This document summarizes an experiment that studied how the composition of molten sulfide liquids affects their ability to wet and spread across olivine, an important mantle mineral. The experiment varied the oxygen and sulfur fugacity during heating to simulate mantle conditions, and added different amounts of nickel, copper and cobalt to the molten sulfide to model natural sulfide compositions. The results showed that sulfide wetting decreased with lower oxygen fugacity and increased metal content in the sulfide. Wetting controls how sulfides distribute and migrate within cooling magmas and the mantle, influencing ore deposit formation and elemental transport.

1. 0361-0128/01/3125/145-13 $6.00 145
Introduction
AS A CONSEQUENCE of changes in fO2, fS2, activity of FeO, or
absolute sulfur content, basaltic magmas may achieve satura-
tion in an immiscible sulfide liquid at some stage in their crys-
tallization history. Owing to their strong affinity for chal-
cophile and siderophile elements, sulfide liquids are expected
to exert a controlling influence on the distribution of these el-
ements within the earth’s crust and upper mantle (e.g., Stone
et al., 1990; Barnes and Picard, 1993; Peach and Mathez, 1993;
Gaetani and Grove, 1997). However, despite both the geo-
chemical and economic importance of magmatic sulfide liq-
uids, our understanding of the conditions that control their
grain-scale distribution and ultimate mobility is far from com-
prehensive. Such constraints have the potential to shed sig-
nificant light on the formation of disseminated as opposed to
massive magmatic sulfide ore deposits as well as the conditions
necessary for sulfide liquid migration in the earth’s mantle.
In general, the local distribution of a liquid phase within a co-
existing solid matrix is controlled by the relative energies of the
solid-liquid and solid-solid interfaces. Subject to the conditions
of both mechanical and chemical equilibrium, and assuming
that the surface energy of the solid matrix is isotropic, the
three-dimensional melt topology can be predicted with
knowledge of the cross sectional geometry of the solid-liquid
interface. The dihedral angle (Θ) is defined as the angle
formed by two intersecting walls of a liquid-filled pore at a
junction with two solid grains. The magnitude of Θ is deter-
mined by the balance of forces at the solid-liquid junction and
is expressed by the relation:
Θ = 2 arcos (γSS/2γSL), (1)
where γSS and γSL are the solid-solid and solid-liquid surface
energies, respectively.
The value of Θ is determined by the amount of solid-liquid
contact required to minimize the total surface energy of the
system. If Θ < 60°, the melt is “wetting” and occupies pris-
matic grain edge channels, which remain open allowing an in-
terconnected melt phase in three dimensions, even at low
melt fractions (i.e., <<1 vol %; von Bargen and Waff, 1986).
Conversely, in cases where Θ > 60°, the melt is “nonwetting”
and grain edges become dry as a result of the liquid phase
“beading-up” at grain-edge intersections. For Θ > 60°, melt
Wetting Properties of Fe-Ni-Co-Cu-O-S Melts against Olivine:
Implications for Sulfide Melt Mobility
LESLEY A. ROSE AND JAMES M. BRENAN†
Department of Geology, University of Toronto, 22 Russell Street, Toronto, Ontario, Canada M5S 3B1
Abstract
The effects of transition metal content and the fugacities of oxygen (fO2) and sulfur (fS2) on the wetting be-
havior of molten sulfide against forsteritic olivine have been investigated in this study. Wetting behavior, or the
mobility of a liquid phase within a coexisting solid matrix, is quantified by the dihedral angle, Θ, which is a func-
tion of the relative solid-solid and solid-liquid surface energies. In order to determine sulfide wetting behavior,
experiments were performed at 1,300°C using a vertical tube furnace employing C-O-S gas mixtures to con-
trol fO2 and fS2. Samples consisted of mixtures of San Carlos olivine and Fe + (Ni, Cu, Co; 0–30 wt %) + S melts
equilibrated for 24 to 168 hr. Experimental conditions ranged from fO2 = 10–8
to 10–10
and fS2 from 10–1.2
to 10–4
in accordance with values appropriate for basalt petrogenesis. Results of our experiments revealed that dihe-
dral angles exhibited a marked increase with decreasing fO2, and variable dependence on melt metal composi-
tion. At fO2 = 10–8
, all sulfide melt compositions were found to be wetting (i.e., Θ < 60°), whereas only those
with <~15 wt percent added Cu, Co, or Ni were wetting at fO2 = 10–9
, and no wetting compositions were en-
countered at fO2 = 10–10
. In agreement with the results of other investigators, we found that values of Θ de-
creased as the mole fraction of oxygen in the melt increased, suggesting that metal oxide species in the melt
are more likely to be surface-active with respect to olivine.
In light of our experimental data, it is expected that the wetting behavior of natural sulfide liquids will de-
pend on both the identity and quantity of nonferrous metal and the abundance of dissolved oxygen. Because
of this latter effect, the prevailing conditions of both fO2 and fS2 are therefore likely to dictate sulfide melt mo-
bility. In terms of the potential for sulfide melt metasomatism in the upper mantle, consideration of the range
of fO2 reflected by natural sulfide liquids, mafic lavas, and upper mantle source regions reveals that conditions
for sulfide melt mobility encompass much of this spectrum, even for many Ni- and Cu-rich natural liquid com-
positions. Thus, such liquids may be potent agents for redistributing siderophile and chalcophile elements in
the upper mantle.
Sulfide melt wetting behavior will also play a role in the final sulfide distribution on solidification of mafic-
ultramafic magmas that achieve saturation in an immiscible sulfide liquid. Efficient sulfide segregation may
occur in reduced magmas only by early sulfide settling through a largely liquid medium, inasmuch as late-
formed sulfide liquid will become trapped in the solid silicate matrix. For the case of relatively oxidized mag-
mas, the wettability of sulfide liquid at these conditions, combined with a low viscosity and high density, sug-
gests that efficient compaction-driven sulfide segregation is possible, even over the relatively short cooling
intervals likely for high level mafic intrusions.
Economic Geology
Vol. 96, 2001, pp. 145–157
†
Corresponding author: e-mail, brenan@zircon.geology.utoronto.ca

2. connectivity is achieved only above a finite fraction (i.e., 2 vol
% for Θ = 65°; von Bargen and Waff, 1986).
In recent studies, dihedral angles have been measured of
simple Fe-S melts at both high (Ballhaus and Ellis, 1996; Mi-
narik et al., 1996) and low pressure (Gaetani and Grove, 1999).
Based on the results of 1-atm experiments in which both fS2
and fO2 were controlled, Gaetani and Grove (1999) proposed
that the main control on the grain-scale distribution of sulfide
liquid within an olivine-rich matrix was the FeO content of
the melt. Specifically, their results demonstrate that the olivine-
sulfide melt dihedral angle decreases significantly with an in-
crease in the concentration of dissolved oxygen, from a high
of 90° with 0.09 wt percent oxygen to values less than 60° for
melts having greater than 2.7 wt percent oxygen. Inasmuch as
the oxygen content of the sulfide liquid was found to vary sys-
tematically with fO2, Gaetani and Grove (1999) concluded that
sulfide liquids are likely to be wetting, and hence mobile, at
fO2 values prevalent in the upper mantle.
Results from simple Fe-S-O systems, however, may not
mirror those obtained from more compositionally complex
natural sulfide liquids. Indeed, observations of the wetting
properties of sulfide liquids containing other additives tend to
support this claim. For example, in experiments that involved
sulfide samples encapsulated in evacuated silica tubes, Ebel
and Naldrett (1996) observed enhanced wetting of the tube
walls with increased Cu content of the sulfide liquid. In addi-
tion, Kress (1997) observed different textures in runs at con-
stant fO2 and fS2 and variable fH2. At high fH2, sulfides ap-
peared as large globules within the silicate liquid whereas at
lower fH2, silicate and sulfide liquids were intermingled on a
much finer scale. Thus, it would appear that both the transi-
tion element and H2 content of the melt could influence sul-
fide wetting behavior against silicate materials.
