American Mineralogist, Volume 86, pages 578–582, 2001

Radiographic study on the viscosity of the Fe-FeS melts at the pressure of 5 to 7 GPa

SATORU URAKAWA,1,* HIDENORI TERASAKI,2 KENICHI FUNAKOSHI,3 TAKUMI KATO,2

AND AKIO SUZUKI4

1
Department of Earth Sciences, Okayama University, Okayama 700-8530, Japan
2
Geoscience Institute, University of Tsukuba, Tsukuba 305-8571, Japan
3
Japan Synchrotron Radiation Research Institute, Mihara 679-5198, Japan
4
Institute of Mineralogy, Petrology and Economic Geology, Tohoku University, Sendai 980-8578, Japan

ABSTRACT
Stokes’ viscometry combined with in situ X-ray radiographic observation, using the 6–8 type
multi-anvil press and synchrotron radiation, has been applied to the viscosity measurement of the Fe-
FeS melt up to pressures of 7 GPa. The viscosity is found to be about 2 × 10–2 Pa-s at 5 to 7 GPa and
temperatures about 1350 K, in marked contrast to previous viscosity measurements, which showed
high viscosity, 0.5 to 14 Pa-s, at 2 to 5 GPa (LeBlanc and Secco 1996). Our viscosity data, however,
is consistent with all other evidence, which include 1 atm viscosity data, X-ray structure analysis, and
ab initio simulations. Recent viscosity measurements (Dobson et al. 2000) also showed the viscosity
of Fe-FeS melt to be about 10–2 Pa-s at 2.5 GPa. Thus, we are confident that the viscosity of the Fe-
FeS melt is close to a typical value (10–2 Pa-s) of viscosity for liquid metal even at high pressures.

INTRODUCTION
The viscosity of the Earth’s outer core is of fundamental
importance to understanding the evolution, present state, and
dynamic process of the Earth’s core. Seismic, geodetic, and
geomagnetic observations and theory, however, give a wide
range in estimated viscosity of the Earth’s outer core from 10–2
to 1012 Pa-s (Secco 1995). Another approach to the viscosity of
the core is to investigate the viscosity of molten iron alloys at
pressures (P) and temperatures (T) conditions considered ap-
propriate for the core. Poirier (1988) estimated the viscosity of
iron melt in the core is 6 × 10–3 Pa-s by the semi-empirical
relationship between activation energy for viscosity of liquid
metal and melting temperature. Alfè and Gillan (1998a, 1998b)
also showed the viscosity of Fe-S melt is about 10–2 Pa-s at
core P and T by a first principles calculation. Thus, theoretical
approaches point to the viscosity of molten iron alloy at core
conditions being similar to that at ambient condition and close
to lower bound of geophysical estimations. Experimental stud-
ies have, however, yielded contradictory results, although these
are limited to several GPa.
LeBlanc and Secco (1996) and Secco et al. (1998) found
the viscosity of the Fe-FeS liquid up to 5 GPa, by Stokes’ vis-
cometry, to be extremely high from 0.5 to 14 Pa-s, which is 2
to 3 order of magnitudes higher than pure iron at 1 atm. They
derived an activation energy of 240–400 kJ/mol and an activa-
tion volume of 1.5–3.6 × 10–6 m3/mol for viscosity, which are
very high values, similar to those of the solid. Secco et al. (1998)
* E-mail: afeg0320@cc.okayama-u.ac.jp
0003-004X/01/0004–578$05.00
predicted the viscosity of Earth’s outer core to range from 10–
4
to 104 Pa-s, based on these data. Brazhkin (1998) also re-
ported a very large activation volume of viscosity for pure iron
liquid, suggesting a glassy state of the inner core.
The aim of this study is to determine the viscosity of mol-
ten iron alloys at high pressure by Stokes’ viscometry using
synchrotron X-ray radiographic observations. This is a reliable
method for investigating the liquid viscosity under pressure
and potentially can measure viscosity as low as 10–2 Pa-s
(Kanzaki et al. 1987; Dobson et al. 1996; Dingwell 1998).
EXPERIMENTS
We carried out the high-pressure X-ray radiography experi-
ments by using synchrotron radiation at BL04B1 of SPring-8,
Japan. The high-pressure X-ray system SPEED-1500 was used,
which is composed a 6–8 double-stage high-pressure appara-
tus driven by a 1500 ton uniaxial press, and an X-ray
diffractometer (Utsumi et al. 1998). The sample cell assembly
is illustrated in Figure 1. The starting material was sub-mi-
crometer fine powder of Fe (99.99% pure) and synthetic FeS
(99.9% pure) mixed to a composition of Fe61S39, which is about
3 at% rich in sulfur relative to the eutectic composition of the
system Fe-FeS at 5 to 7 GPa reported by Usselman (1975).
The Fe61S39 sample was enclosed in the BN sample capsule,
which had been deoxidized under N2 atmosphere at 2273 K.
Temperature was measured just above the BN sample capsule
using a W3%Re–W25%Re thermocouple, with no pressure
correction. Pressure was determined from the equation of state
of BN (unpublished data) and MgO (Jamieson et al. 1982),
which were used as sample capsule and X-ray window, respec-
578
 URAKAWA ET AL.: VISCOSITY OF THE Fe-FeS MELTS AT P 579

