5104(26)
5050(36)
4937(34)
4865(34)
4874(66)
4815(45)
4775(43)
4677(65)
4226(30)
3572(25)
1048(28)
1057(38)
1088(6)
1077(11)
1025(31)
1021(20)
1006(8)
991(11)
999(10)
981(6)
970(26)
958(9)
944(24)
877(11)
741(9)
141(4)
143(5)
144(1)
145(5)
140(5)
138(4)
138(4)
135(3)
136(2)
133(4)
133(3)
128(3)
123(3)
113(20)
092(16)
060(5)
077(5)
091(7)
171(5)
377(9)
443(16)
588(10)
776(9)
810(10)
811(11)
1058(10)
1127(13)
1311(10)
1929(14)
3083(23)
04313(99)
04007(135)
03490(115)
02100(99)
01202(164)
00973(247)
01139(211)
01427(168)
01710(91)
01265(128)
01466(69)
01760(89)
02425(72)
03042(361)
05663(698)
9977
10041
10225
10175
9933
9954
9993
10015
10159
9922
10054
10041
10067
9909
9887
16
15
5
16
15
16
13
14
12
15
16
15
16
10
11
Re
Re
Re
Re
graphite-Pt
graphite-Pt
graphite-Pt
graphite-Pt
graphite-Pt
graphite-Pt
graphite-Pt
Re
graphite-Pt
graphite-Pt
graphite-Pt
Values expressed in weight per cent. Numbers in parentheses are 1SD in the last digit.
loss to the Pt capsule does occur, it will result only in an
increase of f O2 up to the maximum attainable, the CCO
buffer. Further Fe loss will merely produce more CO2rich vapour. Experiments were attempted where the
system was deliberately saturated with CO2 by
employing un-decarbonated starting materials. Such
experiments invariably resulted in melting of the Pt capsule, indicating that the resulting vapour had sufficient
sulfur transport capacity to melt the platinum capsule.
That such reaction between sulfur and the platinum
Journal of Petrology, 2015, Vol. 56, No. 7
capsule was not observed in the experiments employing decarbonated mixes, along with the absence of any
vesicles in the quenched silicate glass, supports the
suggestion that the experimental system is vapour
undersaturated, except in experiments at low FeO content (see below).
Sulfur was added to experiments in a sandwich arrangement, with layers of reagent FeS loaded above
and below the silicate melt. Each layer was 25% of the
mass of silicate material, giving an overall silicate liquid
to sulfide liquid mass ratio of unity. When this approach
was used for Se-bearing experiments, it resulted in selenide melt forming a single large mass, which invariably contained significant quantities of Pt. On many
occasions the selenide melt migrated out of the graphite capsule, and dissolved portions of the Pt capsule
wall, and could be identified dyking through the surrounding MgO in backscattered electron images. The
results of such failed experiments are not reported
here. Subsequently, reagent FeSe was added to the
oxide mixes to give an Se content of 6000 ppm, an
amount (initially inferred by analogy with SCSS and
later demonstrated) sufficient to saturate the silicate
melt at all FeO contents investigated. Two additional experiments were conducted with 12 000 and 24 000 ppm
Se (as FeSe) to investigate the effect of the selenide liquid to silicate liquid ratio.
Sulfur-bearing experiments conducted in graphitein-Pt capsules at very low FeO contents in the silicate
melt were also unsuccessful owing to the melting of the
outer Pt capsule. This is due to the development of an
S2-rich vapour phase, as the relationships presented in
the Introduction imply that sulfide saturation at low FeO
contents in the silicate melt necessitates either high f S2
or low f O2. The low f O2 scenario necessitates the immiscible sulfide melt departing from FeS stoichiometry
towards an Fe–FeS composition (not observed in our
experiments), which is only possible, considering mass
balance in the closed system of a welded Pt capsule, by
developing an S2-rich vapour phase. The development
of an S2-rich vapour phase would therefore seem
inevitable.
To access sulfide saturation at low FeO contents in
the silicate melt, necessary to demonstrate the
U-shaped curve unambiguously, some experiments
were conducted in Re capsules. These were fabricated
by EDM (electrical discharge machining) of 40 mm
diameter Re rod, and consist of a simple flat-bottomed
‘bucket’ capsule with a disc of Re as the lid. The charge
was loaded into the Re capsules in the same geometry
as into the graphite capsules, with two layers of FeS
sandwiching the silicate melt. The f O2 of these experiments is constrained only to be below the Re–ReO2 buffer, (equivalent to DNi–NiO þ 16 at 1400 C, 15 GPa).
High-pressure method
All experiments were conducted at 1400 C and 15 GPa
using an end-loaded piston cylinder apparatus with a
127 mm cylinder bore. A NaCl–Pyrex pressure medium
1411
surrounded a graphite heater, which enclosed 66%
dense MgO pieces surrounding the capsule and
thermocouple. Teflon foil formed a low-friction interface
between the pressure medium and the pressure vessel
wall. Temperatures were monitored via a Eurotherm
PID controller using Pt94Rh6–Pt70Rh30 (Type B) thermocouples (Klemme & O’Neill, 2000) housed inside twobore mullite tubing, with the lower 10 mm replaced with
high-purity Al2O3 tubing. No correction was applied for
the effect of pressure on thermocouple e.m.f.
Experiments were performed by pressurizing twothirds to three-quarters of the final pressure at room
temperature (via a hand pump) before heating at 150 C
min–1 to 600 C, at which stage the Pyrex glass begins
to soften, followed by raising P and T simultaneously,
with the final pressure typically being reached by
1000 C. No friction correction was applied, as the RSES
NaCl–Pyrex–MgO assembly is considered frictionless at
the temperature and duration of the present experiments. Pressure was maintained within 50 MPa of the
nominal value during experiments. Experiment durations were 6 h for sulfur-bearing experiments, and 36 h
for most selenium-bearing experiments. Liu et al. (2007)
demonstrated that 6 h is sufficient time to reach equilibrium in similar experiments at 1250 C and 10 GPa;
Holzheid & Grove (2002) suggested that S contents of
silicate liquids saturated with sulfide liquid become constant after 3 h at 1450 C and 10 GPa; Mavrogenes &
O’Neill (1999) observed constant S contents in glass
for experiments run 4 h or longer at 1400 C and
P > 05 GPa. Experiments were quenched automatically
by the Eurotherm PID controller end of the program. No
attempt was made to ensure isobaric quenching.
Capsules were extracted from the sample assembly
and cast in epoxy resin. For Se-bearing experiments the
surrounding MgO and thermocouple were typically
kept intact and cast along with the capsule. Epoxy discs
were sectioned with a diamond saw to produce two
mounts, each with longitudinal sections of the capsule.
After sectioning the cut surface was vacuum impregnated with epoxy resin, followed by grinding on SiC
paper and polishing with diamond grit (6, 3, 1 and
025 mm) on alumina and cloth laps.
Analytical methods
Sulfur contents of SCSS experiments were analysed via
electron microprobe analysis wavelength-dispersive
spectrometry for sulfur using two methods, one using a
Cameca CAMEBAX instrument employing the method
of O’Neill & Mavrogenes (2002) and the other using the
Cameca SX100 employing a similar peak integral
method to that of Jenner & O’Neill (2012), using barite
as the S standard. Both datasets agree to within 1
standard deviation (1SD), and therefore have been
treated as a single dataset, largely because two mounts
(D0058 and D0063) are no longer available for re-analysis by the SX100. Basaltic glass standard VG2 was
used as a secondary standard for sulfur, and was analysed several times per session. O’Neill & Mavrogenes
1412
(2002) reported an average S content of VG2 of
1403 6 31 ppm. A total of 21 analyses of VG2 using the
SX100 during this study produced an average of
1392 6 61 ppm S. Major elements for the SCSS experiments are taken from the SX100 dataset, for which analytical conditions comprised 15 kV accelerating voltage,
40 nA beam current and a 10 mm beam. Major elements
for D0058 and D0063 are taken from the CAMEBAX
dataset employing the method described by O’Neill &
Mavrogenes (2002).
Selenium-bearing experiments were analysed for Se
using a Cameca SX100 located at the University of
Oregon, a 20 kV accelerating voltage, 40 nA beam current and a 10 mm spot size, with a counting time on the
Se La peak of 300 s. Zinc selenide was employed as the
Se standard. Backscattered electron images were employed to avoid compromising the analysis by including
selenide melt blebs. Analyses with obvious contamination, identified by elevated Se coincident with elevated
Fe–Si or Fe–Ca ratios, were omitted.