Natural sulfide globules occurring in volcanic rocks are the
best-preserved samples of immiscible sulfide liquids and con-
tain appreciable amounts of components other than Fe, S,
and O. In terms of other elements at the major element level,
transition metals such as Ni (up to 30 wt %), Cu (up to 40 wt
%), and Co (up to 2 wt %) are the most common, and minor
amounts of Mn, Ti, and V are also common (Roy-Barmen et
al., 1998; Stone and Fleet, 1991; Peach et al., 1990; Pedersen,
1979; Czamanske and Moore, 1977; Mathez and Yeats, 1976;
see Fig. 1). Furthermore, late-stage fractionation or prolonged
silicate liquid-sulfide liquid exchange can produce more ex-
treme enrichments in these metals, as represented by the
compositions of sulfides from magmatic sulfide ore deposits
(Naldrett, 1989). In light of the compositional complexity of
natural sulfide liquids, we have been motivated to undertake
a study of the effect of transition metals other than Fe on the
wetting of sulfide melt against olivine, the most abundant
mineral in the upper mantle and an important phase in mafic-
ultramafic plutonic rocks. Sulfide liquid compositions were
thus chosen to reflect the range of Co, Cu, and Ni contents
encountered in natural sulfide samples (Fig. 1) and experi-
ments were performed at controlled oxygen and sulfur fugac-
ities appropriate to basalt petrogenesis.
Experimental and Analytical Methods
Experiments were performed in a modified 1-atm vertical
tube furnace using mixtures of CO, CO2, and SO2 gases to
control oxygen and sulfur fugacity. Brenan and Caciagli
(2000) published additional details of the furnace design and
experimental procedures used in this study. Experiments
were performed at oxygen fugacities between 10–8
and 10–10
,
consistent with those estimated for terrestrial basaltic mag-
mas (Christie et al., 1986). Sulfur fugacities used in our ex-
periments were between 10–1.2
and 10–4
, which are sufficient
for sulfide liquid stability and closely resemble fS2 values esti-
mated for natural samples (Wallace and Carmichael, 1992). A
summary of experiment conditions is provided in Table 1.
Relative flow rates of each gas needed to produce specific
oxygen and sulfur fugacities were determined using a pro-
gram that calculates the equilibrium speciation of 28 C-O-H-
S gas species at 1 atm and high temperature (courtesy of Vic-
tor Kress). Flow rates were controlled using flow meters
calibrated in-house and predicted oxygen and sulfur fugaci-
ties from mixing ratios were checked against the nickel-nickel
oxide and (Au, Ag)S2 buffers (Barton and Toulmin, 1964).
Temperatures in the furnace hot spot were continuously mon-
itored using a Pt-Pt90Rh10 thermocouple calibrated against the
melting point of gold. The thermocouple was protected from
the corrosive sulfur atmosphere by an alumina sheath.
Samples were contained in crucibles fabricated from San
Carlos olivine megacrysts, a material chosen because it is
inert with respect to both the experimental sample and the
sulfur-rich atmosphere required for sulfide liquid stability.
Sulfide melt was prepared from high purity (99.99+
%) Fe, Ni,
Cu, Co, and S powders ground together under ethanol. A typ-
ical sample consisted of alternating layers of Fe + (Ni, Cu,
Co) + S (hereafter referred to as “sulfide melt”), powdered
146 ROSE AND BRENAN
0361-0128/98/000/000-00 $6.00 146
mol% Fe
0 10 20 30 40 50 60 70 80 90 100
mol% S + O
0
10
20
30
40
50
60
70
80
90
100
mol%
Ni + Cu + Co
0
10
20
30
40
50
60
70
80
90
100
MORB (Roy-Barman et al., 1998)
MORB (Mathez and Yeats, 1976)
MORB (Czamanske and Moore, 1977)
MORB (Peach et al., 1990)
Kilauea (Stone and Fleet, 1991)
DiskoIsland (Pedersen, 1979)
This Study
FeSXS
X3S2
FIG. 1. Range of natural sulfide compositions from various sources (open
symbols) and sulfide compositions from this study (filled symbols) plotted in
terms of molar quantities of sulfur + oxygen, iron, and other metals (Ni, Cu, Co).

3. San Carlos olivine, and in some cases powdered chromite (re-
sults discussed in Rose and Brenan, in prep.). Sulfide melt
and solid (olivine ± chromite) occurred in proportions of 30
to 50 percent melt and 70 to 50 percent solid. Approximately
20 to 50 mg in total (depending on crucible size) was layered
into the olivine crucible for a given experiment. The olivine
crucible was inserted into a silica cup, which was in turn
placed inside an alumina tube attached to a fused quartz
hook, and then suspended within the top (cold) zone of the
furnace. The furnace was then sealed and gas flow com-
menced. After allowing approximately half an hour for gas
equilibration, the sample was slowly lowered over a period of
15 to 30 min into the predetermined furnace hot spot. At the
end of the experiment, the bottom brass fitting of the furnace
was removed and the quartz rod was quickly plunged into
cold water to quench the sample.
Run products were mounted in epoxy and initially ground
using SiC. In most cases, a second impregnation of epoxy was
necessary after the sample had been exposed, due to the sep-
aration of olivine layers by void space that had originally been
occupied by sulfide melt. After a final polish with 0.3 µm alu-
mina powder (and/or colloidal silica), samples were carbon-
coated and analyzed using the Cameca SX50 electron micro-
probe (EMP) at the University of Toronto. Additional
analyses were performed using the JEOL JXA-8600 micro-
probe at the University of Western Ontario. Analytical condi-
tions for determination of the major elements in the experi-
mental sulfide melt were an accelerating voltage of 20 kV and
a beam current of 30 nA with maximum peak counting times
of 30 s. Depending on sulfide melt pocket size, a defocused
beam of between 5 and 10 µm was used for analysis due to
the textural inhomogeneity resulting from quench crystalliza-
tion. Oxygen was analyzed with an octadecanoate lead
(ODPB) pseudocrystal. Because of differences in peak posi-
tion and shape of the standard and unknown (stemming from
oxygen present in different electronic environments), oxygen
was analyzed by integrating peak areas instead of the more
traditional method of counting at a fixed peak position. Due
to the low energy of the oxygen Kα X-ray, even a small varia-
tion in carbon coat thickness between the standard and un-
known could affect the accuracy of measured oxygen concen-
trations. We found that differences in the carbon-coat
thickness between the hematite standard (used for oxygen
analysis) and mounted run products were negligible, how-
ever, since analyses completed after simultaneous standard-
sample carbon coating, or after only the sample had been re-
coated (without the standard), yielded similar results. Melts
from samples run at fO2 = 10–8
contained abundant dendritic
quench oxide intergrown with quenched sulfide. Melt com-
positions for those samples were calculated by combining
broad-beam analyses of quenched sulfide intergrowths with
focused beam oxide analyses weighted according to their re-
spective modes. Standards used for electron microprobe
analysis were chalcopyrite (Cu, S, Fe), pentlandite (Ni),
cobaltite (Co), and hematite (O). Raw count rates were con-
verted to concentrations using a modified ZAF correction
routine. A summary of measured and calculated sulfide liquid
compositions is provided in Table 2.
