Lunar and Planetary Science XLVIII (2017)
1505.pdf
Experimental Constraints on Metal Percolation through Silicate: Implications for Core Formation on Asteroids and Planetesimals. M. S. Duncan1 and Y. Fei1, 1Geophysical Laboratory, Carnegie Institution of Washington,
Washington, D. C., USA ([email protected], [email protected])
Introduction: Core formation is a seemingly
ubiquitous process in the early solar system, yet the
mechanics of metal segregation on smaller planetary
bodies is debatable. Small bodies, e.g., early-formed
planetesimals, asteroids, planetary embryos etc., were
potentially too cold for complete silicate melting, i.e.,
no magma ocean, yet core formation occurred on ~5
Myr timescales [1]. This invokes scenarios of efficient
core formation by percolation of metallic melt through
a solid silicate matrix, a difficult process as the wetting
ability of Fe melt is negligible due to its high dihedral
angle (Θ>90°) [2-5]. In order to enhance the percolative ability of the metallic melt, several modifications
to the scenario have been proposed. The simplest adjustment increases the volume of metallic melt present
to >5-10 vol.% [5-7] at which point density driven
flow, similar to Rayleigh-Taylor instabilities, would
cause the metallic phase to cross the percolation
threshold. This is likely in the earliest stages of planetary differentiation, but as differentiation proceeds, or
if the body is metal-poor, this density-driven process
will not operate.
Another well-studied scenario is the addition of
light elements (S [3,4,8-15], S-O [5,13-18], Si [19,20])
to the metallic phase that reduced the dihedral angle of
the metallic phase to <60° allowing it to wet grain
boundaries and percolate to the core. At the lower
pressures (<~10 GPa) of core formation on small
planetary bodies, the addition of O and S causes the
strongest reduction in Θ to ≤60° allowing it to wet,
while the addition of S only has a less significant effect (Θ>60°), and the addition of Si has little to no
effect. The effect of C has not been studied, nor has
the combined effect of several light elements, e.g.
S+C+O.
A third, less studied scenario includes the presence
of a silicate melt in addition to the metallic melt [e.g.,
21]. As silicate melts have much lower Θ values
(<45°), they wet grain boundaries very well. Studies
done with Fe-S metallic phase [9,10] did not find any
enhancement of metallic melt wetting in the presence
of silicate melt. However, at high enough melt fractions, the metallic melt will form discrete droplets inside silicate melt channels that may allow for densitydriven separation, though this scenario is likely more
complex [21].
Methods: Two sets of piston cylinder experiments
were conducted to determine the 1) dihedral angle of
C-bearing metallic melts and 2) percolation experi-
ments to determine the silicate melt fraction at which
separation of metallic phase occurs. All experiments
contained a San Carlos olivine-pyroxene mix acting as
the silicate matrix for the metallic phase, were placed
in fired MgO containers, and held at the P and T of
interest for 6 hrs, after a sintering step at 800 °C for 4
hrs. Pressure was varied between 1.0 and 3.0 GPa at
1400 °C and temperature ranged from 1400-1600 °C at
3.0 GPa. Dihedral angle experiments consisted of ~5
vol.% Fe-C mix, with either ~2 or 4.3 wt.% C, i.e., on
the Fe-rich side of the Fe3C eutectic and at the eutectic.
Percolation experiments consisted of layers of metalfree silicate and metal-bearing silicate, where the metallic phase was Fe-C or Fe-C-S.
The compositions and texture of the quenched metallic liquid and silicate matrix and melt were analyzed
with the electron microprobe and SEM. The dihedral
angles were determined using the traditional method of
measurement from a 2-D image slice (apparent angles)
and using a 3-D imaging technique (FIB/SEM) [22].
Results: The dihedral angle experiments showed
that carbon has little to no effect on the dihedral angle
of the metallic melt in textural equilibrium with the
silicate. Measured apparent angles in the Fe-C experiments conducted at 1400 °C range from ~110 to 122°
as pressure decreases from 3.0 to 1.0 GPa (Fig. 1). At
higher temperatures at 3.0 GPa, as the silicate melt
fraction increased (~2 to 15% from 1400 to 1600 °C),
the metal droplets were completely enclosed in silicate
melt pockets (with apparent Θ ~26°, Figs. 1 and 2),
precluding angle measurements of the metallic phase.
Figure 1. Apparent Θ measurements for Fe-C metallic
phase in this study (dark blue: 2 wt.% C, turquoise: 4.3
wt.% C) compared to Θ measurements of immiscible sulfide melt (yellow) and silicate melt (purple) at 3.0 GPa.