the bottom of the sample capsule. These experiments are re-

producible as will be reported in the PhD thesis of H. Terasaki.
After radiographic observation, we acquired an X-ray dif-
fraction pattern of the pressure marker at the sample tempera-
ture at which the viscosity was measured. Samples were then
quenched to room temperature. The recovered samples were
examined by optical microscopy and a JEOL JXA-8621 elec-
tron microprobe.

RESULTS
The Height-temperature-time profile (Fig. 2) shows the fi-
nal stage of radiographic observations at 5.0 GPa, in which the
sphere dropped to the bottom of sample capsule. Textural ob-
servations of the recovered sample from a separate run, which
was quenched just after sphere settling down, show the sample
was completely melted at this stage. Velocities of the falling
spheres are estimated from height-time profile and range from
2 to 3 mm/s. The terminal velocity, transient period, and fall-
ing distance, which spheres accelerate to steady velocity are
calculated as follows.
Simple Stokes flow can be expressed by the equation of
motion:

FIGURE 1. The cross section of high-pressure cell assembly for

m
∂v
∂t
( )( )
= ρ f − ρs g 4 3 πr 3 − 6 πηrv (1)
viscosity measurements. (1) MgO pressure medium, (2) ZrO2 thermal
insulator, (3) MgO X-ray window, (4) graphite Heater, (5) Mo where m is mass of sphere, v is ascent or descent velocity of
electrode, (6) BN sample capsule, (7) W3%Re-W25%Re sphere, ρf is density of fluid, ρs is density of sphere, g is accel-
thermocouple, (8) Fe-FeS sample, (9) Au or Pt sphere. eration of gravity, r is radius of sphere, and η is viscosity of
fluid. The first term of right-hand side is a buoyancy force and
second term is a drag force. The solution of this equation is as
follows

tively. Combination of a straight graphite heater, the ZrO2 ther-

mal insulator, and the MgO cylindrical X-ray window with a
high thermal conductivity reduces the temperature gradient
inside the sample space. X-ray diffraction observation of the
melting of Fe-FeS system in additional experiments showed
that temperature difference between the hot junction of the ther-
mocouple and the sample is less than 30 K.
Stokes’ viscometry experiments were designed based on a
falling sphere method. Use of gold or platinum as the sphere,
gives enough contrast to distinguish the sphere from the Fe-
FeS melt in X-ray radiography. Spheres were made by fusing
the thin wire in flash melting. The diameter of sphere was about
100 µm with a tolerance of 2 µm. We used a few spheres si-
multaneously to decrease statistical error. The falling velocity
of the sphere was calculated from a time-elapsed series of ra-
diograph images recorded by an X-ray CCD camera.
Viscosities were measured above the solidus of the system
Fe-FeS (Usselman 1975) at 5.0 and 6.9 GPa. After pressure
was applied, temperature was increased to about 1073 K and
kept there for several minutes. This temperature is lower than
the ternary eutectic temperature of the system Fe-FeS-FeO
(Urakawa et al. 1987). Next, temperature was raised at a rate
of about 200 K/min while the radiographic images were ac- FIGURE 2. Height-temperature-time profile for the experiment at
quired. The spheres first began moving just above the Fe-FeS- 5.0 GPa. Filled circles show the position of marker sphere. Height is
FeO eutectic temperature and then slowly sank with some relative to the base of sample container. This plot gives the terminal
horizontal motion. Finally, the spheres dropped vertically to velocity of falling sphere through the molten sample.
580 URAKAWA ET AL.: VISCOSITY OF THE Fe-FeS MELTS AT P


