A preview of this full-text is provided by Wiley.
Content available from Journal of Geophysical Research: Planets
This content is subject to copyright. Terms and conditions apply.
Constraints on the Abundances of Carbon and
Silicon in Mercury's Core From Experiments
in the Fe‐Si‐C System
Kathleen E. Vander Kaaden
1
, Francis M. McCubbin
2
, Amber A. Turner
1,3
,
and D. Kent Ross
1,4
1
Jacobs, NASA Johnson Space Center, Houston, TX, USA,
2
NASA Johnson Space Center, Houston, TX, USA,
3
Department of Geoscience, University of Las Vegas, Las Vegas, NV, USA,
4
University of Texas at El Paso‐CASSMAR, El
Paso, TX, USA
Abstract The composition of a planet's core has important implications for the thermal and magmatic
evolution of that planet. Here, we conducted carbon (C) solubility experiments on iron‐silicon (Fe‐Si)
metal mixtures (up to 35 wt% [~52 atom%] Si) at 1 GPa and 800–1800°C to determine the carbon
concentration at graphite saturation (CCGS) in metallic melt and crystalline metal with varying proportions
of Fe and Si to constrain the C content of Mercury's core. Our results, combined with those in the literature,
show that composition is the major controlling factor for carbon solubility in Fe‐rich metal with minimal
effects from temperature and pressure. Moreover, there is a strong anticorrelation between the abundances
of carbon and silicon in iron‐rich metallic systems. Based on the previous estimates of <1–25 wt% Si in
Mercury's core, our results indicate that a carbon‐saturated Mercurian core has 0.5–6.4 wt% C, with 6.4 wt%
C corresponding to an Si‐free, Fe core and 0.5 wt% C corresponding to an Fe‐rich core with 25 wt% Si. The
upper end of estimated FeO abundances in the mantle (up to 2.2 wt%) are consistent with a core that has
<1 wt% Si and up to 6.4 wt% C, which would imply that bulk Mercury has a superchondritic Fe/Si ratio.
However, the lower end of estimated FeO (≤0.05 wt%) supports CB chondrite‐like bulk compositions of
Mercury with core Si abundances in the range of 5–18.5 wt% and C abundances in the range of 0.8–4.0 wt%.
Plain Language Summary The composition of a planet's core can provide clues as to how the
planet has changed over time. In this study, we conducted experiments at high pressures and
temperatures to investigate potential carbon and silicon abundances in the core of Mercury. We utilized a
variety of iron‐silicon metal mixtures (up to 35 wt% silicon) and graphite capsules in order to examine the
concentration of carbon in metallic melts and crystalline metals at graphite saturation with the intention of
constraining the carbon and silicon content of Mercury's core. Combining the results of this study with those
in the literature, we found that composition is the major controlling factor of carbon solubility in
silicon‐bearing, iron‐rich metal, with minimal effects from temperature. More importantly, our results
showed a strong anticorrelation between the abundances of carbon and silicon in iron‐rich metallic systems.
Since Mercury may have formed in a region of the solar system with less oxygen available, it is likely that
some silicon partitioned into Mercury's core as silicon becomes more siderophile under reducing conditions.
These findings, when combined with other elemental data, can be used to place constraints on the bulk
composition of Mercury, which could help to constrain its origin.
1. Introduction
Results from the MErcury Surface, Space ENvironment, GEochemistry and Ranging (MESSENGER) space-
craft indicate elevated abundances of carbon (C) on the surface of Mercury (Klima et al., 2018; Murchie
et al., 2015; Peplowski et al., 2015, 2016). Recent studies show carbon enrichment as high as 4 wt% over
the local mean in the low reflectance materials (LRM) excavated from depth and as much as 2.5 wt% carbon
enrichment in regional deposits associated with the most heavily cratered terrains (Klima et al., 2018).
Furthermore, the X‐Ray Spectrometer on board MESSENGER measured up to 3 wt% sulfur (S) and less than
2 wt% iron (Fe) on Mercury's surface (e.g., Nittler et al., 2011; Vander Kaaden et al., 2017; Weider et al., 2012),
suggesting the planet's oxygen fugacity is between 2.6 and 7.3 log
10
units below the Iron‐Wüstite (IW) buffer
(McCubbin et al., 2012; Zolotov et al., 2013). As discussed by Nittler et al. (2011) and Hauck II et al. (2013),
©2020. American Geophysical Union.
All Rights Reserved.
RESEARCH ARTICLE
10.1029/2019JE006239
Key Points:
•Composition is the main control of
carbon concentration at graphite
saturation in Fe‐rich metal, with
minimal effects from temperature
and pressure
•Mercury requires ≥27 wt% Si and
≤0.5 wt% C in its core if it has a bulk
EH chondritic composition
•Mercury requires 5–18.5 wt% Si and
0.8–4.0 wt% C in its core if it has a
bulk CB chondritic composition
Supporting Information:
•Supporting Information S1
•Table S1
•Table S2
Correspondence to:
K. E. Vander Kaaden,
kathleen.e.vanderkaaden@nasa.gov
Citation:
Vander Kaaden, K. E., McCubbin,
F. M., Turner, A. A., & Ross, D. K.
(2020). Constraints on the abundances
of carbon and silicon in Mercury's core
from experiments in the Fe‐Si‐C system.