Silicate glasses from SeCSeS experiments were analysed for major elements using a JEOL JSM6400 SEM
(scanning electron microscope) equipped with an
Oxford Link-ISIS Pentafet energy-dispersive spectrometry (EDS) system. Beam conditions were an accelerating voltage of 15 kV, a beam current of 1 nA, and a
livetime of 100 s. Silicate glasses were analysed by rastering an 150 nm diameter beam over an area of
15 mm 15 mm, resulting in an average current density of 00012 nA mm–2 (see Morgan & London, 2005).
Sodium loss has been demonstrated to be insignificant
under such analytical conditions (see Wykes, 2014).
Each experiment was analysed 6–15 times. The volcanic
glass standard VG2 was used as a primary quantification standard for SEM EDS major element analysis of
basaltic glasses, and was analysed 8–15 times every
session. The average of all VG2 analyses for the session
was normalized to the published composition
(Jarosewich et al., 1980) and the normalization factors
were applied to all unknowns from that session.
Quenched sulfide and selenide were analysed via
EDS using the JEOL JSM6400 SEM at 15 kV, 1 nA and
100 s livetime. Selenide blebs were typically analysed in
spot mode, owing to their small size (typically<10 mm
diameter). Every attempt was made to analyse areas
that were as large and well polished as possible when
analysing sulfide layers. Pyrite was used as the standard for Fe and S; native selenium as the standard for Se;
Re metal as the standard for Re; Pt metal as the standard for Pt. Synthetic Cu2Se, NiSe and Fe(1–x)Se were
analysed during analytical sessions as secondary standards for Se.
Journal of Petrology, 2015, Vol. 56, No. 7
glasses were free of crystals and vesicles. Run conditions, capsule material, and major element and sulfur
contents of silicate glasses from sulfur-bearing experiments are reported in Table 1. Silicate glasses contained between 900 and 6000 ppm S, and display the
predicted asymmetric U-shaped dependence of SCSS
on the FeO content of the silicate melt, as demonstrated
in Fig. 2. Also demonstrated in Fig. 2 is the similar S
contents of glasses from both Re and graphite-in-Pt
capsules at 10 wt % FeO, suggesting that f O2 conditions in the Re capsule experiments are not sufficiently
different from those of the graphite-in-Pt experiments
to affect SCSS.
The sandwich arrangement of sulfide-bearing experiments was preserved during each experiment, although
<10 mm blebs of sulfide liquid were disseminated
throughout the silicate glass (Fig. 3). Sandwich experiments with FeO contents >15 wt % produced abundant,
semi-regularly spaced sulfide blebs 1 mm in diameter.
The two >15 wt % FeO mounts (D0058 and D0063) were
not available for characterization and re-analysis in the
present study, although Métrich et al. (2009) interpreted
the regularly spaced blebs to reflect a quench artefact.
SEM EDS analyses of sulfide blebs are reported in
Table 2. In almost all experiments, the upper layer of
sulfide (i.e. the layer contacting the lid) became contaminated with Pt (Fig. 4c and d) presumably as a result
of sulfide liquid migrating along the interface between
the graphite capsule and graphite lid, reaching the Pt
outer capsule and dissolving Pt (Fig. 4b). The lower sulfide layer remained uncontaminated in all experiments
(Table 2; Fig. 4a). As a result, the silicate liquid coexisted
with two reservoirs of sulfide liquid with slightly
RESULTS
Sulfide-saturated experiments
Sulfide-saturated experiments produced two immiscible liquids, one a basaltic silicate liquid and the other
an iron sulfide liquid. The recovered quenched silicate
Fig. 2. SCSS of basalt as a function of silicate glass FeO content. Filled symbols, graphite-in-Pt encapsulated experiments;
open symbols, Re encapsulated experiments.
Journal of Petrology, 2015, Vol. 56, No. 7
1413
Table 2: SEM EDS analyses of sulfide liquid blebs
Run ID S
Fe
D0055
D0055
D0057
D0064
D0064
D0065
D0065
D0066
D0066
D0067
D0067
D0068
D0068
D0093
D0094
D0095
C1473
D0052
592(12)
609
618(5)
366(20)
599(9)
444(39)
597(2)
564(6)
588
565(37)
610(3)
455(17)
612(1)
578(3)
578(5)
588(4)
581(3)
584(3)
351(5)
364
360(3)
281(18)
370(7)
319(6)
372(3)
361(3)
366
356(8)
350(15)
308(10)
346(6)
369(2)
370(3)
346(13)
367(3)
369(1)
Re
Pt
30(22)
387(16)
267(51)
59(8)
76(63)
230(24)
19(2)
24(4)
22(5)
24(3)
33(4)
Total n Notes
973
974
978
1034
969
103
969
984
957
997
96
993
958
966
972
956
972
986
5
1
6
3
3
3
3
3
1
3
3
3
3
6
6
6
6
6
upper layer
lower layer
all layers
upper layer
lower layer
upper layer
lower layer
upper layer
lower layer
upper layer
lower layer
upper layer
lower layer
Values expressed in weight per cent Numbers in parentheses are 1SD in the last digit The variation in Pt content between the upper and lower layers of a single sample should be
noted.
Selenide-saturated experiments
Fig. 3. Backscattered electron images of (a) capsule and (b)
sample from Run D0066. Scale bar in both panels represents
1 mm.
different compositions that did not equilibrate during
the course of the experiment. Therefore, a small chemical potential gradient existed across the silicate melt
with respect to FeS, such that overall a(sulfide)
in the sulFeS
fide liquid will be less than unity. This will lower SCSS
relative to the case where a(sulfide)
is unity, and thus
FeS
SCSS values for Pt-contaminated experiments should
be considered a minimum for this silicate melt composition at 1400 C and 15 GPa.
Unlike graphite-in-Pt experiments where the original layered geometry is preserved, FeS liquids wet
the Re metal capsule such that the quenched FeS liquid
forms a continuous shell between the capsule and the
silicate melt (Fig. 5). The study of Fonseca et al. (2007)
reported sulfide melt Re contents of less than 1200 ppm
in 1 atm experiments at 1200–1400 C, whereas SEM
EDS analyses of sulfide from the present experiments,
which were conducted at higher f S2, found a maximum
of 33 wt % Re dissolved in the sulfide, suggesting
that the effect of Re on a(sulfide)
will be higher than
FeS
in 1 atm experiments, but still minor, and another
reason that SCSS values from these experiments
should be considered a minimum for the particular
conditions.
Selenide-saturated experiments similarly produced
crystal- and vesicle-free basaltic glass saturated with an
immiscible iron selenide liquid that occurs as small
(micrometres to tens of micrometres scale) blebs disseminated throughout the glass (Fig. 6). Major element
and Se contents for Se-bearing experiments are reported in Table 3, and Se contents are plotted as a function of FeO content of the silicate glass in Fig. 7. Se
contents of the silicate glass vary from 1000 to
4000 ppm Se, and also exhibit an asymmetric Ushaped relation to the FeO content of the silicate melt,
although the relationship differs slightly from that of
SCSS. On a molar basis, Se2– is about half as soluble as
S2– in a silicate melt of identical composition at the
same P and T. When converted to mass fraction,
SeCSeS is coincidentally the same magnitude as SCSS.
In detail, the SeCSeS ‘U’ is more shallow than that for
sulfur and the enrichment at both high and low FeO is
less than that observed for sulfur. Nevertheless, the
presence of an asymmetric U-shaped dependence on
FeO content suggests a similar solubility mechanism for
selenide and sulfide, involving the same interplay be(silicate liquid)
tween the negative effect of aFeO
in the silicate
melt and the positive effect of the FeO component of the
selenide capacity. High FeO content glasses did not exhibit the semi-regularly spaced 1 mm diameter blebs interpreted by Métrich et al. (2009) to be a quench artefact
seen in high-FeO sulfur-bearing experiments. The lower
overall abundance of Se in the silicate melt suggests
that the selenium equivalent of reaction (5)
FeOðsilicateliquidÞ þ 0 5Se2ðgasÞ $ FeSeðselenideliquidÞ þ 0 5O2ðgasÞ
(10)
favours the right-hand side of the reaction more than
the equivalent S reaction.
1414
Journal of Petrology, 2015, Vol. 56, No. 7
Fig. 4. Backscattered electron images of quenched sulfide liquids from graphite-in-Pt experiments. (a) Pt-free sulfide from lower
layer of Run D0064; (b) sulfide liquid migrating along the interface between graphite capsule and lid and dissolving Pt from outer
capsule (Run D0055); (c) lightly Pt-contaminated upper sulfide layer of Run D0066; (d) heavily Pt-contaminated upper sulfide layer
of Run D0064 [compare (a)]. Scale bar represents 10 mm in (a), (c) and (d); 100 mm in (b).