v = v0 −
(
2 ρ f − ρs gr 2 ) ( )
 9 η  2 ρ f − ρs gr
 exp − t +
2

(2)
2 
 9η   2ρ, r  9η

where v0 is an initial velocity of sphere at t = 0. Independent of

initial velocity v0 and the transient process, the sphere reaches
a terminal velocity, vT, in Stokes flow of

vT =
(
2 ρ f − ρs gr 2 ) (3)
9η
Substituting Equation 3 into 2, gives a simpler form for the
velocity of the sphere

v = (v0 − vT ) exp −
(
 g ρ f − ρs  )
t  + vT (4)
 vT ρs 

Distance of ascent or decent D of sphere is calculating by

integration
T
D = ∫ vdt FIGURE 4. Time and distance to reach terminal velocity (tT and DT)
0
(5) for sphere in Stokes flow. Shaded area shows measurable terminal
(v − v )v ρ   g(ρ f − ρs ) T  − 1 + v T
= − 0 T T s  exp − velocity in Stokes viscometry using SR radiograph.
  T
g(ρ f − ρs )   vT ρs  

Assuming ρf = 6 Mg/m3, ρs = 20 Mg/m3, r = 50 × 10–6 m,

and η = 0.02 Pa-s, we find from Equation 3, that the terminal
velocity vT = – 3.81 × 10–3 m/s. Relationships among velocity,
distance, and time can be calculated from Equation (4) and (5),
assuming v0 = 0 m/s (Fig. 3). These calculations show that time
and distance to reach terminal velocity, tT and DT, are 9 ms and
36 micrometers, respectively. These values are short enough
for our experiments to measure the terminal velocity for Fe-
FeS melt.
From Equation 4 and 5, the time and distance to reach ter-
minal velocity, tT and DT, depend simply on terminal velocity
vT (Fig. 4). Configuration of our high-pressure cell allows us to
measure the viscosity up to vT = 30 × 10–3 m/s. Although this
simple analysis does not include wall or end corrections for
the viscosity, those effects are expected to be minor. Other ex-
perimental conditions, such as time resolution of camera sys-
tem and kinetics of melting, might slightly change these
estimations, but clearly these radiographic measurements give
terminal velocity. In the above analysis, the densities of spheres
were obtained from equations of state of gold (Anderson et al.
1989) and platinum (Jamieson et al. 1982). The density of FeS
rich portion of Fe-FeS melt at 1 atm was reported by Nagamori
(1969), and the isothermal bulk modulus for Fe-FeS melt was
determined by Sanloup et al. (2000). The density of Fe61S39
melt at experimental conditions is calculated based on Birch-
Murnaghan equation of state.
Finally, the viscosities of Fe-FeS melts are calculated using
Stokes’ equation with wall and end correction (Faxen 1925),

2 grs2 (ρs − ρm )  r  
3 5
r  r 
η= 1 − 2.104 s  + 2.09 s  − 0.95 s  
9vT (1 + 3.3(rs h))   rc   rc   rc  

where g is acceleration of gravity, rs is radius of sphere, ρs is

FIGURE 3. Transient period that sphere accelerates to its final steady
or terminal velocity. Solid and dashed lines show distance and time,
respectively.
density of sphere, ρm is density of melt, , and vT is terminal
velocity, rc is radius of sample, and h is fall distance. The val-
ues of viscosity of Fe-FeS melt are 1.6 × 10–2 Pa-s at 5.0 GPa
and 1373k and 2.4 × 10–2 Pa-s at 6.9 GPa and 1333 K.
We observed three falling spheres in the radiography ex-
periment at 5.0 GPa, which allow us to evaluate the statistical
error of the viscosity as ±25%. Reaction of the gold and plati-
num spheres with molten iron-sulfur alloy during the experi-
ments decreased their diameters by up to ~10%. This contributes
~15% error to the viscosity. The sum of the experimental un-
certainties in the measured viscosities is ±40%.
The recovered samples showed typical quench textures and
no sign of coexisting solid phases. Electron microprobe analy-
sis of the recovered sample yielded the atomic ratio of Fe to S
of 61.0 to 39.0, but about 3 at% of oxygen was also detected.
 URAKAWA ET AL.: VISCOSITY OF THE Fe-FeS MELTS AT P 581

Oxygen contamination may have arisen from an oxidized sur-

face layer on the starting iron powder. Qualitative analysis did
not detect any other elements in the recovered sample. How-
ever, gold or platinum spheres were completely dissolved into
Fe-S liquids during the acquisition of X-ray diffraction profile
after radiographic observation, so that the recovered samples
must also contain about 0.03 at% of gold or platinum.