Journal of Geophysical Research:
Planets,125, e2019JE006239. https://
doi.org/10.1029/2019JE006239
Received 14 OCT 2019
Accepted 20 MAR 2020
Accepted article online 30 APR 2020
VANDER KAADEN ET AL. 1of15
under such highly reducing conditions, the majority of the iron available on the planet partitions into the
core, leaving behind a silicate mantle that is highly depleted in FeO, although the magnitude of the depletion
is still an open question. In fact, estimates of the FeO abundance in the mantle range from 1.0–2.2 wt% on the
basis of measured iron abundances from the surface (McCubbin et al., 2017) and from 0.02–0.03 wt% on the
basis of melting experiments conducted at an oxygen fugacity of 5–6 log
10
units below the IW buffer (Nittler
et al., 2019, and references therein). Even more enigmatic, Mercury also has a large core and thin mantle
compared to the other terrestrial planets in the solar system, with the core‐mantle boundary at an estimated
pressure (P) of only 4–7 GPa (e.g., Hauck II et al., 2013). Combined, these constraints have important impli-
cations for the constitution, structure, and thermochemical evolution of Mercury.
One of the implications of Mercury's low‐FeO silicate composition is that the density of silicate melts on
Mercury are lower and within a narrower range than other planetary bodies (Agee, 1998; Agee &
Walker, 1993; Bertka & Fei, 1997; Misawa, 2004; Vander Kaaden et al., 2015; Vander Kaaden &
McCubbin, 2015; Warren et al., 1996). In fact, the melt density was sufficiently low that graphite would have
been the only buoyant major rock‐forming mineral in a low‐FeO Mercurian magma ocean (Vander Kaaden
& McCubbin, 2015). Given the evidence that coarse‐grained graphite was the primary darkening agent in the
LRM (Murchie et al., 2015) and that the pressure of the core‐mantle boundary on Mercury is below the dia-
mond stability field, Vander Kaaden and McCubbin (2015) hypothesized that Mercury had a primary flota-
tion crust of graphite. This hypothesis was later supported by Peplowski et al. (2016) who used the
low‐altitude data collected toward the end of the MESSENGER mission to show that carbon is the only mate-
rial that is consistent with both the visible to near‐infrared spectra and the neutron measurements of low
reflectance material on Mercury. They also showed that graphite was sourced from depth and likely com-
prised a global layer of graphitic material, consistent with a graphite flotation crust. More recently, Klima
et al. (2018) found that the 600‐nm band depth in LRM deposits is related to carbon content and estimated
enrichments in carbon as high as 4 wt% over the local mean in LRM deposits and an average enrichment of
~2.5 wt% carbon in regions associated with the most heavily cratered terrains on Mercury.
The confirmation of carbon on the surface of Mercury leads to many subsequent questions regarding the role
of carbon during the differentiation and evolution of Mercury. In particular, the presence of a primary gra-
phite flotation crust implies that the core was saturated in carbon, at least at the core‐mantle boundary at the
time of differentiation. Given that the solubility of carbon in silicate melts is negligible at low oxygen fugacity
and low hydrogen fugacity (Ardia et al., 2013; Hirschmann et al., 2012; Li et al., 2015, 2017), the bulk carbon
content of Mercury is dependent, primarily, on the solubility of carbon in the Mercurian core, although the
abundances of hydrogen in the mantle would have exerted some control on the storage capacity of carbon in
the silicate mantle at graphite saturation (Ardia et al., 2013; Hirschmann et al., 2012; Li et al., 2015, 2017;
McCubbin & Barnes, 2019). Given the high bulk density of Mercury and relatively thin silicate mantle,
Mercury is assumed to have a large metallic core that is dominated by iron and likely alloyed with at least
one light element (e.g., Chabot et al., 2014; Chen et al., 2008; Hauck & Johnson, 2019; Malavergne
et al., 2010). Prior to, and with preliminary data from, the MESSENGER mission, sulfur was the main light
element of focus for alloying with iron in Mercury's core (e.g., Hauck II et al., 2007; Riner et al., 2008;
Stevenson et al., 1983). However, with the highly reduced nature of Mercury exposed through
MESSENGER results, the focus shifted to silicon as the likely primary light element in Mercury's core
(e.g., Chabot et al., 2014; Margot et al., 2019). Although the exact composition of Mercury's core is unknown,
purportedly, it lies somewhere between an iron‐silicon end‐member and a mixture of iron, silicon, and sul-
fur (Margot et al., 2019).