Selenide blebs were small (typically <10 mm) and
abundant, requiring considerable effort to locate clear
areas of glass sufficiently large for electron microprobe
analysis. Initial attempts at quantifying Se via the Se Ka
peak at 25 kV accelerating voltage produced unacceptably high scatter in the analytical results, presumably
owing to secondary fluorescence of Se in the blebs by
bremsstrahlung radiation (see Llovet et al., 2012).
Significantly improved results were obtained when
counting on the Se La peak at 20 kV accelerating voltage. Unlike the sulfide liquids analysed in this and other
similar studies (Liu et al., 2007), the selenide liquid stoichiometry did not remain at 1:1 metal:chalcogen.
Instead, the Se content of the selenide liquid increased
(Table 4; Fig. 8) with decreasing silicate melt FeO content. A consequence of the non-stoichiometry of iron
(selenide liquid)
selenide liquids is that aFeSe
is less than unity
in all experiments, and decreases with decreasing FeO
content.
The glasses from the silicate liquid–selenide liquid
experiment series at 6000 ppm (C2481), 12 000 ppm
(C2481) and 24 000 ppm (C2481) added Se produced
similar Se contents (1636 ppm, 1549 ppm and 1525 ppm
(selenide liquid)
respectively) in the glass and XFeSe
values
(088, 084 and 086 respectively), suggesting that
(selenide liquid)
SeCSeS and XFeSe
results are not affected by
the small mass of selenide liquid relative to silicate
liquid.
Data fitting
As a complement to the forward modelling of SCSS
using the O’Neill & Mavrogenes (2002) model (Fig. 1),
we have conducted inverse modelling to examine the
effect of FeO on SCSS. The model of O’Neill &
Mavrogenes (2002) was rearranged such that it consists
of only two terms, a0 and a1, in addition to the FeOrelated terms. The expression can be considered to reflect mixing between an FeO-free ‘matrix’ and FeO.
For example, the mole fraction of Ca cations in equation (4) is defined as
XCa ¼ NCa =RNM
where NM are cations of M per 100 g.
(11)
Journal of Petrology, 2015, Vol. 56, No. 7
1415
Fig. 5. Backscattered electron images of (a) longitudinal and (b)
radial sections through a Re capsule experiment showing sulfide liquid wetting Re metal to form a film between the silicate
liquid and capsule. Scale bar represents 1 mm in both panels.
However, for the ‘matrix’ composition the mole fraction of each cation can be defined as
X Ca ¼ NCa =ðRNM –NFe Þ
(12)
which can be simplified to
X Ca ¼ XCa =ð1–XFe Þ:
(13)
We can therefore express the sulfide capacity as
lnCS ¼ A0 þ ð1 –XFe Þ½ACa X Ca þ AMg X Mg þ ... þ AFe XFe (14)
and SCSS may be expressed as
silicate liquid
sulfide liquid
ln ½SSCSS ¼ a0 þ a1 XFe – lnaFeO
þ lnaFeS
(15)
where
a0 ¼ –DG ð5Þ =RT þ A0 þ RAM X M ðM 6¼ FeÞ þ CP =T (16)
and
a1 ¼ –RAM X M ðM 6¼ FeÞ þ AFe :
(17)
The a0 and a1 terms are then fitted by least-squares
regression. Thus, for the case of a haplobasaltic ‘matrix’
Fig. 6. Backscattered electron images of textures from SeCSeS
experiments. (a) Sample area from Run C3590; (b) detail of selenide blebs from Run C3962. Scale bar in (a) represents
10 mm; in (b) 100 mm.
composition to which FeO is added, the model
of O’Neill & Mavrogenes (2002) can be reduced to an
expression with inputs of only cation fraction Fe and
mole fraction FeO in the silicate melt plus a(sulfide)
or
FeS
a(selenide)
of
the
coexisting
chalcogen
liquid
(approxiFeSe
mated as unity for the present SCSS experiments, and
XFeSe for SeCSeS experiments). Inverse modelling by
least-squares fitting of sulfide-saturated experiments
from the present study produced a0 ¼ 385 and
a1 ¼ 1682, with a reduced v2 of 32. The model is known
1416
Journal of Petrology, 2015, Vol. 56, No. 7
Table 3: EPMA analyses of silicate glasses saturated with FeSe liquid
Run ID duration Na2O
(hours)
MgO
Al2O3
SiO2
CaO
TiO2
FeO
Se
Total
C3950
D1246
C3963
D1248
C3962
D1472
D1180
D1178
C4283
C4282
C4281
D1481
D1179
C4447
C3960
C3968
D1549
1053(8)
1063(16)
1025(7)
1007(7)
1030(8)
1023(13)
1011(11)
979(8)
962(6)
962(7)
976(8)
951(6)
959(8)
921(8)
891(9)
857(6)
858(6)
2092(15)
2082(25)
2128(15)
2088(15)
2061(12)
2030(11)
2089(10)
2067(14)
2047(12)
2054(13)
2018(8)
1966(14)
1915(7)
1854(9)
1920(12)
1718(14)
1724(12)
5279(34)
5294(46)
5194(26)
5109(20)
5206(15)
5155(32)
5090(24)
4924(21)
4860(18)
4842(32)
4892(19)
4787(25)
4835(24)
4650(38)
4513(21)
4329(17)
4338(38)
1124(13)
1110(14)
1093(7)
1072(13)
1088(9)
1083(14)
1066(9)
1023(8)
1014(8)
1013(9)
1035(11)
1000(10)
1003(7)
974(7)
943(9)
907(7)
908(6)
145(6)
141(7)
145(6)
137(8)
144(6)
140(7)
140(5)
137(8)
129(9)
128(9)
133(5)
134(6)
136(5)
130(9)
122(6)
121(5)
118(5)
031(10)
061(21)
113(8)
148(7)
235(11)
254(14)
448(12)
616(10)
656(16)
697(18)
708(14)
847(19)
878(15)
1260(15)
1282(22)
1810(20)
1849(16)
01825(22)
01577(51)
01168(23)
01341(22)
01083(21)
01290(17)
01044(16)
01160(54)
01526(47)
01550(41)
01636(21)
01766(21)
01597(17)
02421(42)
02144(17)
03408(22)
03810(69)
9994
10015
9959
9809
10022
994
10097
9999
9914
9942
10009
9932
9977
10037
9914
999
10035
36
6
36
6
36
16
36
36
16
16
16
16
36
36
36
36
36
252(5)
248(4)
249(8)
235(4)
247(4)
241(6)
243(5)
241(4)
230(4)
231(7)
231(4)
229(6)
235(5)
223(4)
222(6)
214(7)
203(8)
n
n
(majors) (selenium)
15
13
15
10
10
9
10
10
7
6
6
6
11
6
15
10
6
9
8
9
9
9
8
9
9
19
19
9
9
9
13
8
9
13
Values expressed in weight per cent. Numbers in parentheses are 1SD in the last digit. All experiments conducted at 15 GPa,
1400 C.
Inverse modelling of the SeCSeS data (Fig. 10) highlights differences between SCSS and SeCSeS. The
asymmetric U-shape is preserved in the inverse modelling, although the overall fit is poor, reflected in the
reduced v2 value of 58; residuals are presented in Fig.
S1 of the Supplementary Data (supplementary data are
available for downloading at http://www.petrology.
oxfordjournals.org). A significant factor in the poor fit
of the inverse SeCSeS model is probably the nonstoichiometry of the selenide liquid, and our complete
lack of knowledge of activity–composition relations of
such FeSe–Se liquids. However, despite the poor fit, the
SeCSeS model returned a0 (319) and a1 (2746) terms
similar to the SCSS forward and inverse models.
DISCUSSION
Comparison with other models
Fig. 7. SeCSeS of basalt as a function of silicate glass FeO content (wt %).
to break down at higher FeO contents (fig. 15 of O’Neill
& Mavrogenes, 2002), and high FeO content glasses are
known to be affected by quench modification (Métrich
et al., 2009), so omitting experiments with FeO > 200 wt
% (a0 ¼ 342, a1 ¼ 2224, v2 ¼ 09) and FeO > 150 wt %
(a0 ¼ 310, a1 ¼ 2735, v2 ¼ 03) results in an improvement of the fit. The fits to the experimental data are
shown in Fig. 9. For comparison, the a0 and a1 terms for
the forward model can be calculated by substituting the
appropriate parameters into equation (15) [i.e. AM and –
DG (5) from O’Neill & Mavrogenes (2002) and the CP/T
parameter of Mavrogenes & O’Neill (1999)] to return a0
of 322 and a1 of 2239.
The results presented in Fig. 2 demonstrate that the sulfur content of FeS-saturated silicate melts exhibits an
asymmetric U-shaped dependence on the FeO content
of the silicate melt, with a minimum between 4 and 6 wt
% FeO, in agreement with the prediction from the
thermodynamic model of O’Neill & Mavrogenes (2002).