WETTING PROPERTIES OF Fe-Ni-CO-Cu-O-S MELTS AGAINST OLIVINE 147
0361-0128/98/000/000-00 $6.00 147
TABLE 1. Summary of Experimental Conditions1
Sample Duration Gas flow rates (ccm) Initial melt composition
no. Type (h) log fO2 log fS2 CO/SO2/CO2 Fe-(Cu,Co,Ni)-S wt %
FeS1 Olivine + chromite 72 –9.05 –2.1 20:03:29 70-0-30
FeS2 Olivine + chromite 72 –9.05 –2.1 20:03:29 70-0-30
FeS3 Olivine 72 –8 –2.0 13:12:21 70-0-30
Cu1 Olivine 72 –8 –2.0 13:12:21 65-5-30
Cu2 Olivine + chromite 71 –9.05 –2.1 20:03:29 65-5-30
Cu3 Olivine 72 –8 –2.0 13:12:21 50-20-30
Cu5 Olivine + chromite 72 –9.05 –2.06 20:03:29 40-30-30
Cu7 Olivine 72 –8 –2 13:12:21 55-15-30
Cu8 Olivine + chromite 71 –9.05 –2.06 20:03:29 55-15-30
Cu9c Olivine + chromite 24 –10.1 –2.06 35:02:17 55-15-30
Cu9a Olivine + chromite 96 –10.1 –2.06 35:02:17 55-15-30
Cu9b Olivine + chromite 168 –10.1 –2.06 35:02:17 55-15-30
Cu16B Olivine 72 –8.97 –1.24 22:10:00 65-5-30
Co2 Olivine + chromite 72 –9.05 –2.06 20:03:29 65-5-30
Co5 Olivine + chromite 73 –9.05 –2.06 20:03:29 40-30-30
Co7 Olivine 72 –8 –2 13:12:21 55-15-30
Co8 Olivine 72 –9.05 –2.06 20:03:29 55-15-30
Ni3 Olivine 72 –8 –2.02 13:12:21 50-20-30
D.ol 72 Olivine 48 –8.2 –2.1 14:09:30 55-15-30
D.ol 92 Olivine 48 –9.05 –2.06 20:03:29 40-30-30
D.ol 102 Olivine 48 –9.05 –2.06 20:03:29 65-5-30
D.ol 112 Olivine 48 –10.1 –2.06 35:02:17 65-5-30
D.ol 202 Olivine 120 –10.1 –2.06 35:02:17 65-5-30
D.ol 232 Olivine 96 –9 –4 00:34.0 40-30-30
1 All runs performed at 1,300°C
2 Samples from Brenan and Caciagli (2000)

4. In order to determine the true dihedral angle for each sam-
ple, approximately 100 angles were measured from digitized
secondary electron images using variable magnifications (de-
pending on olivine grain size). Analysis was performed using
the protractor function in ScionImage (ported from NIH
Image developed at the U.S. National Institute of Health and
available for download at http://www.scioncorp.com).
Harker and Parker (1945) showed that the median value of
a range of apparent measured angles represents the true di-
hedral angle and the associated uncertainty was calculated
using the method of Riegger and Van Vlack (1960). On the
basis of replicate measurements, the error associated with a
single measurement was found to be ±2°. A distribution of
apparent angles is expected (Fig. 2) as each sample was sec-
tioned along a plane where grains are randomly oriented with
respect to the bisectrix of the true dihedral angle. A summary
of dihedral angles is provided in Table 3.
The olivine-sulfide melt system behaves anisotropically
with respect to surface energy, as shown by samples produced
in this study containing some olivine grains with crystal facets.
Indeed, in any given sample, we observed that between 5 and
10 percent of the grain-grain junctions involved facetted crys-
tal edges. Dihedral angles involving such interfaces were not
included when determining the median value used to infer
the permeability of a given sample. The presence of faceted
edges introduces some uncertainty in this regard since their
effect on melt connectivity is difficult to predict, particularly
at low melt fractions. Waff and Faul (1992) have argued that
the presence of facets serves to increase matrix permeability
by reducing the occurrence of pinch-off along grain edge in-
tersections. However, in experiments that monitored bulk dif-
fusion in fluid-bearing pyroxene aggregates, Watson and
Lupulescu (1993) showed that the median dihedral angle in-
volving curved interfaces was a reliable predictor of perme-
ability, despite the presence of abundant faceted interfaces.
Experimental Results
Run products consisted of solid layers containing olivine or
chromite and quenched sulfide liquid. Over the course of an
experiment, melts appeared to have migrated from their ini-
tial position between the packed mineral layers, leaving inter-
layer void space behind. The resulting melt-impregnated lay-
ers maintained their initial position and appeared suspended
within the olivine crucible. Brenan and Caciagli (2000) had
initially observed that silicate melt was present in sample D.ol
9 at triple point junctions between intersecting olivine grains,
yet upon further investigation, this was found not to be the
case, as only sulfide melt was present.
Sulfide melts produced in our experiments consisted mainly
of Fe, S, and either Ni, Cu, or Co, in addition to varying amounts
of oxygen, depending on fO2 and the identity of the added
metals (Table 2). The aim of this study was to determine the
148 ROSE AND BRENAN
0361-0128/98/000/000-00 $6.00 148
TABLE 2. Analyses of Sulfide Melt (wt % element)1
Sample no. log fO2 log fS2 n2 Fe3 Cu, Co, Ni S O Total
FeS1 –9.05 –2.06 19 63.63 (1.12) –––––––
27.46 (1.88) 7.95 (1.13) 99.04
FeS2 –9.05 –2.06 16 65.14 (0.86) –––––––
26.70 (1.12) 8.07 (0.96) 99.91
FeS3 –8 –2 66.25 (1.72) –––––––
19.55 (0.50) 12.44 (0.83) 98.24
Cu1 –8 –2.02 64.58 (1.82) 2.32 (1.93) 19.06 (0.90) 12.86 (0.84) 98.82
Cu2 –9.05 –2.06 15 59.02 (1.48) 4.22 (1.10) 29.74 (1.41) 6.78 (1.22) 99.76
Cu3 –8 –2.02 16 37.49 (4.67) 30.92 (6.07) 27.94 (1.90) 3.21 (1.27) 99.56
Cu5 –9.05 –2.06 15 44.72 (2.35) 22.16 (2.58) 28.49 (1.58) 4.17 (1.06) 99.54
Cu7 –8 –2 53.81 (2.14) 13.74 (1.92) 23.81 (0.83) 7.99 (0.58) 99.35
Cu8 –9.05 –2.06 13 50.11 (1.98) 15.70 (2.76) 31.42 (1.66) 3.16 (0.20) 100.39
Cu9c –10.1 –2.06 13 49.66 (5.40) 14.86 (6.56) 32.23 (2.08) 3.04 (1.41) 99.79
Cu9a –10.1 –2.06 15 45.95 (5.40) 19.81 (6.29) 31.81 (1.24) 2.14 (0.45) 99.70
Cu9b –10.1 –2.06 13 52.00 (2.90) 12.20 (3.02) 30.83 (1.24) 4.35 (0.87) 99.38
Cu16B –8.97 –1.24 15 55.64 (1.82) 8.92 (2.11) 32.03 (1.11) 3.62 (1.01) 100.21
Co2 –9.05 –2.06 18 59.14 (0.89) 4.69 (0.39) 30.61 (1.86) 5.60 (1.49) 100.04
Co5 –9.05 –2.06 10 37.51 (2.92) 25.59 (2.62) 33.87 (0.99) 2.09 (0.52) 99.06
Co7 –8 –2 54.07 (1.52) 12.31 (0.74) 24.44 (0.69) 8.93 (0.82) 99.75
Co8 –9.05 –2.06 22 52.90 (1.09) 12.31 (0.89) 30.46 (1.33) 4.88 (1.13) 100.55
Ni3 –8 –2.02 16 30.31 (1.77) 39.69 (4.08) 29.08 (2.70) 0.96 (0.41) 100.06
D.ol 7 –8.2 –2.1 17 46.97 (3.66) 22.05 (5.10) 29.49 (1.15) 1.16 (0.78) 99.67
D.ol 9 –9.05 –2.06 16 35.93 (0.64) 32.03 (2.31) 32.56 (1.52) <0.054 100.52
D.ol 10 –9.05 –2.06 16 60.99 (1.09) 5.43 (2.56) 28.32 (1.72) 5.79 (0.78) 100.54
D.ol 11 –10.1 –2.06 17 59.44 (0.64) 4.72 (0.60) 34.56 (0.75) 0.63 (0.59) 99.35
D.ol 20 –10.1 –2.06 8 58.91 (0.61) 5.41 (0.77) 33.93 (0.62) 1.79 (0.57) 100.04
D.ol 23 –9 –4 23 22.63 (0.64) 52.25 (1.68) 25.93 (1.64) 0.27 (0.17) 101.08
1 Melt compositions from experiment FeS3, Cu1, Cu7 and Co7 were determined by electron microprobe analysis of individual quench phases, combined
with their modal abundances, all other melt compositions based on multiple broad-beam electron microprobe analyses
2 n is the number of analyses
3 Number in parentheses refers to the error based on one σ of n analyses
4 Minimum detection limit based on 2σ above background count rate

5. wetting behavior of sulfide melts whose compositions closely
resemble those found in nature. Good agreement exists be-
tween sulfide melt compositions produced in this study and
those of natural samples, as seen in Figure 1. Typically, most
natural samples fall on or near the FeS-XS and FeS-X3S2
joins. Sulfides from MORB from the FAMOUS area of the
mid-Atlantic ridge (Peach et al., 1990) and samples from
Disko Island, Greenland (Pedersen, 1979), however, are
more metal poor and metal rich, respectively, than these
joins, which probably reflect deviations in fO2 or fS2 from val-
ues typical for basalt petrogenesis.