In the present study, we seek to evaluate the implications of a graphite flotation crust on the composition of
Mercury's core with a primary focus on the Fe‐Si‐C system and a secondary focus on the Fe‐Ni‐S‐C‐Si system.
Importantly, numerous researchers have examined carbon‐bearing metallic systems in the past (e.g.,
Bouchard & Bale, 1995; Chabot et al., 2006, 2017; Dasgupta et al., 2013; Dasgupta & Walker, 2008;
Kawanishi et al., 2009; Li et al., 2015, 2016, 2017; Righter & Drake, 2000; Wood, 1993, and references
therein), but they have not focused exclusively on the implications for carbon in the Mercurian core.
Furthermore, many previous studies did not have access to the Field Emission Gun Electron Microprobe
and cold finger attachment used in this study to strengthen the vacuum and minimize carbon contamination
allowing for higher precision carbon measurements in the metallic phases; nor did they report experiments
over a wide range of temperatures (T). Therefore, we supplement existing experimental data with new
10.1029/2019JE006239
Journal of Geophysical Research: Planets
VANDER KAADEN ET AL. 2of15
experimental data on carbon solubility within iron‐silicon metal mixtures
(up to 35 wt% silicon) at 1 GPa and 800–1800°C. These experiments were
conducted within graphite capsules to determine the carbon concentra-
tion at graphite saturation (CCGS) in metallic melt and crystalline metal
with varying proportions of iron and silicon. Moreover, carbon abun-
dances in the metals were determined using an analytical method devel-
oped specifically for the accurate measurement of carbon within
metallic phases in petrologic experiments (Li et al., 2015). Our experimen-
tal results are combined with previously published data on the
Fe‐Ni‐S‐C‐Si system to determine the effects of pressure, temperature, and composition on carbon solubility
in metallic phases at graphite saturation. We use these data to evaluate the implications of a
graphite‐saturated core on Mercury to place constraints on the C and Si abundances in Mercury's core over
a range of proposed bulk Mercury compositions. Although we do not have definitive constraints on the bulk
composition of Mercury at this time, our results are used to determine under what conditions Mercury has
chondritic or superchondritic Fe/Si and C/Si ratios when the core is saturated in carbon. Our results are also
used to place broad constraints on the abundances of C and Si in Mercury's core on the basis of the Fe‐Si‐C
system as well as the Fe‐Ni‐S‐C‐Si system.
2. Methods
2.1. Starting Materials
The synthetic starting materials (Table 1) used to investigate CCGS were prepared at the Institute of
Meteoritics, University of New Mexico (UNM), using high‐purity Fe and Si metal powders that were hand
mixed in a glass vial for several hours for homogenization. Four different metallic mixtures, each with a dis-
tinct Fe:Si ratio, were produced in order to examine a range of possible Si abundances in the Mercurian core.
The Low Si (5 wt% Si) and Intermediate (Int)‐High Si (22 wt% Si) compositions were chosen based on the
thermal minima of the liquidus temperatures along the Fe‐Si metallic‐binary join (Predel, 1995, and refer-
ences therein). The Int‐High Si composition is also close to the upper limit of Si in a Mercurian core (e.g.,
up to 25 wt% Si; Chabot et al., 2014; Margot et al., 2019). The Low‐Int Si and High Si compositions were cho-
sen to expand the range of possible core compositions on Mercury to be examined.
2.2. Piston Cylinder Experimental Methods
With the exception of the 1300°C experiments, which were conducted at UNM following the procedures out-
lined in Vander Kaaden and McCubbin (2016), all piston cylinder (PC) CCGS experiments were conducted
in the high‐pressure laboratory at NASA Johnson Space Center (JSC). Each experiment began by first pack-
ing one of the Si‐Fe metal mixtures into a graphite capsule using a Teflon coated spatula and wooden tamper
to minimize Fe loss due to magnetization with the spatula/tamper. The main difference between the UNM
setup and the JSC setup is the pressure medium used in the experiments. For the CCGS experiments at JSC,
the loaded graphite capsules were placed within barium carbonate (BaCO
3
) cell pressure media, with crush-
able MgO parts and a graphite furnace. A hard‐fired alumina disk was placed between the top of the thermo-
couple wire and the graphite capsule to ensure no contact during the run that could result in oxidation or
corrosion of the thermocouple wire. Experiments were conducted in a non‐end‐loaded piston cylinder appa-
ratus, calibration given in Appendix A. The temperature and pressure calibration methods for the JSC PC
and the UNM PC were identical (see McCubbin et al., 2015; Vander Kaaden et al., 2015), which lends con-
fidence to the direct comparison of our experimental results. A Type C (W‐5%Re vs. W‐26%Re) thermocouple
was used to monitor temperature throughout the run and was controlled by either a Love controller or a
Eurotherm 2416 controller throughout the duration of each run. Each experiment was quenched by shutting
off power to the furnace and slowly decompressing the run. Experiments were run at ~1.0 GPa in the tem-
perature range of 800–1800°C, with run durations of 8–24 hr, with the exception of the sole 1800°C experi-
ment that was held for ~1.25 hr.