DISCUSSION
The present radiographic study revealed that the viscosity
of the molten Fe-FeS alloy is about 2 × 10–2 Pa-s at 5 to 7 GPa,
which is about three or four times larger than those of typical
molten iron alloys at atmospheric pressure. The values of vis-
cosity are shown on an Arrhenius plot in Figure 5 and are com-
pared with viscosities of the Fe61S39 melt at 2 to 5 GPa previously
determined by LeBlanc and Secco (1996). Their values for the
viscosity of the Fe-FeS melt are by three orders of magnitude
higher than our data at 5 GPa and 1373 K. LeBlanc and Secco
(1996) applied an electro-detection method to in situ Stokes’
viscometry. Their experimental configuration is similar to ours,
except they used ruby spheres rising in the melt. The large dis- FIGURE 5. Arrhenius plot of the viscosities of the molten Fe-S
alloys. Filled circles show viscosities for Fe-FeS melts determined by
crepancy between the two studies, prompts us to consider pos-
this study at 5.0 GPa and 1373 K, and 6.9 GPa and 1333 K. Filled
sible experimental effects, as follows. squares, diamonds, triangles, and inverted triangles are Fe-FeS melts
If convection occurs in the liquid sample, it will affect the at 2, 3, 3.9, and 5 GPa, respectively, measured by LeBlanc and Secco
falling velocity of sphere. Using the temperature gradient of (1996). Open circles are Fe melts at 1 atm (Nakanishi et al. 1967), and
30 K, thermal expansion of 1.2 × 10–4 /K (Nagamori 1969), open squares are FeS melts at 1 atm (Barmin et al. 1970). The large
and thermal diffusivity of 6.0 × 10–6 m2/s–1 (Touloukian 1970), error bar at right-hand side denotes one estimate for the effect of
the Rayleigh number was evaluated to be 6 for our cell, which alloying with O, Au, and Pt on viscosity (see text).
is much lower than the critical Rayliegh number of 1200 for
the onset of convection. Hence, convection is not a factor. Ad-
1988).
dition of other component should change the viscosity. In the
Recently Vočadlo et al. (2000) evaluated the viscosity of
present study, the Fe-FeS melts contain O, Au, and Pt. Effect
Fe-FeS melt at 5 GPa by first principles calculations, giving
of addition of these components on viscosity is unknown at
1.1 × 10–2 Pa-s at 1300 K and 0.4 × 10–2 Pa-s at 1500 K. These
high pressure. It is known, however, that the effect of such di-
calculations agree with our data within factor of two. The struc-
lute liquid alloys on the viscosity at atmospheric pressure gen-
tural study of Urakawa et al. (in review), which shows the
erally does not exceed an order of magnitude (e.g., Iida and
simple structure for the Fe-FeS melt, also supports the low vis-
Guthrie 1988; Secco 1995). Even if an order of magnitude ef-
cosity. Furthermore, Dobson et al. (2000) independently mea-
fect due to alloying is allowed for, a large difference still re-
sured the viscosity of molten Fe-FeS alloys using an X-ray
mains between LeBlanc and Secco’s and our results (Fig. 5).
radiographic method, and showed the viscosity to be about 10–2
LeBlanc and Secco (1996) and Secco et al. (1998) attrib-
Pa-s at 2.5 GPa and 1450 K. Thus the more recent data and
uted their anomalous high viscosity of Fe-FeS melt to its struc-
calculations are consistent. We are confident that the viscosity
ture. They suggested the possibility that sulfur conditions large
of the Fe-FeS melt is as low as 10–2 Pa-s at moderately high
viscous flow unit resulting from its polymerization tendency,
pressures.
giving high viscosity to Fe-S melts similar to molecular liquid
sulfur. Structure analysis using in situ X-ray diffraction, how- [
An Arrhenius relation, η = η0 exp (Qac + PVac ) RT is used ]