Olivine-sulfide melt dihedral angles determined from our
experiments ranged from 44° to 80° (Table 3), and secondary
electron images of sectioned and polished run products de-
picting this variation are shown in Figure 3. By comparing
samples A and B, the effect of variable fO2 values can be seen
as the dihedral angle decreased from 60° to 48° with an in-
crease in fO2 from 10–9
to 10–8
. The effect of added transition
metals (in this case Co) to the sulfide melt at an fO2 of 10–9
can
be seen when comparing samples B, C, and D. As the metal
content increased from 5 to 12 to 26 wt percent, dihedral an-
gles increased from 58° to 60° to 73°. Similar effects were ob-
served when increased concentrations of copper and nickel
were added to the sulfide melt.
To test whether textural equilibrium had been achieved,
time series experiments were conducted for durations ranging
from 24 to 168 hr at constant fO2 and fS2 and initial melt com-
position. Figure 4 illustrates the results of these experiments.
Dihedral angles were invariant with experiment duration,
implying that equilibrium dihedral angles were established
quickly. With increasing maturity of the microstructure, the
distribution of apparent angles became narrower with respect
to the theoretical apparent angle distribution. Results from
the study of Gaetani and Grove (1999) showed similar behav-
ior and, as in that study, an experimental duration of 72 hr was
considered sufficient for producing textural equilibrium.
In order to determine the effect of added transition metals
on the resulting olivine-sulfide melt dihedral angles, experi-
ments were performed at constant fS2, fO2, and variable sulfide
melt composition. The effect of variable fO2 was also investi-
gated and results are displayed in Figure 5 for all experiments
conducted at an fS2 of 10–2
. Two experiments were conducted
at fO2 = 10–9
containing only Fe and S in the initial sulfide
melt, which yielded a median dihedral angle of 58°, whereas
a similar experiment at fO2 of 10–8
yielded a median angle of
WETTING PROPERTIES OF Fe-Ni-CO-Cu-O-S MELTS AGAINST OLIVINE 149
0361-0128/98/000/000-00 $6.00 149
0
5
10
15
20
25
30
35
40
0 40 80 120 160
Frequency
Apparent Angle (degrees)
Median = 73
o
Co5
0
5
10
15
20
25
30
35
40
0 40 80 120 160
Frequency
Apparent Angle (degrees)
Median = 58
o
Co2
A
B
FIG. 2. Comparison of measured and theoretical frequency distributions
for (A) sample Co5 where ~120 apparent angles were measured, and (B)
sample Co2 where ~100 apparent angles were measured. A distribution of
dihedral angles is expected because each run product was sectioned along a
random plane, resulting in a variety of apparent dihedral angles. The median
value of this distribution is taken to represent the true dihedral angle of the
sample and is inset on each histogram.
TABLE 3. Dihedral Angle Measurements
Sample no. n1 Dihedral angle2
FeS1 83 58.5 (1.9)
FeS2 106 56.5 (1.7)
FeS3 94 44.1 (1.2)
Cu1 93 47.3 (1.5)
Cu2 91 55.3 (2.1)
Cu3 144 44.5 (1.3)
Cu5 108 66.1 (1.9)
Cu7 83 50.0 (1.5)
Cu8 122 61.0 (1.9)
Cu9c 83 73.7 (1.8)
Cu9a 90 78.3 (1.6)
Cu9b 73 75.4 (1.8)
Cu16B 114 57.0 (1.8)
Co2 111 57.5 (1.7)
Co5 124 73.4 (1.9)
Co7 100 47.9 (1.6)
Co8 81 60.0 (1.7)
Ni3 171 46.4 (1.2)
D.ol 7 95 50.0 (1.5)
D.ol 9 101 79.5 (2.1)
D.ol 10 98 61.2 (2.3)
D.ol 11 75 72.1 (2.9)
D.ol 20 100 71.1 (1.8)
D.ol 23 101 57.0 (1.6)
1 Number of measured dihedral angles included in median calculation
2 Number in parentheses refers to the error based on one σ of n angles
given by the equation from Riegger and Van Vlack (1960)

6. 150 ROSE AND BRENAN
0361-0128/98/000/000-00 $6.00 150
olivine
olivine
olivine
A B
D
sulfide
melt
sulfide
melt
sulfide
melt
sulfide
melt
olivine
C
FIG. 3. Secondary electron images of run products consisting of Co-bearing sulfide melt and olivine. The effect of fO2 vari-
ation is shown when samples in (A) and (B) are compared, and the effect of the addition of transition metals (Co in this case)
to the sulfide melt (at constant fO2) is seen when comparing samples in (B), (C), and (D). Samples include (A) Co7, where
fO2 = 10–8
, Θ ~ 48°, and ~12 wt percent Co is added to sulfide melt, (B) Co8 where fO2 = 10–9
, Θ = ~60°, and ~12 wt per-
cent Co is added to sulfide melt, (C) Co5 where fO2 = 10–9
, Θ = ~ 73°, and ~26 wt percent Co is added to sulfide melt, and
(D) Co2 where fO2 = 10–9
, Θ = ~58°, and ~5 wt percent is added to sulfide melt. Scale bars represent 1 µm.

7. 44°. Increases in the abundance of transition metals replacing
Fe produced progressive increases in measured dihedral an-
gles for oxygen fugacities of 10–9
and 10–10
. However, mea-
sured dihedral angles for experiments performed at fO2 = 10–8
did not vary with increasing nonferrous metal content. For all
compositions investigated, dihedral angles increased as fO2
decreased. All sulfide melts produced at fO2 = 10–8
yielded Θ
values < 60° whereas all melts at 10–10
were nonwetting, re-
gardless of transition metal content. Experiments conducted
at an fO2 of 10–9
remained wetting (as defined by Θ < 60°)
until more than approximately 15 wt percent Cu, Ni, or Co
was incorporated into the sulfide liquid.
WETTING PROPERTIES OF Fe-Ni-CO-Cu-O-S MELTS AGAINST OLIVINE 151
0361-0128/98/000/000-00 $6.00 151
65
70
75
80
85
90
95
0 50 100 150 200 250
Cu9a, b, c
(this study)
FeS4, FeS7,
FeS1, FeS3
(Gaetani and
Grove, 1999)
D.ol 11,D.ol 20
(Brenan and
Caciagli, 2000)
DihedralAngle(degrees)
Time (hours)
FIG. 4. Variation in dihedral angle as a function of run duration for Cu-doped experiments from this study, Ni-doped ex-
periments from Brenan and Caciagli (2000), and Fe-bearing experiments from Gaetani and Grove (1999). The error associ-
ated with dihedral angle measurement is based on one standard deviation of n measured apparent angles given by the equa-
tion from Riegger and Van Vlack (1960).
Copper
Nickel
40
50
60
70
80
0 10 20 30 40 50
Fe Only
Copper
Cobalt
Nickel
Fe Only
Copper
Cobalt
Nickel
DihedralAngle(degrees)
wt% non-ferrous metal analyzed
within sulfide melt
fO
2
=10
-10
fO
2
=10
-9
fO
2
=10
-8
FIG. 5. Dihedral angle plotted against wt percent transition metal added to the sulfide melt for three distinct oxygen fu-
gacities at constant fS2 of 10–2
. As fO2 increases, dihedral angles decrease. At fO2 values of 10–9
and 10–10
, the dihedral angle
increases as nonferrous transition metals are added to the sulfide melt. At an fO2 of 10–8
, however, dihedral angles remain
fairly constant over the same compositional range. The error associated with dihedral angle measurement is based on one
standard deviation of n measured apparent angles given by the equation from Riegger and Van Vlack (1960); the error on
melt metal content is based on one standard deviation of n analyses.