2.3. Approach to a Steady State
The geometric configuration of our experiments can be used to determine the minimum amount of time
required to reach a steady state. We computed the time it would take for a single atom of carbon to diffuse
from the graphite capsule across the entirety of the experimental charge along the longest dimension of the
Table 1
Composition of the Metal Starting Materials Used in the CCGS Study
Low Si Low‐Int Si Int‐High Si High Si
Si 5 (9.5) 10 (18.1) 22 (35.9) 35 (51.7)
Fe 95 (90.5) 90 (81.9) 78 (64.1) 65 (48.3)
Total 100 100 100 100
Note. Values are in wt% with atom% listed parenthetically.
10.1029/2019JE006239
Journal of Geophysical Research: Planets
VANDER KAADEN ET AL. 3of15
capsule (typically ~0.25 cm), which is a conservative measurement given that graphite is present on all sides
and hence only half this distance is required. Previous studies have reported a range in diffusion coefficients
for C in Fe metal (Tibbetts, 1980; Wert, 1950). At 1500°C it takes an average of 0.71 hr for C to diffuse across
the longest dimension of the graphite capsule with a maximum time of 1.11 hr, depending on the diffusion
coefficient utilized. At higher temperatures of 1800°C however, it takes, on average, 0.18 hr for C to diffuse
across the longest dimension of the graphite capsule with a maximum of 0.28 hr.
Additionally, a set of time series experiments was conducted on the Int‐High composition at 1 GPa. The time
series consisted of experiments run at 1500°C and held at target pressure‐temperature (P‐T) conditions for 1,
4, 8, and 11 hr to assure the calculated time of 0.75 hr using carbon diffusion coefficients was sufficient for
approaching a steady state in these experiments. The results from the time series experiments are shown in
Figure 1. These data indicate that carbon, iron, and silicon abundances are very similar for the 1‐hr duration
experiments and the 11‐hr duration experiments. Therefore, only experiments held for longer than 1 hr were
considered in this study.
2.4. Analysis of CCGS Experiments
All run products were polished using hexagonal boron nitride powder as a lubricant instead of water to
ensure no carbon‐bearing phases were lost from the experimental charges (Murthy et al., 2003). All phases
within the CCGS experiments were analyzed for Fe, Si, and C using a JEOL 8530F microprobe at NASA's
JSC. For the analysis of C, we adapted existing procedures that have been developed and verified through
comparison with secondary ion mass spectrometry (e.g., Dasgupta & Walker, 2008; Li et al., 2015). Those
techniques were developed at NASA JSC, and sample preparation protocols were specifically optimized
for the quantitative analysis of C. We did not coat our samples with carbon or any other conducting material,
so each sample was painted with Pelco® colloidal silver slurry from the capsule to the edge of the 1‐inch
round epoxy plug to ensure overlap and electrical contact with the sample holder. To minimize charging
Figure 1. CCGS time series on Si
22
Fe
78
metal at 1 GPa and 1500°C. Symbols correspond to a particular element. Error
bars are the standard deviation of the data points collected for a given element in that experiment.
10.1029/2019JE006239
Journal of Geophysical Research: Planets
VANDER KAADEN ET AL. 4of15
effects during analyses, the edges of the capsule were also surrounded by silver paint. Since each experiment
only contained metal, there was no need to coat the samples in an additional conductive material. All
analyses were collected at 15 keV and 30 nA, and we kept the cold finger on the electron microprobe
filled with liquid N
2
(LN
2
) throughout the duration of our probe sessions to strengthen the vacuum and
minimize carbon contamination during the residence time within the instrument (Robaut et al., 2006).
Specifically, the cold finger was filled with LN
2
prior to the beginning of each session (causing the
vacuum to initially increase), allowed to cool for approximately 1 hr while the vacuum on the microprobe
strengthened to optimal operating conditions (9E
−5
Pa), and then refilled periodically throughout each
analytical session. The use of the cold finger significantly diminishes ambient carbon contamination
during electron probe analysis. For example, analysis done in this laboratory at a single location to
measure C in Fe metal showed an increase in C concentrations of <0.8 wt% after 600 s when the cold
finger was in use but >3 wt% after 600 s when the cold finger was not in use.