to extrapolate our viscosity data to core pressure. We assume
ever, has not found any evidence of sulfur polymerization in
the viscosity of Fe-FeS melt is independent of composition,
Fe-FeS melt at 3 GPa (Urakawa et al. 1998, 2001). Recent first
and we adopt the pure Fe melt viscosity as 1 atm reference
principles simulations (Alfè and Gillan 1998a; Vočadlo et al.
data. The activation energy of 40 kJ/mol is determined mainly
2000) also indicate that the sulfur polymerization is unlikely.
from 1 atm Fe data. The activation volume is 0.7 × 10–6 m3/mol
In contrast, our viscosity data for Fe-FeS melts are consis-
for viscosity of Fe-FeS melt by non-linear least square fitting.
tent with the viscosity of iron alloys at 1 atm. The values of
These values give a viscosity of 2 × 10–1 Pa-s at the inner core
viscosity of pure Fe melt (Nakanishi et al. 1967) and FeS melt
boundary (330 GPa and 5000 K) for the Fe-FeS melt, which is
(Barmin et al. 1970) are also shown in Figure 3. Assuming an
close to the previous theoretical estimations (Poirier 1988; Alfè
Arrhenius temperature dependence of viscosity, extrapolated 1
and Gillan 1998a, 1998b).
atm Fe and FeS viscosities to low temperature are slightly lower
than our data. This small difference is consistent with theoreti- ACKNOWLEDGMENTS
cal prediction of the pressure dependency of viscosity, which The X-ray experiments were carried out at SPring-8 (proposal no.
is expected to be small and positive for liquid metal (e.g., Poirier 1999A0297-ND-np). The authors thank M. Ito, M. Kanzaki, S. Kasai, T. Katsura,
582 URAKAWA ET AL.: VISCOSITY OF THE Fe-FeS MELTS AT P
T. Kubo, K. Mibe, N. Yagi, S. Yamashita, and A. Yasuda for assistance to radi-
ography experiments. The authors are also grateful to D. Dobson, L. Vočadlo,
and A.P. Jones for helpful discussion. The authors appreciate the critical com-
ments on manuscript from R. Secco and D. Dingwell. The JEOL microprobe is
a supported facility of the Chemical Analysis Center, University of Tsukuba.
REFERENCES CITED
Alfè, D. and Gillan, M.J. (1998a) First principles simulations of liquid Fe-S under
earth’s core conditions. Physical Review B, 58, 8248–8256.
———(1998b) First-principles calculation of transport coefficients. Physical Re-
view Letters, 81, 5161–5164.
Anderson, O.L., Isaak, D.G., and Yamamoto, S. (1989) Anharmonicity and the equa-
tion of state for gold. Journal of Applied Physics, 65, 1534–1543.
Barmin, L.N., Estin, O.A., and Dobrovinskii, I.E. (1970) Primenenie teorii reguljarnix
rastvorov k oposaniju izoterm vjazkosti i moljarnix obemov binarnix suldnix
rasplavov. Zhurnal Fizicheskoi Khimii, 44, 2560–2563.
Brazhkin, V.V. (1998) Investigation of the crystallization of liquid iron under pres-
sure: extrapolation of the melt viscosity into the megabar range. JETP Letters,
68, 502–507.
Dingwell, D.B. (1998) Melt viscosity and diffusion under elevated pressures. Re-
views in Mineralogy, 37, 397–424.
Dobson, D.P., Jones, A. P., Rabe, R., Sekine, T., Kurita, K., Taniguchi, T., Kondo, T.,
Kato, T., Shimomura, O., and Urakawa, S. (1996) In-situ measurement of vis-
cosity and density of carbonate melts at high pressure. Earth and Planetary
Science Letters, 143, 207–215.
Dobson, D.P., Crichton, W. A., Vočadlo, L., Jones, A. P., Wang, Y., Uchida, T., Riv-
ers, M., Sutton, S., and Brodholt, J.P. (2000) In situ measurement of viscosity of
liquids in the Fe-FeS system at high pressures and temperatures. American
Mineralogist, 85, 1838–1842.
Faxen, H. (1925) Gegenseitige Einwirkung zweier Kugeln, die in einer zahen
fkussigkeit fallen, Arkiv for Matematik, Astronomi och Fysik, 19, 1–8.