8. Discussion
Variation of Θ with melt composition
Assuming that the solid-solid interfacial energy does not
vary appreciably with olivine composition, the observed vari-
ation in Θ with melt metal content, fO2, and fS2 can be attrib-
uted to the effects of these variables on the solid-liquid inter-
facial energy. In turn, it is reasonable to suggest that metal
content, fO2, and fS2 dictate the proportions of certain olivine
surface-active species within the melt. For the melts pro-
duced in our experiments, the relevant metal speciation reac-
tions take the forms:
Fe1 – xOsulfide melt + 1⁄2S2(g) = Fe1 – xSsulfide melt + 1⁄2O2(g), (2)
Ni1 – xOsulfide melt + 1⁄2S2(g) = Ni1 – xSsulfide melt + 1⁄2O2(g), (3)
Cu1 – xOsulfide melt + 1⁄2S2(g) = Cu1 – xSsulfide melt + 1⁄2O2(g), (4)
and
Co1 – xOsulfide melt + 1⁄2S2(g) = Co1 – xSsulfide melt + 1⁄2O2(g), (5)
in which the subscript 1 – x is used to emphasize the uncer-
tainty in the stoichiometry of a particular sulfide or oxide
species. Based on a well-defined correlation between Θ and
the oxygen content of the melt, Gaetani and Grove (1999)
concluded that the most likely olivine surface-active compo-
nent in Fe-S-O-bearing melts is an Fe oxide species of un-
known stoichiometry. As portrayed in Figure 6A and B, for
sulfide melts equilibrated at fO2 = 10–9
and 10–8
, respectively,
the mole percent of oxygen within the sulfide melt is posi-
tively correlated with the mole percent of iron, suggesting
that the dominant oxide species in these mixed metal melts is
Fe bearing. Although there is insufficient data to establish the
same trends at an fO2 of 10–10
, it is notable that Cu-rich melts
produced at these conditions contain more oxygen than melts
dominated by Fe, suggesting the presence of a Cu oxide
species. In order to assess whether an iron oxide species is im-
portant in controlling the wetting behavior of the melts pro-
duced in our experiments, we have plotted Θ as a function of
the mole percent oxygen in the sulfide liquid (Fig. 7A and B).
Dihedral angles from experiments containing 0 to 5 wt per-
cent Co, Cu, or Ni (Fig. 7A) decrease as the oxygen (and
hence Fe oxide) content of the melt increases; the data are in
accord with the correlation displayed by results from Gaetani
and Grove (1999) for Fe-bearing sulfide melts with minor
amounts of Ni (0.4–1.3 wt %). In contrast, experiments con-
taining high abundances of Ni, Cu, or Co are displaced from
this correlation, with runs at fO2 of 10–9
and 10–10
generally
plotting above the curve, and those at fO2 of 10–8
plotting
below the curve. Two effects are probably important in estab-
lishing these relations. First, the addition of a nonferrous
metal component may decrease the activity of Fe1 – xO, thus
leading to an increase in Θ for a specific melt oxygen content.
This effect is also expressed by the systematic increase in Θ
with an increased nonferrous metal component in the sulfide
melt seen in the data at fO2 of 10–9
and 10–10
(Fig. 5). Con-
versely, with increased fO2 (or decreased fS2) the proportion of
other nonferrous metal oxide species will increase, as equa-
tions 3, 4, and 5 shift to the left. If such oxide species are sur-
face active, then an increase in their abundance will corre-
spondingly decrease Θ. This effect not only explains the
general decrease in Θ with increased fO2, but also the de-
crease in Θ for the sample run at fS2 of 10–4
(D.ol 23, Table 3).
The fact that other metal oxide species may equal Fe1 – xO in
their capacity to reduce the olivine-melt surface energy may
account for the apparent invariance in Θ with a nonferrous
metal content displayed by the data at fO2 of 10–8
(Fig. 5).
In the context of our experimental results, it is useful to re-
call the observations of Ebel and Naldrett (1996) regarding
the enhanced wetting capacity of Cu-rich melts against silica
glass. As previously mentioned, these workers observed that
the wetting of Fe-Cu-S melts against the walls of the silica
glass tubes employed in their experiments increased when
high levels of Cu were present in the melt, implying an in-
crease in the glass-vapor surface energy or a decrease in the
glass-melt or melt-vapor surface energies. Ebel and Naldrett
(1996) went on to use this result to suggest that surface ten-
sion-induced spreading onto fracture surfaces in silicate host
rocks may provide a mechanism for separation of Cu-rich sul-
fide liquids from an Ni-rich sulfide cumulate. Our results in-
dicate that the addition of Cu to the sulfide melt either in-
creased Θ (at low fO2) or had no effect on Θ (high fO2), thus
152 ROSE AND BRENAN
0361-0128/98/000/000-00 $6.00 152
20
25
30
35
40
45
50
-5 0 5 10 15 20 25 30 35
Fe Only
Copper
Cobalt
Nickel
mol%Fe(sulfidemelt)
mol% O (sulfide melt)
A
fO
2
=10
-9
20
25
30
35
40
45
50
-5 0 5 10 15 20 25 30 35
Fe Only
Copper
Cobalt
Nickel
mol%Fe(sulfidemelt)
mol% O (sulfide melt)
B
fO
2
=10
-8
FIG. 6. Variation in mole percent oxygen as a function of mole percent
iron for sulfide liquids produced at (A) fO2 = 10–9
and (B) fO2 = 10–8
and con-
stant fS2 (10–2
). The errors associated with both iron and oxygen contents are
based on one standard deviation of n analyses. Note that the Ni-bearing sam-
ple in (B) with 36 mole percent iron (D.ol 7) has a somewhat lower oxygen
content than that of the main sample trend. This discrepancy is probably a
result of the somewhat lower fO2 (10–8.2
) for this experiment.

9. suggesting that the spreading of sulfide melts against silica
glass cannot be used to predict wetting of such melts against
mafic silicate minerals, such as olivine. However, our results
do not dispute the separation hypothesis put forth by Ebel
and Naldrett (1996), provided Cu-rich melts in contact with
natural fracture surfaces within crystalline rocks retain the
same relative interfacial energies as for silica glass.
Implications for sulfide melt mobility
The upper mantle environment: It is now well established
that sulfides may be an important minor phase in upper man-
tle rocks (e.g., Fleet and Stone, 1990), and it is also clear that
mantle-derived magmas may leave their source regions at or
near saturation in an immiscible sulfide liquid (e.g., Peach et
al., 1990). Given that sulfide liquids could play a significant
role in redistributing chalcophile and siderophile elements in
the upper mantle, it is worth assessing to what extent natural
sulfide liquids are likely to be wetting (i.e., Θ < 60°), and
hence, mobile via porous flow in this environment. The com-
bined results of this study and that of Gaetani and Grove
(1999) indicate that both fO2, and the identity and quantity of
the dissolved transition metal will play a role in determining
the mobility of natural sulfide liquids. Thus, information
bearing on the magnitude of all of the variables will be re-
quired in order to assess the potential for sulfide liquids to be
mobile in the upper mantle.
Estimates of the oxygen fugacity that prevails during
oceanic basalt genesis have been provided using Fe3+
/Fe2+
ratios in MORB glasses and olivine-orthopyroxene-spinel
barometry from peridotite xenoliths. Using the former tech-
nique, Christie et al. (1986) found a range in MORB fO2 from
1 to 3 log units more reducing than the nickel-nickel oxide
(NNO) buffer but with a predominance of values near (NNO
= –1.5 to –2.5. Studies of spinel peridotites by Wood et al.
(1990) yielded a similar fO2 range, if a small pressure correc-
tion was applied to the Christie et al (1986) values. Sulfur fu-
gacities estimated by Wallace and Carmichael (1992) from
Atlantic MORB glasses vary from 10–1
to 10–3
. Our results
suggest that sulfide liquids can be wetting in olivine-rich
lithologies at a fO2 as low as 10–9
, provided the nonferrous
metal content is less than ~15 wt percent. At 1,300°C, this
corresponds to ∆NNO of –2.3, indicating that Fe-rich sulfide
melts can be wetting at the conditions of fO2 and fS2 appro-
priate for the oceanic upper mantle, in accord with similar
conclusions reached by Gaetani and Grove (1999). It is im-
portant to note, however, that this conclusion is made with
the explicit assumption that the fO2 vs. melt oxygen content
determined at 1 atm is applicable to sulfide melts at high
pressure.