Si, Fe, and C were analyzed in each experiment. Si (Si‐metal standard) and Fe (Fe metal standard) were ana-
lyzed using the TAP and LIFH crystals, respectively. C was analyzed using the LDE2 synthetic multilayer
crystal and was standardized using a synthetic cohenite (Fe
3
C) synthesized at 1 GPa, 1162°C, for 339 hr in
the piston cylinder apparatus at JSC (Righter et al., 2017). An additional check on the C standardization
was provided by a synthetic moissanite (SiC) (Figure 2d) that was synthesized at 1 GPa, 1800°C, for
~40 min in the piston cylinder apparatus at JSC. The crystal structure and composition of the synthetic cohe-
nite and moissanite standards were verified by selected area electron diffraction utilizing a JEOL 2500SE
field emission scanning transmission electron microscope at JSC under the direction of Dr. Lindsay
Keller. Owing to the wide X‐ray peaks on the LDE2 crystal during Electron Probe MicroAnalysis (EPMA),
as well as an interference between the backgrounds of C and Si, an optimal background was chosen to
Figure 2. Backscattered electron images of experimental run products (a) CSM‐1121 (1 GPa, 800°C, Si
10
Fe
90
metal), (b)
CSM‐1003 (1 GPa, 1700°C, Si
10
Fe
90
metal), (c) CS‐3 (1 GPa, 1300°C, Si
5
Fe
95
metal), and (d) SiC standard used in EPMA
analyses.
10.1029/2019JE006239
Journal of Geophysical Research: Planets
VANDER KAADEN ET AL. 5of15
ensure this overlap was avoided. Peak and background counts for major elements were 30 and 15 s,
respectively, and peak and background counts for carbon were 60 and 30 s, respectively. Error for each
element in each analysis is reported in Table 2 as the standard deviation of a given element from “n”
analyses. All data were corrected using the phi‐rho‐z correction method, which is ideal when analyzing
light elements (Merlet, 1994).
3. Results
3.1. High‐Temperature (1300–1800°C) Experiments
Each CCGS experiment conducted at or above 1300°C resulted in a single liquid at run conditions that dis-
played a variety of textures when quenched to a solid (Figures 2a–2c). This liquid metal was a single phase at
the P‐Tconditions of our runs and typically quenched to a dendritic texture of Si‐Fe‐rich, C‐poor dendrites
surrounded by C‐rich, Fe‐Si interstitial phases. Although present throughout the experimental charge,
higher concentrations of dendrites are typically found near the edges of the capsules or around graphite
grains suggesting they are nucleating on the available solid surfaces. In the experiments containing
≥22 wt% Si, the interstitial phases were typically too small to accurately analyze using EPMA. In some
experimental charges, vermicular intergrowths of graphite in metal were also present. These textures are
similar to superliquidus metals produced by Zhang et al. (2018) when quenched to room temperature.
In order to fully characterize each experimental charge, we analyzed the samples using both broad beam
analyses (~15–20 μm) representative of the melt composition at high P‐Tconditions and spot analyses
(~1–5μm) on the dendrites and surrounding regions (see supporting information). For all discussions of
composition moving forward, only the broad beam analyses are considered. For the high‐temperature
experiments, the broad beam analyses show a range in C abundances from 0.6–4.5 wt% (Table 2;
Figure 3). Broad beam analyses for the high‐Texperiments, representative of the average interstitial compo-
sition at the experimental P‐Tconditions, incorporating both quenched melt and dendritic phases, con-
tained ~4.5–22.4 wt% Si and ~79.2–92.8 wt% Fe, indicative of the wide range in starting compositions that
were used in this study (Table 1).
Table 2
EPMA Data for All Broad Beam (~15–20 μm) Analyses on the CCGS Experiments (wt%)
Exp #
Nominal mix
composition (wt%)
T1
(°C)
Hold time
(hr)
T2
(°C)
Hold
time2 (hr) n Si (wt%) Std C (wt%) Std Fe (wt%) Std Total (wt%) Std
CS‐3 Si5Fe95 1300 24 23 4.58 0.4 4.5 0.5 92.09 0.6 101.2 0.6
CSM‐982 Si5Fe95 1500 8 16 5.07 0.1 3.5 0.3 92.50 0.3 101.1 0.2
CSM‐983 Si5Fe95 1600 8 11 4.79 0.1 3.5 0.4 92.31 0.8 100.6 1.0
CSM‐993 Si5Fe95 1700 8 15 5.24 0.4 3.6 0.8 91.91 0.9 100.8 1.0
CSM‐997 Si5Fe95 1800 1.25 13 9.16 0.3 2.5 0.3 89.44 0.6 101.1 0.4
CSM‐1120 Si5Fe95 1500 0.25 800 9 10 4.96 0.4 1.6 0.4 94.23 1.3 100.8 1.1
CSM‐1123 Si5Fe95 1500 0.25 900 9 15 5.05 0.9 1.7 1.0 94.00 1.0 100.7 1.1