Importantly, high-pressure experiments are necessary
to achieve the high f S2 conditions necessary to saturate
basaltic silicate liquids in an immiscible FeS liquid at
low FeO contents. Extant empirical models have had
limited experimental data for the high-SCSS, low-FeO
limb of the ‘U’ as input (i.e. Tsujimura & Kitakaze, 2005),
and thus may not correctly predict the increase in SCSS
at low FeO content. In this section, the performance of
several models for SCSS from the literature is compared with that of the O’Neill & Mavrogenes (2002)
model.
Figure 11 shows a comparison of the O’Neill &
Mavrogenes (2002) model and those of Li & Ripley
(2005), Liu et al. (2007), Li & Ripley (2009) and Ariskin
Journal of Petrology, 2015, Vol. 56, No. 7
1417
Table 4: SEM EDS analyses of selenide liquid blebs
Run ID
S
Fe
Se
Pt
Total
Fe/(FeþSe)
(molar)
XFeSe
(selenide)
XSe
(selenide)
n
C3950
D1246
C3963
D1248
C3962
D1472
D1180
D1178
C4282
D1179
C4447
C4283
D1481
D1549
C4281
C3960
C3968
006(4)
016(3)
017(1)
017(2)
016(2)
018(3)
013(3)
013(5)
007(3)
021(5)
019(2)
015(5)
024(3)
016(4)
019(3)
012(2)
005(4)
2796(52)
2883(34)
2885(41)
3024(56)
3126(69)
3295(60)
3394(34)
3510(42)
3707(37)
3618(33)
3587(120)
3696(100)
3657(31)
3597(17)
3741(32)
3808(72)
3877(41)
6759(134)
6940(49)
6899(139)
6809(22)
6515(163)
6422(92)
6450(58)
6227(95)
6259(117)
6086(119)
6018(231)
6105(245)
6021(66)
5827(136)
6043(31)
5836(348)
5692(44)
058(26)
019(12)
048(31)
018(9)
029(18)
046(24)
058(17)
056(15)
017(11)
060(14)
084(12)
028(13)
070(23)
035(13)
047(14)
083(7)
044(22)
962
9858
9849
9868
9686
9781
9915
9805
9989
9785
9708
9843
9773
9475
985
9739
9618
037(1)
037
037
039
040(1)
042(1)
043
044
046
046
046
046(2)
046
047(1)
047
048(1)
049
059(2)
059(1)
059(1)
063(1)
068(1)
073(2)
074(1)
080(2)
084(1)
084(1)
084(2)
086(6)
086(2)
087(2)
088(1)
093(5)
096(1)
042(2)
041(1)
041(1)
037(1)
032(1)
027(2)
026(1)
020(2)
016(1)
016(1)
016(2)
014(6)
014(2)
013(2)
013(1)
008(5)
004(1)
5
5
5
5
5
9
10
6
6
6
6
9
4
2
5
5
6
Values expressed in weight per cent. Numbers in parentheses are 1SD in the last digit.
Fig. 8. Mole fraction FeSe and Se on the FeSe–Se binary of selenide blebs plotted against FeO (wt %) in the glass.
et al. (2013) in SCSS–FeO space. The ‘matrix’ input composition for each model was the average, normalized
FeO-free composition determined from the silicate
glasses of the SCSS experiments conducted in this
study (Table 5). The calculations assume all Fe is Fe2þ
and 250 ppm H2O in the silicate melt for models requiring a value for H2O content. Plots of the component
terms for the Li & Ripley (2005), Liu et al. (2007) and Li &
Ripley (2009) models versus silicate melt FeO content
are presented in the Supplementary Data as Figs S1, S2
and S3.
For the range of FeO contents observed in terrestrial
basalts (typically 5–10 wt % FeO), all the models produce approximately the correct slope and magnitude
Fig. 9. Comparison of forward modelling and inverse modelling of SCSS data. Forward model: a0 ¼ 316, a1 ¼ 2485,
¼ 5126; all points: a0 ¼ 385, a1 ¼ 1682, v2 ¼ 320; FeO < 20 wt
%: a0 ¼ 343, a1 ¼ 2224, v2 ¼ 087; FeO < 15 wt %: a0 ¼ 311,
a1 ¼ 2736, v2 ¼ 025. The forward model and outputs from the
three fits were calculated using the ‘matrix’ composition listed
in Table 5, which is the average of normalized, FeO-free compositions determined from SCSS experiments.
for the dependence of SCSS on FeO content. This behaviour is unsurprising, as most of the experimental
data used to generate the models involve basalts of terrestrial composition with FeO contents between 5 and
10 wt %. There is significant divergence at low (<2 wt
%) FeO content. This is due to the lack of a negative ln
(silicate melt)
aFeO
term in other models, whereas it is present
in equation (8) of this paper. The Liu et al. (2007) model
predicts a concave-downwards dependence at very low
FeO contents, the opposite of what is observed
1418
Fig. 10. Comparison of forward modelling (eight-term) and inverse modelling of SeCSeS data. Forward model: a0 ¼ 318,
a1 ¼ 2484, v2 ¼ 7669; all points: a0 ¼ 319, a1 ¼ 2746, v2 ¼ 5804;
FeO < 15 wt %: a0 ¼ 298, a1 ¼ 3347, v2 ¼ 3796. The forward
model and outputs from the three fits were calculated using
the ‘matrix’ composition listed in Table 5, which is the average
of normalized, FeO-free compositions determined from
SeCSeS experiments. SeCSeS datapoints from this study were
corrected to aFeSe ¼ 10, assuming XFeSe ¼ aFeSe. The weight of
input SeCSeS values was set to 5% or 1SD, whichever was
greater.
experimentally, as a result of the positive coefficient of
(silicate melt)
the ln XFeO
term (Supplementary Data Fig. S2).
Interestingly, the Li & Ripley (2005) model does include
(silicate melt)
a negative ln aFeO
term, and consequently produces an asymmetric U-shaped curve (Supplementary
Data Fig. S1); however, this term was dropped in the
(silicate melt)
2009 version of their model for an XFeO
term, resulting in the prediction of increasing SCSS with additional FeO for all FeO contents (Supplementary Data
Fig. S3). The output from COMAGMAT (Ariskin et al.,
2013) successfully produces an asymmetric U-shape, although shifted to lower FeO contents than the results of
the present experiments. Unlike the other models
examined herein, the model of Ariskin et al. (2013) is
not based exclusively on high-temperature experimental data, but also includes 50 S-saturated mid-ocean
ridge basalt (MORB) glasses, many of which represent
liquidus temperatures below 1200 C and are saturated
with Fe–Ni–Cu–S liquids. The resulting model more accurately represents SCSS in natural melts at the expense of high-temperature synthetic melts.
At high FeO contents (10–20 wt % FeO), the models of
Li & Ripley (2005), Liu et al. (2007) and Li & Ripley (2009)
perform better than that of O’Neill & Mavrogenes (2002).
Inspection of Fig. 1 reveals that the value for SCSS at
high FeO contents is dominated by the sulfide capacity
term (ln CS). O’Neill & Mavrogenes (2002) showed that
their parameterization of CS is valid only up to 15 wt %
Journal of Petrology, 2015, Vol. 56, No. 7
Fig. 11. Comparison of SCSS models calculated at 15 GPa and
1400 C using the average, normalized FeO-free composition
determined from SCSS experiments (Table 5). All models assume that all Fe is Fe2þ and H2O content is 250 ppm for models
that require it as an input.
FeO (see fig. 15 of O’Neill & Mavrogenes, 2002), with CS
progressively overestimated at >15 wt % FeO, consistent with the discrepancy identified here. This implies
that the Fincham–Richardson relationship, and the reciprocal solid solution model derived from it, breaks down
at these high FeO contents. Further experiments at high
FeO content will be necessary to guide the modelling of
CS at these higher FeO contents, but such experiments
may be difficult given the tendency for the sulfur dissolved in high-FeO silicate melts to exsolve on quenching (see Markus & Baker, 1989).
Although the model of O’Neill & Mavrogenes (2002)
accurately describes SCSS for the silicate liquid investigated in this study (at FeO < 15 wt %), it currently has
limited use as a geochemical modelling tool, as AM coefficients exist only for 1400 C. It cannot be stressed
enough that, despite the occurrence of a T term in the
expression for DG (5) the O’Neill & Mavrogenes (2002)
model can be employed only at 1400 C in its present
state. Thus, in the interim while the model calibration is
extended to other temperatures, the reader is directed
towards the Li & Ripley (2005) model or COMAGMAT if
they wish to conduct geochemical modelling over a
range of P, T and silicate liquid compositions.