An additional, and potentially more direct, assessment of
sulfide liquid mobility in the upper mantle considers the vari-
ation in fO2 and nonferrous metal content exhibited by natural
sulfide liquids occurring in primitive mafic lavas. To assess
such data quantitatively, we have plotted the values of fO2 (in
terms of ∆NNO) and the nonferrous metal content of our ex-
periments (Fig. 8A) to determine the crossover point (i.e., Θ
< 60°) for sulfide melt wettability against olivine. We also
used the fO2 dependence of the partitioning of Fe and Ni be-
tween olivine and coexisting sulfide liquid, as calibrated by
Brenan and Caciagli (2000), to estimate fO2 for a variety of
olivine and sulfide-saturated, primitive oceanic volcanic
rocks. Those estimates, along with the total nickel + copper
contents of the associated sulfide liquid are portrayed in Fig-
ure 8B (other components, such as Co, are minor in these
samples). Along with these data we have superimposed the
dividing line between wetting and nonwetting liquids, as de-
termined from Figure 8A. With the exception of the most
nonferrous metal-rich compositions, measured fO2 of natural
samples are high enough that their elevated nonferrous metal
abundance does not hinder their wettability, thus further
strengthening our conclusions regarding the likelihood of sul-
fide melt mobility based on fO2 estimates alone. Moreover, it
is also important to note that samples with the highest non-
ferrous metal content (those from the Kilauea Iki lava lake)
plotting at or below the crossover line are copper rich (i.e., at.
% Cu/(Cu + Ni) >> 1), and thus unlikely candidates for prim-
itive sulfide liquids.
WETTING PROPERTIES OF Fe-Ni-CO-Cu-O-S MELTS AGAINST OLIVINE 153
0361-0128/98/000/000-00 $6.00 153
40
50
60
70
80
90
100
-5 0 5 10 15 20 25 30 35
12-31% Cu
12-26% Co
22-40% Ni
DihedralAngle(degrees)
mole % oxygen
22
14
16
15
2012
26
12
12
22
32
B
31
40
-9
-10
-8
40
50
60
70
80
90
100
-5 0 5 10 15 20 25 30 35
Fe
Fe (Gaetani and
Grove, 1999)
< 5% Cu
< 5% Co
< 5% Ni
DihedralAngle(degrees)
mole % oxygen
-9
-10
-8
A
FIG. 7. Variation of dihedral angle with mole fraction of oxygen within the
sulfide melt. A. Experiments involving Fe-bearing melts containing 0 to 5 wt
percent of a nonferrous metal component from this study, Brenan and Ca-
ciagli (2000), and Gaetani and Grove (1999). The dashed line is a best-fit
curve describing data for Fe-bearing compositions only. B. Data for experi-
ments containing substantial amounts of nonferrous metals. Numbers beside
each data point correspond to the abundance of that metal in the melt. The
dashed line corresponds to the trend for the Fe-rich melts portrayed in (A).

10. Formation of massive vs.
disseminated magmatic Fe-Ni-Cu-PGE sulfide deposits
Timing of sulfide saturation relative to solidification of a
high-temperature silicate cumulate matrix and sulfide liquid
wetting behavior will each have a direct bearing on the ten-
dency to form massive or disseminated sulfide ore deposits
associated with mafic magmas. When sulfide saturation oc-
curs prior to matrix solidification, droplets of sulfide liquid
can settle (provided the crystal + liquid matrix viscosity is low
enough) and be concentrated at a position of gravitational sta-
bility, thus forming a massive segregation. Whether the sul-
fide liquid remains massive or begins to disperse is deter-
mined by its wetting behavior with respect to the surrounding
solid matrix. If the crystallizing cumulate matrix is olivine rich
and conditions are relatively oxidizing, the melt may begin to
disperse by the process of surface tension-driven infiltration
(Watson, 1982). Conversely, if the surrounding environment
is relatively reducing, the sulfide melt will subsequently be
nonwetting, with no tendency to disperse, thus maintaining
the initially massive segregation.
If sulfide saturation occurred after a substantial amount of
the silicate matrix had crystallized, the resulting immiscible
sulfide melt will initially be dispersed between the silicate
minerals. Owing to the considerable density contrast between
molten sulfide and silicate minerals, however, buoyancy-dri-
ven compaction and melt segregation may occur, provided
the sulfide liquid is interconnected (i.e., if Θ < 60°). If non-
wetting conditions prevail, the sulfide liquid will remain dis-
persed upon complete solidification. Given that Fe-rich sul-
fide liquids can be interconnected under relatively oxidizing
conditions, we have used the compaction model of McKenzie
(1985) to assess specifically the time scales required for sig-
nificant segregation of sulfide-bearing layers having a given
thickness and porosity. It is important to note at the outset
that the viability of these calculations for modeling natural
segregation processes depends on the following two impor-
tant assumptions: that the temperature dependence of Θ is
relatively small and thus values measured at 1,300°C can be
used to predict melt connectivity at lower temperatures, and
that values of Θ measured for olivine provide an accurate pre-
diction of the permeability of olivine-dominated rocks (i.e.,
the presence of small amounts of potentially nonwetting
phases does not seriously hinder melt flow).
In McKenzie’s model, the compacting layer with a constant
initial porosity (φ) rests below an impermeable roof (perhaps
defined by a nonwetting lithology). As a consequence of the
negative density contrast between sulfide melt and olivine-
rich solid, the melt will migrate downward accompanied by
compaction of the solid to accommodate the loss of melt. The
thickness of the zone over which compaction is most rapid is
defined as the compaction length scale (δc) and is given by the
relation:
δc = (ηk/µ)0.5
, (6)
where η is the kinematic viscosity of the matrix (~1018
Pa⋅s;
McKenzie, 1985), k is the permeability of the matrix (m2
), and
µ is the kinematic viscosity of the sulfide melt. McKenzie’s
original relation between permeability, porosity, and grain
size (a; assumed to be 1 mm) takes the form:
k = a2
φ3
/1,000. (7)
As shown by von Bargen and Waff (1986), however, perme-
abilities calculated by this equation are about 100× lower
than values predicted using a fluid distribution model dic-
tated by surface energy minimization. However, forms of the
k vs. φ functions are nearly identical, and as such, permeabil-
ities calculated by equation 7 are simply multiplied by 100 to
154 ROSE AND BRENAN
0361-0128/98/000/000-00 $6.00 154
-7
-6
-5
-4
-3
-2
-1
0
1
0 10 20 30 40 50 60
wetting
non-wetting
∆NNO
wt% Ni + Cu in sulfide melt
A
?
-7
-6
-5
-4
-3
-2
-1
0
1
0 10 20 30 40 50 60
Kilauea pumice (primitive)
Kilauea pumice (evolved)
Kilauea lava lake (primitive)
Kilauea lava lake (evolved)
FAMOUS
Nazca
Lamont
Disko
NNO∆
wt% Ni + Cu in sulfide liquid
non-wetting
wetting
B
Boundary
from Plot A
?
FIG. 8. A. Wetting vs. nonwetting samples produced in this study, in terms
of fO2 relative to the NNO buffer (∆NNO = log fO2 sample – log fO2 NNO),
and the abundance of Ni + Cu within the sulfide melt. Above the dashed
curve, sulfide liquids are wetting (Θ < 60°), whereas below the curve sulfide
liquids are nonwetting (Θ > 60°). B. Summary of oxygen fugacities recorded
by sulfide-saturated volcanic rocks as a function of the Ni + Cu content of the
associated sulfide liquid. Oxygen fugacities were calculated based on Fe-Ni
partitioning between coexisting olivine and sulfide liquid, as calibrated by
Brenan and Caciagli (2000). Values of fO2 are expressed with respect to the
NNO buffer calculated at the quenching temperature of the lava (as esti-
mated from phase equilibria or olivine thermometry). In general, Ni is the
dominant nonferrous transition metal in these sulfide liquids, and samples
with >10 wt percent Cu are labeled in terms of their melt Cu content. Ki-
lauea samples were produced during the 1959 eruption and correspond to
subaerial tephra (pumice) and samples from the lava lake; FAMOUS (vol-
canic), Nazca, and Lamont are ocean floor basalts; Disko is a native iron-
bearing Tertiary mafic subvolcanic dike from Disko Island, west Greenland.