CS‐4 Si10Fe90 1300 24 20 9.87 0.3 2.4 0.2 89.45 0.2 101.7 0.2
CSM‐998 Si10Fe90 1500 8 11 4.46 0.4 3.4 0.5 92.83 0.7 100.7 0.3
CSM‐1002 Si10Fe90 1600 8 14 9.95 0.7 2.3 0.4 88.79 0.4 101.0 0.4
CSM‐1003 Si10Fe90 1700 8 13 9.32 0.5 2.4 0.3 87.86 0.3 99.6 0.6
CSM‐1121 Si10Fe90 1500 0.25 800 9 6 8.86 0.6 0.4 0.03 91.82 1.2 101.1 0.9
CSM‐1129 Si10Fe90 1500 0.25 900 9 21 8.10 0.6 0.6 1.2 91.75 1.5 100.5 1.0
CS‐5 Si22Fe78 1300 24 16 20.42 0.5 0.9 0.1 81.07 0.5 102.4 0.4
CSM‐984 Si22Fe78 1500 8 14 18.43 0.1 1.0 0.1 79.28 0.3 98.7 0.3
CSM‐996 Si22Fe78 1600 8 12 19.02 1.1 1.0 0.1 80.40 1.0 100.4 1.1
CSM‐995 Si22Fe78 1700 8 11 15.18 1.2 1.3 0.2 83.49 0.4 99.9 1.3
CSM‐1004 Si22Fe78 1500 4 13 19.57 0.4 1.1 0.2 79.17 1.1 99.8 1.3
CSM‐1005 Si22Fe78 1500 1 6 18.60 0.4 1.3 0.1 79.76 1.3 99.7 1.2
CSM‐1019 Si22Fe78 1500 11 6 17.47 0.4 1.5 0.5 80.81 1.5 99.8 1.3
CSM‐1122 Si22Fe78 1500 0.25 800 9 14 15.82 1.7 0.4 0.1 84.02 2.1 100.3 1.0
CSM‐1133 Si22Fe78 1500 0.25 900 9 13 18.55 1.8 0.5 0.04 81.93 2.0 100.9 0.7
CS‐2 Si35Fe65 1300 24 16 22.37 1.3 0.6 0.1 79.94 1.3 102.9 0.3
Note. Std = standard deviation; n = number of analyses included in average
10.1029/2019JE006239
Journal of Geophysical Research: Planets
VANDER KAADEN ET AL. 6of15
3.2. Low‐Temperature (800–900°C) Experiments
Each CCGS experiment with a dwell temperature at or below 900°C
resulted in a single solid metal phase. The broad beam analyses for the
low‐temperature experiments, representative of the average composition
at the experimental P‐Tconditions, contained ~5.0–18.6 wt% Si and
~84.0–94.2 wt% Fe, again indicative of the wide range in starting composi-
tions that were used in this study (Table 1). These broad beam analyses of
the lower temperature experiments show a range in C abundances from
0.4–1.7 wt% (Table 2; Figure 3).
4. Discussion
4.1. Compositional Effects on C Solubility in Fe‐Rich
Metallic Phases
4.1.1. Effect of Carbon on Melting in the Iron‐Silicon System
at 1 GPa
The addition of C into the Fe‐Si system lowers the melting temperature of
the system (Ahn et al., 2014). Based on the C‐free 1‐bar Fe‐Si phase dia-
gram, the runs at 1300°C in a Si
5
Fe
95
and Si
35
Fe
65
mixture are subsolidus
(Hansen, 1958). However, the quenched dendritic texture in these runs
(Figure 2a–2c) indicate these compositions were liquid at the P‐Tcondi-
tions of the experiments. Given that our experiments were at 1 GPa, this
observation indicates that the addition of C into the Fe‐Si system depresses the melting temperature such
that it out‐competes the effect of pressure to raise the melting temperature, indicating that C has a strong
effect on liquidus depression in the Fe‐Si‐C system at the conditions of core formation in celestial bodies
(e.g., Greiner et al., 1933; Hansen, 1958). This observation has important implications for the thermochemi-
cal evolution of Mercury's core, in particular the ratio of crystalline metal to molten metal over the
planet's history.
4.1.2. Distribution of Carbon Between Quenched Metallic Melt and
Crystalline Metallic Alloys
The experiments that were quenched directly from a superliquidus state
(i.e., experiments run at ≥1300°C) have more carbon than the experi-
ments that were held under subsolidus conditions before quenching
(i.e., experiments run at 800–900°C; Figure 3). These results are consistent
with previous studies on the partitioning of C between metallic melt and
crystalline metal (e.g, Chabot et al., 2017; Greiner et al., 1933;
Hansen, 1958, and references therein) and indicate that the solid‐liquid
metal ratio in the Mercurian core will likely play an important role in
its capacity to store C.
4.1.3. Solubility of Silicon and Carbon in Iron‐Rich Metallic Phases
There is a clear and consistent trend (Figure 3–5) that indicates an antic-
orrelation between Si abundance and C abundance in Fe‐rich metallic
melt and crystalline Fe‐rich metal alloys, an effect that has also been
observed and/or predicted in numerous other studies (e.g., Ahn et al., 2014;
Bouchard & Bale, 1995; Kawanishi et al., 2009; Li et al., 2015, 2016;
Shürmann & Kramer, 1969; Wade & Wood, 2005). Although evidenced
across the entire data set, this is much clearer when examining only the
experiments that contained a molten metallic phase at the time of quench
(Figure 3–5) rather than those that were subsolidus at the time of quench.