Ultimately, all the empirical models work well for dry
basaltic liquids with 5–12 wt % FeO, and it is only at low
FeO contents that the modeller needs to exercise
discretion.
(sulfide)
The effect of aFeS
The modelling and experiments presented so far
involved sulfide liquids with a composition of pure FeS
Journal of Petrology, 2015, Vol. 56, No. 7
1419
Table 5: Silicate melt compositions employed in modelling SCSS and SeCSeS as a function of FeO
Na2O
MgO
Al2O3
SiO2
K2O
CaO
TiO2
MnO
Total
261
264
31
1168
41
047
1051
1065
842
–
032
1557
2117
2106
1722
2028
1341
161
5331
5294
5662
6803
7777
6156
–
–
–
–
255
–
1097
1119
1296
–
185
46
143
151
168
–
–
096
–
–
–
–
–
074
100
100
100
100
100
100
023
2178
1334
5721
–
587
09
069
100
–
073
2005
785
1409
877
5388
7783
–
–
1162
474
016
0
02
007
100
100
Reference
Notes
this study
this study
Jenner et al., 2010
Bradbury, 1983
Clemente et al., 2004
Stockstill-Cahill
et al., 2012
Stockstill-Cahill
et al., 2012
McCoy et al., 1999
Berthet et al., 2009
SCSS experiments; Figures 1, 9, 11
SeCSeS experiments; Figure 10
FeO-free MORB; Figure 12
FeS-saturated experiments; Figure 13
FeS-saturated experiments; Figure 13
Northern Volcanic Plains
Intercrater Plains
Experiment 269 (1400 C; 1 bar)
Experiment 316 (1400 C; 10 kbar)
Values expressed in weight percent. Compositions in this table are calculated as the average of the population of normalized,
FeO-free compositions.
[i.e. a(sulfide)
¼ 1]. Sulfide blebs from the glassy rims of
FeS
MORB pillow basalts contain significant Ni and Cu
(Patten et al., 2012, 2013), resulting in reduced a(sulfide)
.
FeS
Modelling of SCSS as a function of FeO content, con(sulfide)
toured for aFeS
values between 10 and 01, is presented in Fig. 12. The base silicate melt or ‘matrix’
composition used in the modelling is the FeO-free
average major element composition of 616 ocean
floor basaltic glasses studied by Jenner & O’Neill
(2012); the pressure used in the calculation is 100 MPa
and the temperature is 1400 C, admittedly a high temperature for MORB, but it is the only temperature currently supported by the O’Neill & Mavrogenes (2002)
model.
In most natural magmatic systems, the NiO and
Cu2O contents of the silicate melt are sufficiently low to
be an insignificant contribution to the sulfide capacity
(see Evans et al., 2008), such that the FeO content of the
silicate melt remains the dominant component of the
sulfide capacity term (CS). However, strong partitioning
of Ni and Cu into the sulfide phase has the potential to
significantly reduce the a(sulfide)
term, owing to the nonFeS
ideal nature of Ni–Cu–Fe–S liquids (Hsieh & Chang,
1987; Brenan, 2003; Kress, 2007). The overall effect of Ni
and Cu addition to the sulfide liquid is to stabilize immiscible sulfide liquids at lower values of SCSS in the
coexisting silicate liquid than the simple FeS liquid case
(Ariskin et al., 2013). Overlaid on the a(sulfide)
contours of
FeS
Fig. 12 is a density plot of sulfur and FeO(total) contents
measured in 329 sea-floor glasses sourced from Jenner
& O’Neill (2012). Inspection of the SCSS contours in Fig.
12 reveals that decreasing a(sulfide)
results in decreasing
FeS
sensitivity of SCSS to the FeO content of the silicate
melt, particularly in the 5–15 wt % FeO range, and also
results in a shift of the low-FeO SCSS enrichment to
even lower FeO contents. A decrease in temperature (at
constant melt composition) will result in a decrease in
SCSS (e.g. fig. 1 of Wendlandt, 1982; Baker & Moretti,
2011), resulting in the cloud of MORB data points coinciding with higher a(sulfide)
contours than in the present
FeS
figure. For the O’Neill & Mavrogenes (2002) model to
describe the observed MORB sulfur contents a detailed
activity–composition model to account for the effect of
Ni, Cu, O and excess S on the activity of FeS in MORB
Fig. 12. Output of the O’Neill & Mavrogenes (2002) model, calculated using the average MORB composition from Jenner &
O’Neill (2012), contoured for aFeS. Also shown are FeO and S
contents of 329 sea-floor glasses from Jenner & O’Neill (2012),
and point density contours for that dataset.
sulfide blebs, along with a calibration of the CS parameterization at MORB temperatures, is necessary.
Further consideration of the form of the O’Neill &
Mavrogenes (2002) model reveals that when presented
melt)
in FeO–SCSS space, only the ln a(silicate
and ln CS
FeO
terms vary with FeO content (see Fig. 1). Thus, varying
any of the other FeO independent terms will produce
similar variations in the shape of the predicted SCSS
versus FeO content curves to those illustrated in Fig. 12.
It then follows that it is entirely unsurprising that the Li
& Ripley (2005) model produces a result with a shape
similar to one of the lower a(sulfide)
contours, as the form
FeS
of that model is effectively identical to that of O’Neill &
Mavrogenes (2002), consisting of a negative natural log
of the FeO content, a positive slope straight line compositional term, and FeO insensitive terms representing
P, T and a constant, as shown in Fig. S2 of the
Supplementary Data, for comparison with Fig. 1.
1420
SCSS in hydrous silicic melts
The strong dependence of SCSS on the FeO content of
anhydrous, basaltic silicate melts has been clearly demonstrated in this study. The interplay between the two
FeO-related terms is a fundamental characteristic of the
dependence of SCSS on FeO content. An important
question is whether this relationship holds for lower
temperature, hydrous, silicic melts.
The unpublished PhD thesis of Bradbury (1983)
investigated the solubility of pyrrhotite in hydrous albite
limelt as a function of pressure, temperature, XH(silicate
2O
quid)
and f S2. All the experiments of Bradbury (1983)
were saturated with pyrrhotite (of varying composition)
and a H2O–H2S fluid phase. Loss of sulfur from pyrrhotite to form H2S in the vapour phase resulted in the dissolution of FeO in the albite melt. When plotted in FeO–
SCSS space, the results form the characteristic negative
logarithm, low-FeO side of the asymmetric U-shaped relationship. Clemente et al. (2004) conducted a detailed
study into the solubility of sulfur in metaluminous rhyolitic melts as a function of pressure, temperature, f O2
and f S2.
As demonstrated in the foregoing sections and in
Fig. 1, the shape of the ‘U’ is fixed by the relationship
(silicate melt)
between CS and ln aFeO
as a function of FeO content, whereas the ‘magnitude’ is determined by a combination of these two parameters and the remaining
parameters, such as ln aFeS, pressure dependence, and
the equilibrium constant for reaction (5). At low FeO
contents the shape of the ‘U’ is dominated by the nega(silicate melt)
tive ln aFeO
term, rather than the sulfide capacity
(we would expect the activities of oxide components,
and thus AM coefficients of different cations making up
CS, to vary with temperature; O’Neill & Mavrogenes,
2002, p. 1067).
In a crude attempt to investigate the relationship between SCSS and FeO content in silicic melts, the SCSS
data from the pyrrhotite-saturated experiments of
Bradbury (1983) and Clemente et al. (2004) were recast
as anhydrous compositions and fit using the ‘matrix’FeO inverse modelling method presented above. This is
a valid way to explore the relationship between FeO and
SCSS as the only inputs are XFe, XFeO and SCSS,
and the only assumptions are the relationship
(silicate melt)
between SCSS, XFe, ln aFeO
and ln a(sulfide)
as exFeS
pressed in equation (15), and that the effect of H2O is insignificant. In contrast to earlier inverse modelling, the
(silicate melt)
ln aFeO
term was also set as a variable in the regression. The result of the regression is plotted as a continuous function of FeO content by calculating the
average FeO-free composition from each input dataset
(i.e. the ‘matrix’ composition), and calculating SCSS as
FeO is added to the ‘matrix’. The average, normalized,
FeO- and H2O-free silicate liquid compositions used to
calculate the model output are listed in Table 5. The results are presented in Fig. 13.