Data sources are as follows: Kilauea, FAMOUS, Nazca (references in Fleet
and Stone, 1990); Lamont (J. Allen, pers. commun.; Allen et al., 1989); Disko
(Pederson, 1979).

11. bring them in line with values predicted by the von Bargen
and Waff (1986) model.
The compaction time scale (τh) is defined as the time re-
quired to change the porosity of the entirely molten layer (of
thickness h) from φ to φ/e and is given by the relation:
τh = τo(h/δc + δc/h), (8)
where τo is the initial compaction rate at the base of the layer
defined as:
τo = δc/ωo(1 – φ), (9)
and where ωo is the separation velocity defined as:
ωo = k(1 – φ)∆ρg/µφ, (10)
in which ∆ρ is the density difference between the melt and
the solid matrix and g is the acceleration due to gravity.
For comparative purposes, values of the compaction length
and time scales were determined for both sulfide and mafic
silicate melts. Viscosity and density of the olivine-bearing ma-
trix and silicate melt are taken from McKenzie (1985). Den-
sities and viscosities of compositionally complex sulfide liq-
uids are unknown, so we have chosen to model the sulfide
liquid using values obtained for molten FeS extrapolated to
1,300°C. The sulfide melt density was calculated to be 3,705
kg/m3
based on the data of Kucharski et al. (1984); using kine-
matic viscosity values from Vostryakov et al. (1964), we have
calculated a sulfide melt viscosity of 2.7 × 10–3
Pa⋅s. Results
of our length- and time-scale calculations are shown in Figure
9.
Figure 9A depicts the compaction length scale for both sul-
fide and silicate melt over a range of porosities and shows that
δc is ~10× thicker for molten sulfide than for molten silicate.
Therefore, efficient compaction is expected to occur over a
significantly greater thickness of sulfide melt-bearing layers
as opposed to equivalent layers containing silicate melt. Fig-
ure 9B displays the compaction time scales for both sulfide
and silicate melt-bearing layers having variable thickness and
porosity. For all layer thickness and low porosity (i.e., 0.1 and
0.01%), sulfide melt is extracted at a much greater rate than
silicate melt. For example, a sulfide melt-bearing layer 1 km
thick with a porosity of 0.1 percent requires ~180 yr for ap-
proximately one-third of the melt to be expelled, whereas an
equivalent silicate melt-bearing layer requires ~77 Ka to
achieve the same result. For all layer thickness and high
porosity (i.e., 1 and 10%), both sulfide and silicate melt are
extracted at the same rate. The time scale for solidification of
a mafic-ultramafic intrusion will depend on its size, depth of
emplacement, and availability of circulating fluids, among
other factors. However, for kilometer-sized mafic bodies in-
truding at shallow depths, as exemplified by the Skaergaard
and Stillwater complexes, cooling time scales of 150 to 200 Ka
are probably reasonable (e.g., Hess, 1972; Norton and Taylor,
1979). Our calculations would therefore suggest that sulfide
liquid segregation by compaction is possible within the solid-
ification time interval, even for layers that are hundreds of
meters thick with low porosities.
Given the sensitivity of olivine-sulfide melt dihedral angles
to the melt oxygen content, variations in magma fO2 during so-
lidification may play an important role in controlling the final
distribution of sulfide melt. For example, it is clear that some
mafic intrusions have undergone assimilation of a sulfide-
bearing country rock, which may have resulted in sulfide sat-
uration, but also displaced magma fO2 to more reduced values
(e.g., Voisey’s Bay; Brenan and Li, 2000). If assimilation and
sulfide saturation occur while the system is still largely
molten, gravity separation of massive sulfide can be efficient.
However, if saturation occurs after a significant amount of sil-
icate crystallization, the sulfide will get stuck (i.e., remain dis-
seminated) if the magma fO2 has been lowered by assimilation
of a reducing agent. In contrast, relatively oxidized magmas
may be able to segregate their interstitial sulfide as a result of
WETTING PROPERTIES OF Fe-Ni-CO-Cu-O-S MELTS AGAINST OLIVINE 155
0361-0128/98/000/000-00 $6.00 155
1
2
3
4
5
6
7
-4 -3 -2 -1
Sulfide Melt
Basaltic Melt
Compaction Time-scale
log
10h
(yrs)τ
log10
φ
B
Stillwater,
Skaergaard
solidified
100 m
10 m
1k m
-2
-1
0
1
2
3
4
5
Compaction Length-scale
Sulfide Melt
Basaltic Melt
log
10c
(m)δ
A
FIG. 9. Variation of (A) compaction length scale (δc) and (B) compaction
time scale (τh) as a function of porosity for systems containing molten silicate
and sulfide. Note that in each case, an interconnected porosity is assumed
(i.e., Θ < 60°), and values of δc and τh were calculated according to equations
6 and 8 in the text, respectively. Curves in A show that, for a given porosity,
the length scale over which compaction is most rapid is ~10× larger for sys-
tems containing molten sulfide than for those containing silicate melt. Curves
in B are labeled according to the thickness of the compacting layer and show
that, for low porosities, the compaction time scale for segregating a sulfide
liquid is nearly 1,000× faster than that for segregating silicate melt. However,
at higher porosities, both sulfide and silicate melts are extracted at the same
rates for similar layer thicknesses. The time scale for cooling a moderate-
sized basaltic intrusion (e.g., Stillwater or Skaergaard) to below the silicate
solidus is ~160 Ka, which is ample time for sulfide liquid segregation, even
for relatively thick layers (i.e., 5 km) having low porosities.

12. compaction. This analysis clearly affirms the notion that signs
of early sulfide saturation (i.e., Ni depletion in olivine; Nal-
drett, 1989) are critical to assessing the massive sulfide po-
tential of a particular igneous body, especially if significant
wall-rock assimilation is suspected to have occurred during
emplacement.
Summary and Conclusions
Our results indicate that both oxygen fugacity and the iden-
tity and abundance of nonferrous transition metals play im-
portant roles in the wetting behavior of sulfide melt against
forsteritic olivine. The presence of nonferrous metal species
within sulfide melts at low fO2 values counters the effect of sur-
face-active Fe oxide on sulfide melt wettability. At higher val-
ues of fO2 however, Fe oxide couples with other nonferrous
metal oxides to enhance sulfide melt wetting properties. Given
the relatively oxidized conditions inferred for the petrogenesis
of oceanic basalts, sulfide liquids associated with these mag-
mas are likely to contain a high abundance of metal oxide
species, and thus be wetting. Such wetting liquids may have
a profound effect on the distribution of chalcophile and
siderophile elements in the mantle source regions for oceanic
basalts, particularly in terms of disturbing the parent-daughter
ratios of the U/Pb, Th/Pb, and Re/Os isotopic systems (e.g.,
Gaetani and Grove, 1999). During the solidification of mafic-
ultramafic intrusive bodies, the final distribution of sulfides
will depend on both the timing of sulfide liquid saturation and
the sulfide liquid wetting properties. If sulfide liquid satura-
tion occurs while the silicate portion of the system is largely
molten, droplets of sulfide liquid can fall through the largely
liquid matrix and be concentrated at a position of gravitational
stability, thus forming a massive segregation. If conditions are
relatively reducing, then subsequent dispersal of sulfide liquid
into the solid silicate pile is avoided since surface-energy-
induced infiltration (Watson, 1982) is inhibited when Θ < 60°.
Conversely, if sulfide liquid saturation occurs after a substan-
tial amount of silicate matrix has formed, buoyancy-driven
compaction and melt segregation can take place if conditions
are relatively oxidized, and hence Θ < 60°. We have found that
efficient compaction segregation can occur on time scales that
are short relative to the cooling interval of a moderate-sized
mafic intrusion, suggesting that this process could be an effec-
tive means of sulfide segregation in such systems.
Acknowledgments
L. A. R. thanks Richard Bailey for helping to clarify some
important mathematical concepts as well as Claudio Cer-
mignani (University of Toronto) and Yves Thibault (Univer-
sity of Western Ontario) for their help with EMP analyses.