Given these results, under reducing conditions, where Si exhibits more
siderophile behavior and hence partitions into a metallic phase (e.g.,
Chabot et al., 2014), one should expect lower amounts of carbon to be pre-
sent than in a similarly Si‐poor, Fe‐rich metallic phase. This result is simi-
lar to a study by Chen et al. (2015) who examined C solubility in a molten
Figure 3. C‐Fe‐Si (molar) ternary. The color of the symbol corresponds to
temperature. All data (see Table 2) analyzed using a broad beam (~15–
20 μm).
Figure 4. Plot of mol% carbon versus mol% Si in the Fe‐Si‐C system at 0–
1 GPa to examine the effects of composition and temperature on CCGS.
All data on experiments quenched from a superliquidus state in the present
study (i.e., experiments run at 1300–1800°C) are included along with error
bars that represent 1σstandard deviations on the mean value from replicate
analyses of each experiment. Diamond symbols represent experimental
data at 0 GPa in the Fe‐Si‐C system from Shürmann and Kramer (1969).
Solid black lines represent computed solubilities of C in molten metallic
Fe‐Si at 0 GPa and 1800°C as well as at 1300°C based on the model of Wade
and Wood (2005).
10.1029/2019JE006239
Journal of Geophysical Research: Planets
VANDER KAADEN ET AL. 7of15
Fe‐Ni‐Si alloy for steel making purposes and found lower C contents with
higher Si contents. Although we did not have silicates present in our
experiments, we reexamined the metal present in some graphite‐saturated
experiments from Vander Kaaden and McCubbin (2016). These experi-
ments had pure Si‐metal as one of the main starting materials in order
to reduce the oxygen fugacity of the experimental charges to
Mercurian‐like conditions. EPMA analyses of these metals yielded unde-
tectable amounts of carbon, consistent with the results from this study.
The combination of our experimental results with those in the literature
shows that bulk composition has a substantial effect on CCGS in metallic
phases.
4.1.4. Solubility of Carbon in Iron‐Rich Planetary Cores
As described in Table 1, our experiments were conducted on a simplified
system, similar to that used in Knibbe and van Westrenen (2018) to model
the thermal evolution of Mercury's core. Specifically, we used the Fe‐Si‐C
system to evaluate the effect of temperature and Si on CCGS in Fe‐rich
metals and metallic melts. However, in nature, one would expect a more
chemically complex core with substantial amounts of Ni (i.e., several wt
%) and potentially low amounts of S. Dasgupta et al. (2013) investigated
Fe‐rich Ni‐bearing metallic melts at similar pressures to the experiments
in this study (i.e., 1 and 3 GPa) under lower oxygen fugacities than those
typical for the Earth and Mars (i.e., ~IW‐1.5 to IW‐1.9). Their 1600°C data
showed that at these pressures and oxygen fugacities, C solubility in
Fe‐rich metallic melts decreases with increasing Ni content, the same gen-
eral trend that is seen for Si in the present study. This observation is
further supported by Kim et al. (2014) who examined the solubility of carbon in liquid Fe that also contained
vanadium, molybdenum, and nickel. Kim et al. (2014) also showed that carbon solubility decreases with
increasing Ni content. Contrary to these results, Malavergne et al. (2014) conducted experiments at 1 GPa
from 1300–2000°C on an enstatite chondrite‐like composition and showed increasing carbon contents with
increasing Ni contents within the metallic phases of their experiments. However, these metals also con-
tained S, Pd, Mo, Cr, and V along with being Fe‐dominated. The experimental results from Malavergne
et al. (2014) support a decrease in C content (2.8 to 0.1 wt%) with increasing Si (6.04 to 23.66 wt%) in the
metal phases, complimentary to the results in this study. Zhang et al. (2018, and references therein) showed
a similar effect in the Fe‐Ni‐S system conducted from 2–7 GPa and 1200–1600°C. These observations coupled
with our own indicate that there are likely structural controls on the incorporation of C in metallic melts that
are consistent with a paucity of bonds between C and non‐Fe metallic melt components (Elardo &
Shahar, 2017; Hume‐Rothery, 1966). Although this is not a surprising result for Si given the high melting
temperatures in the Si‐C system (Hansen, 1958), C is fairly soluble in Ni metallic melts at low pressure with
a similar solubility as that exhibited in the Fe‐C system for a given temperature (Hansen, 1958). Nonetheless,
the results of our study and previous studies on the solubility of C in metallic melts indicates that C solubility
decreases with increasing substituents (of any element) for Fe in the metallic melt of planetary cores.