The regression produces a fit to both datasets that
describes the shape of the relationship between silicate
Journal of Petrology, 2015, Vol. 56, No. 7
melt FeO content and SCSS very well, given the simplifications. This is a significant finding, as it can be inter(silicate melt)
preted to suggest that the ln aFeO
term controls
SCSS in hydrous, low-FeO, metaluminous silicic magmas, just as it does in high-temperature, dry, basaltic
melts. Even at very high molar H2O contents (Burnham,
1979a, 1979b) the effect of H2O on SCSS does not over(silicate melt)
whelm the effect of FeO via the ln aFeO
term. H2O
does not have a sufficiently large effect on CS (i.e. via an
(silicate melt)
AH2O term) to overwhelm the effect of ln aFeO
at
low FeO contents. Therefore, in the presence of an Febearing saturating phase such as pyrrhotite or pyrite,
reduced sulfur dissolves in silicate melts as S2–, rather
than species such as H2S or HS–, as the silicate melt
must necessarily also contain FeO. Additionally, spectroscopic investigations suggest that H2S or HS– species are observed only in glasses from Fe-free systems
(Klimm & Botcharnikov, 2010) or systems with Fe/S of
two or less (Klimm et al., 2012).
The relationship of increasing SCSS with decreasing
FeO content shown in Fig. 13 may be significant during
the evolution of granitic melts, and could conceivably
result in the dissolution of sulfides should a granitic
melt evolve to a sufficiently low FeO content. This presents an avenue for future investigation, along with
more general investigation of SCSS as a function of
FeO content in hydrous melts, particularly at higher FeO
contents to confirm the expected increase in SCSS.
The FeO-poor, sulfur-rich surface of Mercury
Remote sensing of the planet Mercury suggests a low
overall FeO content of the surface (02–45 wt % FeO),
which is interpreted to largely consist of volcanic material (Nittler et al., 2011; Evans et al., 2012; Weider et al.,
2012). The surface of Mercury exhibits unusually high
sulfur contents (1–6 wt %), and correlations between Ca,
Mg and S have led to suggestions of Ca and Mg sulfide
minerals on the surface of Mercury. As demonstrated
by the present work, silicate melts are expected to dissolve relatively high amounts of sulfur at low FeO contents when saturated with an Fe-rich sulfide phase. The
model of O’Neill & Mavrogenes (2002) can be applied to
silicate melt compositions relevant to the surface of
Mercury to examine whether proposed surficial S contents can plausibly represent sulfur dissolved in lowFeO silicate magmas saturated with an Fe–S phase.
Zolotov et al. (2013) explored the solubility of sulfur
in liquids representing the surface of Mercury via several models from the metallurgical literature attempting
to constrain both f O2 and f S2 in addition to SCSS. As
presented in the Introduction, the O’Neill & Mavrogenes
(2002) model is insensitive to f O2 and f S2 when a melt
is saturated with stoichiometric FeS (solid or liquid); furthermore, the addition of other components to a liquid
sulfide phase, such as oxygen, Ni, Cu and additional Fe
or S, will serve to lower the activity of FeS, and thus
SCSS. Here, the SCSS in four compositions relevant to
the surface of Mercury were modelled using the model
Journal of Petrology, 2015, Vol. 56, No. 7
Fig. 13. Inverse modelling of SCSS for low-temperature, hydrous silicic glasses. FeO contents of the model output were
calculated using the average, normalized FeO- and H2O-free
composition of included experiments from (a) Bradbury (1983)
and (b) Clemente et al. (2004).
of O’Neill & Mavrogenes (2002); two are modelled compositions derived from MESSENGER X-ray spectrometry data, ‘Northern Volcanic Plains’ and ‘Intercrater
Plains and Heavily Cratered Terrain’ (Stockstill-Cahill
et al., 2012); two are silicate liquids produced from partial melting of the Indarch enstatite chondrite, the
1400 C silicate liquid composition from Run 269 of
McCoy et al. (1999), and the 10 GPa, 1400 C silicate liquid composition from Run 316 of Berthet et al. (2009)
(see Table 5 for compositions). The SCSS as a function
of FeO content for the four compositions is presented in
Fig. 14. Sulfur contents greater than 1 wt % are achieved
in all silicate melt compositions, although only at very
1421
low FeO contents, ranging from 1300 ppm FeO in the
Berthet et al. (2009) composition to 05 wt % FeO in
the composition of McCoy et al. (1999). Thus, it is plausible that silicate melts with low FeO contents can contain significant dissolved sulfur when the saturating
sulfide phase is essentially FeS.
However, it is likely that the sulfide solubility regime
occurring on Mercury and enstatite chondrite parent
bodies is unlike the regime familiar on Earth. For almost
all terrestrial magmas, the range of f O2 conditions is
such that iron dissolves into silicate melts as
Fe2þ 6 Fe3þ, and Fe metal is not a stable phase. Fe metal
is observed in terrestrial basaltic magmas only in rare
cases involving assimilation of carbon-rich sedimentary
rocks (e.g. Disko Island; Pedersen, 1979). In the case of
enstatite chondrite parent bodies, f O2 is extremely low,
such that silicate melts are almost Fe-free, owing to the
formation of an Fe-rich metal phase that frequently contains several weight per cent silicon. When the sulfur
solubility limit of the metal phase and silicate liquid is
exceeded (i.e. at low f O2 and high f S2) saturation in refractory sulfide phases such as oldhamite (CaS) and
niningerite (MgS) occurs. Analyses of glasses produced
from experiments involving partial melting of the
Indarch enstatite chondrite (Fogel et al., 1996; McCoy
et al., 1999; Malavergne et al., 2007; Berthet et al., 2009)
return values for SCSS significantly greater than predicted by the O’Neill & Mavrogenes (2002) model; for
example, up to 782 wt % sulfur was reported from the
experiments of Fogel et al. (1996). One explanation for
the results is that the FeO content of the glasses may be
overestimated owing to secondary fluorescence of Fealloy blebs (see Cottrell & Walker, 2006), and the high
SCSS values are occurring at very low (i.e. <01 wt %)
FeO contents as predicted by the O’Neill & Mavrogenes
(2002) model. Alternatively, SCSS values determined
from experimental glasses could be similarly overestimated owing to secondary fluorescence. The present
model suggests that as the FeO content of the silicate liquid tends towards zero, SCSS tends towards infinity.
In reality, infinite SCSS does not occur, and rather the
silicate liquid will saturate in another sulfide phase,
probably a refractory Ca or Mg sulfide phase. Thus, to
explore the plausible upper level of SCSS, values for
SCSS from experiments with CaS- and MgS-saturated
silicate melts are necessary (see Malavergne et al.,
2007).
SUMMARY
The sulfur content at sulfide saturation (SCSS) and the
selenide content at selenide saturation (SeCSeS) of a
haplobasaltic melt were investigated as a function of
silicate melt FeO content at 1400 C and 15 GPa.
1. SCSS and SeCSeS are shown to have an asymmetric U-shaped dependence on the FeO content of the
silicate melt, with a minimum at 5 wt % FeO. The
asymmetric U-shaped dependence arises as a result
1422
2.
3.
4.
5.
6.
of the interplay between the competing role of FeO
in two terms of the expression for SCSS: the positive
effect of FeO on CS, and the negative effect of ln
(silicate melt)
aFeO
.
Models for SCSS that include only a single term for
the mole fraction or mass fraction of FeO will not accurately describe SCSS at both low and high FeO
contents.
SCSS in low-temperature hydrous silicic melts appears to be similarly controlled by melt FeO content.
The experiments of Bradbury (1983) and Clemente
et al. (2004) conducted at FeO contents <30 wt %
FeO exhibit the characteristic negative logarithm dependence of SCSS on FeO.
The overall controls on SeCSeS are similar to those
on SCSS, such that the behavior of Se would be expected to follow that of S in reduced, dry basaltic
magmatic systems.
The thermochemistry of iron selenide melts is
clearly different from that of sulfide melts, manifest
as increasing non-stoichiometry with decreasing
FeO content, and thus increasing f Se2.
Re solubility in sulfide liquid increases with increasing pressure.
Journal of Petrology, 2015, Vol. 56, No. 7
ACKNOWLEDGEMENTS
The authors acknowledge Frank Brink (Centre for
Advanced Microscopy, ANU) for assistance with SEM
EDS; Nick Ware, Robert Rapp (Research School of Earth
Sciences, ANU), John Donovan and Julie Barkman
(CAMCOR, University of Oregon) for assistance with
EPMA; David Clark and Dean Scott (Research School of
Earth Sciences, ANU) for maintenance of high-pressure
equipment. Alexey Ariskin, Sébastian Jégo and Bruno
Scaillet are thanked for their constructive reviews, and
Alastair Lumsden and Marjorie Wilson are thanked for
editorial handling.
FUNDING
This work was supported by the block grant to the
Research School of Earth Sciences, Australian National
University. J.L.W. was supported by an Australian
Postgraduate Award, a Jaeger Scholarship and an ANU
Supplementary Scholarship.
SUPPLEMENTARY DATA
Supplementary data for this paper are available at
Journal of Petrology online.