Lunar and Planetary Science XLVIII (2017)
Olivine
Olivine
1505.pdf
Silicate
melt
Silicate melt
C-bearing
C-bearing
25 μm
25 μm
Olivine
S-rich
C-rich
25 μm
Figure 2. SEM images of experimental run products. Top
(this study) left: PC1591, Fe + 2 wt.% C, 3.0 GPa,
1400°C; right: PC1580, Fe + 2 wt.% C, 3.0 GPa, 1600°C.
Bottom center [23]: FeSO6, Fe + S + C, 3.0 GPa, 1600°C.
A similar situation occurs in the Fe-S-C system
from previous work (Fig. 2, bottom) in that at low
pressures (<6 GPa), metal-metal immiscibility causes
separation of an Fe-S phase and an Fe-C phase [23].
As the interfacial energy between the olivine and the
Fe-S phase is lower than that between the olivine and
the Fe-C phase, the Fe-S phase ‘wets’ (median Θ
~90°) and encloses the Fe-C phase.
Discussion/Implications: For both the silicate melt
– Fe-C and the Fe-S – Fe-C systems, the Fe-C phase
sits in the interior of these ‘channels’. In the case of
the silicate melt – Fe-C system, the silicate melt (ρ~2.9
g/cm3) is buoyant with respect to the solid olivine
(ρ~3.2 g/cm3), while the Fe-C melt (ρ~7 g/cm3) is significantly denser than both. If the Fe-C droplets are
small enough, the silicate melt will likely carry them
upwards away from the core. However, if the density
difference between the Fe-C melt and the silicate melt
is the driving factor, the Fe-C melt will travel down the
silicate channels either to the core or to a choke point
where the melt fraction has decreased and the metal
can no longer flow. This will effectively stop core formation, unless this state persists allowing a build-up of
metal at the choke point (increasing volume fraction of
metal to form a large metal droplet), at which point the
density-driven process will become active.
In the immiscible metal case, both the Fe-S (ρ~4.2
g/cm3) and Fe-C phases are denser than the olivine,
though the Fe-C is denser than the Fe-S. This may
result in the Fe-C phase moving through the Fe-S
phase to the core, though if the Fe-S melt is wetting,
the two phases will likely track together and mix with
the core.
These scenarios will also control the trace element
composition of the core as the partitioning behavior
between olivine and Fe-S is different than the partitioning between silicate melt and Fe-C. For example,
the partition coefficient for Co between FeS melt and
olivine is ~9 [14], while it is closer to 80 [24] between
Fe-C melt and silicate melt, depending on the P, T, and
silicate melt composition. While only a one order of
magnitude difference, it would suggest that there
would be less Co in the residual silicate if the Fe-C
phase can separate from the silicate melt and flow to
the core.
It is clear that carbon alone cannot enhance the
wetting ability of metallic melt, as would be expected
from the metallurgy literature. However, a mix of
volatiles under conditions that allow for immiscibility,
and/or the presence of a silicate partial melt may create
channels that allow an interior metallic phase (in this
case Fe-C) to flow down to the core.
References: [1] Kleine et al. (2002) Nature, 418,
952-955. [2] Jurewicz and Jones (1995) LPSC XXVI,
709-710. [3] Shannon and Agee (1996) GRL, 23,
2717-2720. [4] Ballhaus and Ellis (1996) EPSL, 143,
137-145. [5] Minarik et al. (1996) Science, 272, 530533. [6] von Bargen and Waff (1986) JGR, 91, 92619276. [7] Yoshino et al. (2003) Nature, 422, 154-157.
[8] Shannon and Agee (1998) Science, 280, 10591061. [9] Holzheid et al. (2000) JGR, 105, 1355513567. [10] Holzheid (2013) EJM, 25, 267-277.
[11] Roberts et al. (2007) GRL, 34, L14306.
[12] Bagdassarov et al. (2009) PEPI, 177, 139-146.
[13] Shi et al. (2013) Nat. Geo., 6, 971-975.
[14] Gaetani and Grove (1999) EPSL, 169, 147-163.
[15] Terasaki et al. (2005) EPSL, 232, 379-392.
[16] Rose and Brenan (2001) Econ. Geol., 96, 145157. [17] Terasaki et al. (2007) PEPI, 161, 170-176.
[18] Terasaki et al. (2008) EPSL, 273, 132-137.
[19] Takafuji et al. (2004) EPSL, 224, 249-257.
[20] Mann et al. (2008) PEPI, 167, 1-7.
[21] Stevenson (1990) Origin of the Earth, p. 231-250.
[22] Fei (2013) J. Vis. Exp., 81, e50778. [23] Fei and
Shibazaki (2016) LPS XLVII, #1719. [24] Seibert et al.
(2011) GCA, 75, 1451-1489.