We thank both C. Ballhaus and an anonymous referee for
their constructive reviews. Research was supported by
NSERC operating grant OGP 0194228 to J. M. B., and L. A.
R. is grateful for support from a University of Toronto Open
Scholarship.
March 7, July 25, 2000
REFERENCES
Allen, J.F., Batiza, R., Perfit, M.R., Fornari, D.J., and Sack, R.O., 1989,
Petrology of lavas from the Lamont seamount chain and adjacent East Pa-
cific Rise: Journal of Petrology, v. 30, p. 1245–1298.
Ballhaus, C., and Ellis, D.J., 1996, Mobility of core melts during earth’s ac-
cretion: Earth and Planetary Science Letters, v. 143, p. 137–145.
von Bargen, N., and Waff, H.S., 1986, Permeabilities, interfacial areas and
curvatures of partially molten systems: Results of numerical computations
of equilibrium microstructures: Journal of Geophysical Research, v. 91, no.
B9, p. 9261–9276.
Barnes, S.-J., and Picard, C.P., 1993, The behaviour of platinum-group ele-
ments during partial melting, crystal fractionation, and sulphide segrega-
tion: An example from the Cape Smith fold belt, northern Quebec:
Geochimica et Cosmochimica Acta, v. 57, p. 79–87.
Barton, P.B., Jr., and Toulmin, P., III, 1964, The electrum-tarnish method for
the determination of the fugacity of sulfur in laboratory sulfide systems:
Geochimica et Cosmochimica Acta, v. 28, p. 619–640.
Brenan, J.M., and Caciagli, N.C., 2000, Fe-Ni exchange between olivine and
sulfide liquid: Implications for oxygen barometry in sulfide-saturated mag-
mas: Geochimica et Cosmochimica Acta, v. 64, p. 307–320.
Brenan, J.M., and Li, C., 2000, Constraints on oxygen fugacity during sulfide
segregation in the Voisey’s Bay intrusion, Labrador, Canada: ECONOMIC
GEOLOGY, v. 95, p. 901–915.
Christie, D.M., Carmichael, I.S.E., and Langmuir, C.H., 1986, Oxidation
state of mid-ocean ridge basalt glasses: Earth and Planetary Science Let-
ters, v. 79, p. 397–411.
Czamanske, G.K., and 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, v. 88, p. 587–599.
Ebel, D.S., and Naldrett, A.J., 1996, Fractional crystallization of sulfide ore
liquids at high temperature: ECONOMIC GEOLOGY, v. 91, p. 607–621.
Fleet, M.E., and Stone, W.E., 1990, Nickeliferous sulfides in xenoliths,
olivine megacrysts and basaltic glass: Contributions to Mineralogy and
Petrology, v. 105, p. 629–636.
Gaetani, G.A., and Grove, T.L., 1997, Partitioning of moderately siderophile
elements among olivine, silicate melt, and sulfide melt: Constraints on core
formation in the earth and Mars: Geochimica et Cosmochimica Acta, v. 61,
p. 1829–1846.
——1999, Wetting of mantle olivine by sulfide melt: Implications for Re/Os
ratios in mantle peridotite and late-stage core formation: Earth and Plane-
tary Science Letters, v. 169, p. 147–163.
Harker, D., and Parker, E.R., 1945, Grain shape and grain growth: American
Society of Metallogeny Transactions, v. 34, p. 156–201.
Hess, G.B., 1972, Heat and mass transport during crystallization of the Still-
water igneous complex: Geological Society of America Memoirs, v. 132, p.
503–520.
Kress, V., 1997, Thermochemistry of sulfide liquids. I. The system O-S-Fe at
1 bar: Contributions to Mineralogy and Petrology, v. 127, p. 176–186.
Kucharski, M., Ip, S.W., and Toguri, J.M., 1994, The surface tension and den-
sity of Cu2S, FeS, Ni3S2 and their mixtures: Canadian Metallurgical Quar-
terly, v. 33, no. 3, p. 197–203.
Mathez, E.A., and Yeats, R.S., 1976, Magmatic sulfides in basalt glass from
DSDP hole 319A and site 320, Nazca plate: Deep Sea Drilling Project Ini-
tial Reports, v. 34, Dec. 1973–Feb. 1974, p. 363–373.
McKenzie, D., 1985, The extraction of magma from the crust and mantle:
Earth and Planetary Science Letters, v. 74, p. 81–91.
Minarik, W.G., Ryerson, F.J., and Watson, E.B., 1996, Textural entrapment
of core-forming melts: Science, v. 272, p. 530–533.
Naldrett, A.J., 1989, Magmatic sulfide deposits: New York, Clarendon Press,
186 p.
Norton, D., and Taylor, H.P., Jr., 1979, Quantitative simulation of the hy-
drothermal systems of crystallizing magmas on the basis of transport theory
and oxygen isotope data: An analysis of the Skaergaard intrusion: Journal of
Petrology, v. 20, pt. 3, p. 421–486.
Peach, C.L., and Mathez, E.A., 1993, Sulfide melt-silicate melt distribution
coefficients for nickel and iron and implications for the distribution of other
chalcophile elements: Geochimica et Cosmochimica Acta, v. 57, p.
3013–3021.
Peach, C.L., Mathez, E.A., and Keays, R.R., 1990, Sulfide melt-silicate melt
distribution coefficients for noble metals and other chalcophile elements as
deduced from MORB: Implications for partial melting: Geochimica et Cos-
mochimica Acta, v. 54, p. 3379–3389.
Pedersen, A.K., 1979, Basaltic glass with high temperature equilibrated im-
miscible sulfide bodies with native iron from Disko, Central West Green-
land: Contributions to Mineralogy and Petrology, v. 69, p. 397–407.
Riegger, O.K., and Van Vlack, L.H., 1960, Dihedral angle measurement:
AIME Transactions, v. 218, p. 933–935.
156 ROSE AND BRENAN
0361-0128/98/000/000-00 $6.00 156

13. Roy-Barman, M., Wasserburg, G.J., Papanastassion, D.A., and Chaussidon,
M., 1998, Osmium isotopic compositions and Re-Os concentrations in sul-
fide globules from basaltic glasses: Earth and Planetary Science Letters, v.
154, p. 331–347.
Stone, W.E., and Fleet, M.E., 1991, Nickel-copper sulfides from the 1959
eruption of Kilauea volcano, Hawaii: Contrasting compositions and phase
relations in eruption pumice and Kilauea Iki lava lake: American Mineral-
ogist, v. 76, p. 1363–1372.
Stone, W.E., Crockett, J.H., and Fleet, M.E., 1990, Partitioning of palladium,
iridium, platinum and gold between sulfide liquid and basalt melt at
1200°C: Geochimica et Cosmochimica Acta, v. 54, p. 2341–2344.
Vostryakov, A.A., Vatolin, N.A., and Yesin, O.A., 1964, Viscosity and electri-
cal resistivity of molten alloys of iron with phosphorous and sulfur: Physics
of Metals and Metallography, v. 18, no. 3, p. 167–169.
Waff, H.S., and Faul, U.H., 1992, Effects of crystalline anisotropy on fluid
distribution in ultramafic partial melts: Journal of Geophysical Research, v.
97, no. B6, p. 9003–9014.
Wallace, P., and Carmichael, I.S.E., 1992, Sulfur in basaltic magmas:
Geochimica et Cosmochimica Acta, v. 56, p. 1863–1874.
Watson, E.B., 1982, Melt infiltration and magma evolution: Geology, v. 10, p.
236–240.
Watson, E.B., and Lupulescu, A., 1993, Aqueous fluid connectivity and
chemical transportin clinopyroxene-rich rocks: Earth and Planetary Sci-
ence Letters, v. 117, p. 279–294.
Wood, B.J., Bryndzia, L.T., and Johnson, K.E., 1990, Mantle oxidation state
and its relationship to tectonic environment and fluid speciation: Science,
v. 248, p. 337–345.
WETTING PROPERTIES OF Fe-Ni-CO-Cu-O-S MELTS AGAINST OLIVINE 157
0361-0128/98/000/000-00 $6.00 157

14. 0361-0128/98/000/000-00 $6.00 158