REFERENCES
Fig. 14. Application of the O’Neill & Mavrogenes (2002) model
to silicate liquid compositions relevant to Mercury at conditions of 1400 C, 1 bar and aFeS ¼ 1. The Intercrater Plains and
Northern Volcanic Plains compositions of Stockstill-Cahill et al.
(2012) were determined by remote sensing, whereas the compositions from McCoy et al. (1999) and Berthet et al. (2009)
were generated by partial melting of the Indarch enstatite
chondrite. EPMA analyses from the 1400 C experiments from
the two partial melting studies are shown for comparison.
Compositions are given in Table 5. The cation coefficient for
Mn from Evans et al. (2008) was used in the modelling.
Ariskin, A. A., Danyushevsky, L. V., Bychkov, K. A., McNeill, A.
W., Barmina, G. S. & Nikolaev, G. S. (2013). Modeling solubility of Fe–Ni sulfides in basaltic magmas: the effect of
nickel. Economic Geology 108, 1983–2003.
Baker, D. R. & Moretti, R. (2011). Modelling the solubility of sulfur in magmas: A 50-year old geochemical challenge. In:
Behrens, H. & Webster, J. D. (eds) Sulfur in Magmas and
Melts: its Importance for Natural and Technical Processes.
Mineralogical Society of America and Geochemical Society,
Reviews in Mineralogy and Geochemistry 73, 167–213.
Berry, A. J., O’Neill, H. S. C., Jayasuriya, K. D., Campbell, S. J. &
Foran, G. J. (2003). XANES calibrations for the oxidation
state of iron in a silicate glass. American Mineralogist 88,
967–977.
Berthet, S., Malavergne, V. & Righter, K. (2009). Melting of the
Indarch meteorite (EH4 chondrite) at 1 GPa and variable oxygen fugacity: Implications for early planetary differentiation
processes. Geochimica et Cosmochimica Acta 73, 6402–
6420.
Bingham, P. A., Connelly, A. J., Cassingham, N. J. & Hyatt, N. C.
(2011). Oxidation state and local environment of selenium in
alkali borosilicate glasses for radioactive waste immobilisation. Journal of Non-Crystalline Solids 357, 2726–2734.
Bradbury, J. W. (1983). Pyrrhotite solubility in hydrous albite
melts. PhD thesis, Pennsylvania State University, State
College, PA.
Brenan, J. M. (2003). Effects of f O2, f S2, temperature, and melt
composition on Fe–Ni exchange between olivine and sulfide
liquid: Implications for natural olivine–sulfide assemblages.
Geochimica et Cosmochimica Acta 67, 2663–2681.
Burnham, C. W. (1979a). Magmas and hydrothermal fluids. In:
Barnes, H. L. (ed.) Geochemistry of Hydrothermal Ore
Deposits. John Wiley, pp. 71–136.
Burnham, C. W. (1979b). The importance of volatile constituents. In: Yoder, H. S. (ed.) The Evolution of the Igneous
Journal of Petrology, 2015, Vol. 56, No. 7
Rocks: Fiftieth Anniversary Perspectives. Princeton
University Press, pp. 439–482.
Clemente, B., Scaillet, B. & Pichavant, M. (2004). The solubility
of sulphur in hydrous rhyolitic melts. Journal of Petrology
45, 2171–2196.
Cottrell, E. & Walker, D. (2006). Constraints on core formation from Pt partitioning in mafic silicate liquids at high
temperatures. Geochimica et Cosmochimica Acta 70,
1565–1580.
Czamanske, G. K. & Moore, J. G. (1977). Composition and phase
chemistry of sulfide globules in basalt from the Mid-Atlantic
Ridge rift valley near 37 N lat. Geological Society of
America Bulletin 88, 587–599.
Doyle, C. D. & Naldrett, A. J. (1987). Ideal mixing of divalent cations in mafic magma. II. The solution of NiO and the partitioning of nickel between coexisting olivine and liquid.
Geochimica et Cosmochimica Acta 51, 213–219.
Evans, K. A., O’Neill, H. S. C. & Mavrogenes, J. A. (2008).
Sulphur solubility and sulphide immiscibility in silicate
melts as a function of the concentration of manganese,
nickel, tungsten and copper at 1 atm and 1400 C. Chemical
Geology 255, 236–249.
Evans, L. G., Peplowski, P. N., Rhodes, E. A., et al. (2012). Majorelement abundances on the surface of Mercury: Results
from the MESSENGER Gamma-Ray Spectrometer. Journal
of Geophysical Research: Planets 117, doi:10.1029/
2012JE004178.
Fang, L. & Lynch, D. C. (1987). Evaluation of the behavior of
selenium in silicate slag. Metallurgical Transactions B 18,
181–187.
Fincham, C. J. B. & Richardson, F. D. (1954). The behaviour of
sulphur in silicate and aluminate melts. Proceedings of the
Royal Society of London, Series A 223, 40–62.
Fleet, M. E., Liu, X., Harmer, S. L. & King, P. L. (2005). Sulfur Kedge XANES spectroscopy: Chemical state and content of
sulfur in silicate glasses. Canadian Mineralogist 43, 1605–
1618.
Fogel, R. A., Weisberg, M. K. & Prinz, M. (1996). The solubility of
CaS in aubrite silicate melts. Proceedings of the Lunar and
Planetary Science Conference 27, 371–372.
Fonseca, R. O. C., Mallmann, G., O’Neill, H. S. C. & Campbell, I.
H. (2007). How chalcophile is rhenium? An experimental
study of the solubility of Re in sulphide mattes. Earth and
Planetary Science Letters 260, 537–548.
Francis, R. D. (1990). Sulfide globules in mid-ocean ridge basalts (MORB), and the effect of oxygen abundance in Fe–S–O
liquids on the ability of those liquids to partition metals from
MORB and komatiite magmas. Chemical Geology 85,
199–213.
Holzheid, A. & Grove, T. L. (2002). Sulfur saturation limits in silicate melts and their implications for core formation scenarios for terrestrial planets. American Mineralogist 87,
227–237.
Hsieh, K.-C. & Chang, Y. A. (1987). Thermochemical description
of the ternary iron–nickel–sulfur system. Canadian
Metallurgical Quarterly 26, 311–327.
Jakobsson, S. & Oskarsson, N. (1994). The system C–O in equilibrium with graphite at high pressure and temperature: An
experimental study. Geochimica et Cosmochimica Acta 58,
9–17.
Jarosewich, E., Nelen, J. A. & Norberg, J. A. (1980). Reference
samples for electron microprobe analysis. Geostandards
Newsletter 4, 43–47.
Jenner, F. E. & O’Neill, H. S. C. (2012). Analysis of 60 elements
in 616 ocean floor basaltic glasses. Geochemistry,
Geophysics, Geosystems 13, Q02005.
1423
Jenner, F. E., O’Neill, H. S. C., Arculus, R. J. & Mavrogenes, J. A.
(2010). The magnetite crisis in the evolution of arc-related
magmas and the initial concentration of Au, Ag and Cu.
Journal of Petrology 51, 2445–2464.
Johnston, M. D., Jahanshahi, S. & Lincoln, F. J. (2007).
Thermodynamics of selenium and tellurium in calcium ferrite slags. Metallurgical and Materials Transactions B 38,
433–442.
Johnston, M. D., Jahanshahi, S., Zhang, L. & Lincoln, F. J.
(2010). Effect of slag basicity on phase equilibria and selenium and tellurium distribution in magnesia-saturated calcium iron silicate slags. Metallurgical and Materials
Transactions B 41, 625–635.
Jugo, P. J., Wilke, M. & Botcharnikov, R. E. (2010). Sulfur Kedge XANES analysis of natural and synthetic basaltic
glasses: Implications for S speciation and S content as function of oxygen fugacity. Geochimica et Cosmochimica Acta
74, 5926–5938.
Klemme, S. & O’Neill, H. S. C. (2000). The effect of Cr on the
solubility of Al in orthopyroxene: experiments and thermodynamic modelling. Contributions to Mineralogy and
Petrology 140, 84–98.
Klimm, K. & Botcharnikov, R. E. (2010). The determination of
sulfate and sulfide species in hydrous silicate glasses using
Raman spectroscopy. American Mineralogist 95, 1574–
1579.
Klimm, K., Kohn, S. C. & Botcharnikov, R. E. (2012). The dissolution mechanism of sulphur in hydrous silicate melts. II:
Solubility and speciation of sulphur in hydrous silicate melts
as a function of f O2. Chemical Geology 322–323, 250–267.
Kress, V. (2007). Thermochemistry of sulfide liquids III: Ni-bearing liquids at 1 bar. Contributions to Mineralogy and
Petrology 154, 191–204.