Iida, T. and Guthrie, I. L. (1988) The Physical Properties of Liquid Metals. 288 p.
Clarendon Press, Oxford.
Jamieson, J.C., Fritz, J.N., and Manghnani, M.H. (1982) Pressure measurement at
high temperature in X-ray diffraction study: gold as a primary standard. In S.
Akimoto and M.H. Manghnani, Eds., In High-Pressure Research in Geophys-
ics, 27–48 p. Center for Academic Publishing, Tokyo.
Kanzaki, M., Kurita, K., Fujii, T., Kato, T., Shimomura, O., and Akimoto, S. (1987)
A new technique to measure the viscosity and density of silicate melts at high
pressure. In M.H. Manghnani and Y. Syono, Eds., High-Pressure Research in
Mineral Physics, p. 195–200. American Geophysical Union, Washington, D.C.
LeBlanc, G.E. and Secco, R.A. (1996) Viscosity of an Fe-S liquid up to 1300°C and
5 GPa. Geophysical Research Letters, 23, 213–216.
Nagamori, M. (1969) Density of molten Ag-S, Cu-S, Fe-S, and Ni-S systems. Trans-
actions of the Metallurgical Society of AIME, 245, 1897–1902.
Nakanishi, K., Sato, T., and Shiraishi, Y. (1967) On the viscosity of molten iron and
its dilute binary alloys of aluminum, silicon, and oxygen. Journal of the Japan
Institute of Metals, 37, 881–887 (in Japanese).
Poirier, J.P. (1988) Transport properties of liquid metals and viscosity of the earth’s
core. Geophysical Journal, 92, 99–105.
Sanloup, C., Guyot, F., Gillet, P., Fiquet, G., Mezouar, M., and Martinez, I. (2000)
Density measurements of liquid Fe-S alloys at high-pressure. Geophysical Re-
search Letters, 27, 811,814.
Secco, R.A. (1995) Viscosity of the outer core. In T.J. Ahrens, Ed., Mineral Physics
and Crystallography, A Handbook of Physical Constants, p. 218–226. AGU
Reference Shelf Series, Vol. 2, Washington, D.C.
Secco, R.A., LeBlanc, G.E., Yang, H., and Seibel, J.N. (1998) High pressure viscos-
ity of an Fe-S liquid: experimentally derived estimate of the viscosity of earth’s
outer core. In Properties of Earth and Planetary Materials at High Pressure and
Temperature, Geophysical Monograph 101, p. 495–505. American Geophysi-
cal Union, Washington, D.C.
Touloukian, Y. S. (1970) Thermal Diffusivity: Thermodynamical Properties of Mat-
ter, 10, 83–100, IFI/Plenum, New York.
Urakawa, S., Kato, M., and Kumazawa, M. (1987) Experimental study on the phase
relation in the system Fe-Ni-O-S up to 15 GPa. In M.H. Manghnani and Y.
Syono, Eds., High-Pressure Research in Mineral Physics, p. 95–111. American
Geophysical Union, Washington, D.C.
Urakawa, S., Igawa, N., Kusaba, K., Ohno, H., and Shimomura, O. (1998) Structure
of molten iron sulfide under pressure. Review of High Pressure Science and
Technology, 7, 286–288.
Urakawa, S., Funakoshi, K., Shimomura, O., Ohno, H., and Kikegawa, T. (2001)
Structure of molten iron-sulfur alloys at 3 GPa submitted to Earth and Plan-
etary Science Letters, in press.
Usselman, T.M. (1975) Experimental approach to the state of the core: Part I. The
liquidus relations of the Fe-rich portion of Fe-Ni-S system from 30 to 100 kb.
American Journal of Science, 275, 278–290.
Utsumi, W., Funakoshi, K., Urakawa, S., Yamakata, M., Tsuji, K., Konishi, H., and
Shimomura, O. (1998) SPring-8 beamlines for high pressure science with multi-
anvil apparatus. Review of High Pressure Science and Technology, 7, 1484–
1486.
Vočadlo, L., Alfè, D., Price, G.D., and Gillan, M.J. (2000) First principles calcula-
tions on the diffusivity and viscosity of liquid Fe-S at experimentally accessible
conditions. Physics of the Earth and Planetary Interiors, 120, 145–152.
MANUSCRIPT RECEIVED SEPTEMBER 7, 2000
MANUSCRIPT ACCEPTED JANUARY 8, 2001
PAPER HANDLED BY ANNE M. HOFMEISTER