Kress, V. C. & Carmichael, I. S. E. (1991). The compressibility of
silicate liquids containing Fe2O3 and the effect of composition, temperature, oxygen fugacity and pressure on their
redox states. Contributions to Mineralogy and Petrology
108, 82–92.
Lesne, P., Kohn, S. C., Blundy, J., Witham, F., Botcharnikov, R.
E. & Behrens, H. (2011). Experimental simulation of closedsystem degassing in the system basalt–H2O–CO2–S–Cl.
Journal of Petrology 52, 1737–1762.
Li, C. & Ripley, E. M. (2005). Empirical equations to predict the
sulfur content of mafic magmas at sulfide saturation and applications to magmatic sulfide deposits. Mineralium
Deposita 40, 218–230.
Li, C. & Ripley, E. M. (2009). Sulfur contents at sulfide-liquid
or anhydrite saturation in silicate melts: empirical equations and example applications. Economic Geology 104,
405–412.
Liu, Y., Samaha, N.-T. & Baker, D. R. (2007). Sulfur concentration at sulfide saturation (SCSS) in magmatic silicate melts.
Geochimica et Cosmochimica Acta 71, 1783–1799.
Llovet, X., Pinard, P. T., Donovan, J. J. & Salvat, F. (2012).
Secondary fluorescence in electron probe microanalysis of
material couples. Journal of Physics D: Applied Physics 45,
225301.
Malavergne, V., Berthet, S. & Righter, K. (2007). Formation of
CaS–MgS in enstatite chondrites and achondrites as a function of redox conditions and temperature: constraints on
their evolution in a planetesimal and in a proto-planet.
Proceedings of the Lunar and Planetary Science Conference
38, 1737–1738.
Markus, R. S. & Baker, D. R. (1989). Sulfur solubility in anhydrous andesitic melts. EOS Transactions, American
Geophysical Union 70, 1402.
1424
Mathez, E. A. (1976). Sulfur solubility and magmatic sulfides in
submarine basalt glass. Journal of Geophysical Research
81, 4269–4276.
Mavrogenes, J. A. & O’Neill, H. S. C. (1999). The relative effects
of pressure, temperature and oxygen fugacity on the solubility of sulfide in mafic magmas. Geochimica et
Cosmochimica Acta 63, 1173–1180.
McCoy, T. J., Dickinson, T. L. & Lofgren, G. E. (1999). Partial
melting of the Indarch (EH4) meteorite: A textural, chemical,
and phase relations view of melting and melt migration.
Meteoritics and Planetary Science 34, 735–746.
McNeill, A. W., Danyushevsky, L. V., Ariskin, A. A., Bychkov, K.
A. & Perfit, M. (2010). The Siqueiros Transform Fault MORB;
a tale of sulfur-saturation. 11th International Platinum
Symposium, Ontario Geological Survey, Miscellaneous
Release Data 269.
Médard, E., McCammon, C. A., Barr, J. A. & Grove, T. L. (2008).
Oxygen fugacity, temperature reproducibility, and H2O contents of nominally anhydrous piston-cylinder experiments
using graphite capsules. American Mineralogist 93, 1838–
1844.
Métrich, N., Berry, A. J., O’Neill, H. S. C. & Susini, J. (2009). The
oxidation state of sulfur in synthetic and natural glasses
determined by X-ray absorption spectroscopy. Geochimica
et Cosmochimica Acta 73, 2382–2399.
Morgan, G. B. & London, D. (2005). Effect of current density on
the electron microprobe analysis of alkali aluminosilicate
glasses. American Mineralogist 90, 1131–1138.
Müller-Simon, H., Bauer, J. & Baumann, P. (2001). Redox behaviour of selenium in industrial soda–lime–silica glasses.
Glastechnische Berichte 74, 283–291.
Nagamori, M. & Mackey, P. J. (1977). Distribution equilibria of
Sn, Se and Te between FeO–Fe2O3–SiO2–Al2O3–CuO05
slag and metallic copper. Metallurgical Transactions B 8,
39–46.
Nittler, L. R., Richard, D. S., Shoshana, Z. et al. (2011). The
Major-Element Composition of Mercury s Surface from
MESSENGER X-Ray Spectrometry. Science 333, 1847–1850.
O’Neill, H. S. C. & Eggins, S. M. (2002). The effect of melt composition on trace element partitioning: an experimental investigation of the activity coefficients of FeO, NiO, CoO,
MoO2 and MoO3 in silicate melts. Chemical Geology 186,
151–181.
O’Neill, H. S. C. & Mavrogenes, J. A. (2002). The sulfide capacity
and sulfur content at sulfide saturation of silicate melts at
1400 C and 1 bar. Journal of Petrology 43, 1049–1087.
Patten, C., Barnes, S. -J. & Mathez, E. A. (2012). Textural variations in MORB sulfide droplets due to differences in crystallization history. Canadian Mineralogist 50, 675–692.
Patten, C., Barnes, S. -J., Mathez, E. A. & Jenner, F. E. (2013).
Partition coefficients of chalcophile elements between sulfide and silicate melts and the early crystallization history of
sulfide liquid: LA-ICP-MS analysis of MORB sulfide droplets.
Chemical Geology 358, 170–188.
Pedersen, A. K. (1979). Basaltic glass with high-temperature
equilibrated immiscible sulphide bodies with native iron
Journal of Petrology, 2015, Vol. 56, No. 7
from Disko, central West Greenland. Contributions to
Mineralogy and Petrology 69, 397–407.
Rüssel, C. (2000). Electrochemical study on the redox behavior
of selenium-containing soda–lime–silica melts. Glastechnische Berichte 74, 1–5.
Schreiber, H. D. (1987). An electrochemical series of redox couples in silicate melts: A review and applications to geochemistry. Journal of Geophysical Research: Solid Earth 92,
9225–9232.
Schreiber, H. D., Leonhard, P. G., Nofsinger, R. G., Henning, M. W.,
Schreiber, C. W. & Kozak, S. J. (1988). Oxidation–reduction
chemistry of nonmetals in a reference borosilicate melt. In:
Bickford, D. F., Horsafall, W. E. & Woolley, F. E. (eds) Advances
in the Fusion of Glass: Proceedings of the 1st International
Conference on Advances in the Fusion of Glass, American
Ceramic Society pp. 29.1–29.14.
Schreiber, H. D. & Schreiber, C. W. (1993). Polyselenide formation in borosilicate glasses. Journal of Non-Crystalline
Solids 155, 209–220.
Shima, H. & Naldrett, A. J. (1975). Solubility of sulfur in an ultramafic melt and the relevance of the system Fe–S–O.
Economic Geology 70, 960–967.
Stockstill-Cahill, K. R., McCoy, T. J., Nittler, L. R., Weider, S. Z. &
Hauck, S. A. (2012). Magnesium-rich crustal compositions
on Mercury: Implications for magmatism from petrologic
modeling. Journal of Geophysical Research: Planets 117,
doi:10.1029/2012JE004140.
Tsujimura, T. & Kitakaze, A. (2005). Experimental study of sulfur
solubility in silicate melts coexisting with graphite as a function of silicate melt composition. Resource Geology 55, 55–
60.
Ulmer, P. & Luth, R. W. (1991). The graphite–COH fluid equilibrium in P, T, f O2 space. Contributions to Mineralogy and
Petrology 106, 265–272.
Weider, S. Z., Nittler, L. R., Starr, R. D., McCoy, T. J., StockstillCahill, K. R., Byrne, P. K., Denevi, B. W., Head, J. W. &
Solomon, S. C. (2012). Chemical heterogeneity on Mercury’s
surface revealed by the MESSENGER X-Ray Spectrometer.
Journal of Geophysical Research: Planets 117, doi:10.1029/
2012JE004153.
Wendlandt, R. F. (1982). Sulfide saturation of basalt and andesite melts at high pressures and temperatures. American
Mineralogist 67, 877–885.
Weyl, W. A. (1959). Glasses Coloured by Selenium and
Selenides. In: Coloured Glasses. Dawson, pp. 282–323.
Wilke, M., Partzsch, G. M., Bernhardt, R. & Lattard, D. (2004).
Determination of the iron oxidation state in basaltic
glasses using XANES at the K-edge. Chemical Geology 213,
71–87.
Wykes, J. L. (2014). Selenium and sulfur in silicate melts. PhD
thesis, Research School of Earth Sciences, Australian
National University, Acton, ACT.
Zolotov, M. Y., Sprague, A. L., Hauck, S. A., Nittler, L. R.,
Solomon, S. C. & Weider, S. Z. (2013). The redox state, FeO
content, and origin of sulfur-rich magmas on Mercury.
Journal of Geophysical Research: Planets 118, 138–146.