4.2. Temperature Effects on Carbon Solubility in Iron‐Silicon Metallic Phases
The solubility of C in molten metallic Fe increases as a function of increasing temperature by approximately
1 wt% C every 200°C over the range of 1300–2400°C at 0 GPa (e.g., Kawanishi et al., 2009; Wade &
Wood, 2005). At 2 GPa, an increase of ~1 wt% C in molten metallic Fe occurs every 250°C over the range
of 2000–2410°C (e.g., Dasgupta & Walker, 2008), indicating that the effect of temperature on C solubility
is subdued at higher pressure in the Fe‐C system. In contrast, previous experimental studies have reported
an increase in C solubility in molten metallic Fe‐Si alloy of ~1 wt% C every 390°C over the range of
1300°C to 1800°C at 0 GPa (Bouchard & Bale, 1995; Kawanishi et al., 2009; Shürmann & Kramer, 1969;
Wade & Wood, 2005), indicating that the effect of temperature on C solubility is diminished in the
Fe‐Si‐C system compared to the Fe‐C system. Furthermore, the temperature effect on C solubility dimin-
ished with increasing Si abundance in the alloy from an increase of 1 wt% C every 380°C at 10 mol% Si to
1 wt% C every 410°C at 30 mol% Si over the range of 1300°C to 1800°C at 0 GPa (Kawanishi et al., 2009;
Figure 5. Plot of mol% carbon versus mol% Si in the Fe‐Ni‐P‐Si‐S‐C system
at 1500–1600°C to examine the effects of composition and pressure on
CCGS. Only data on experiments quenched from a superliquidus state at
1500–1600°C from the present study are shown along with error bars that
represent 1σstandard deviations on the mean value from replicate analyses
of each experiment. Diamond symbols represent experimental data at
0 GPa in the Fe‐Si‐C system from Shürmann and Kramer (1969) and data
from the Fe‐P‐Si‐C system from Bouchard and Bale (1995). Square symbols
represent experimental data at 3 GPa from Li et al. (2015), and the error
bars on the square symbols are based on the errors reported by Li
et al. (2015).
10.1029/2019JE006239
Journal of Geophysical Research: Planets
VANDER KAADEN ET AL. 8of15
Shürmann & Kramer, 1969; Wade & Wood, 2005). Although the overall trends on the effect of temperature
on C solubility in molten Fe‐Si alloys are in agreement among previous studies at 0 GPa, the CCGS in Fe‐Si
alloys as a function of temperature and composition were not always in agreement (Ahn et al., 2014;
Figure 4). Our superliquidus experiments span a temperature range of 1300–1800°C at a pressure of
1 GPa, and we did not detect a measurable temperature dependence on the solubility of C in molten metallic
Fe‐Si alloy (Figures 3–4). Both the Fe:Si ratio of our system and the temperature range of our experiments
encompass the range of temperature‐composition conditions of the previous studies pertaining to 0 GPa
(Bouchard & Bale, 1995; Kawanishi et al., 2009; Shürmann & Kramer, 1969; Wade & Wood, 2005).
Consequently, we infer that either the temperature effect on C solubility was overestimated in previous stu-
dies or pressure diminishes the positive temperature dependence on C solubility in molten Fe‐Si alloys. In
support of the latter point, experimental data support minimal temperature dependence on C solubility in
the Fe‐Si‐S‐C system between 920–1700°C and 3.5–20 GPa (Deng et al., 2013). Although temperature likely
exerts some control over the C solubility of Si‐Fe bearing metals, composition plays a more substantial role
than temperature.
4.3. Pressure Effects on Carbon Solubility in Iron‐Silicon Metallic Phases
Previous studies that report the effect of pressure on the solubility of C in molten metallic Fe do not agree
and indicate either a slight increase in C solubility with pressure (Dasgupta et al., 2013; Siebert et al., 2011;
Wood, 1993) or a slight decrease in C solubility as a function of pressure (Dasgupta & Walker, 2008;
Nakajima et al., 2009). Previous studies on the ternary Fe‐Si‐C system have been evaluated at 0 GPa, which
we compare to our experiments conducted at 1 GPa to assess for a pressure effect on C solubility in molten
Fe‐Si alloys. However, given the disparities among the various 0 GPa data regarding the effect of temperature
on C solubility in Fe‐Si alloys (Ahn et al., 2014; Bouchard & Bale, 1995; Kawanishi et al., 2009; Shürmann &
Kramer, 1969; Wade & Wood, 2005) coupled with the near absence of a temperature effect in our 1 GPa
experiments (Figure 4), we cannot begin to assess for any isothermal pressure effect on C solubility in molten
Fe‐Si alloys. As noted in section 4.2, one possible explanation for our observation of no detectable effect of
temperature on C solubility in molten Fe‐Si alloys is that pressure causes a reduction in the intensity of
the temperature effect on C solubility, which was noted previously in the Fe‐C system (Dasgupta
& Walker, 2008).
Additional data on CCGS for more complex Si‐bearing metallic systems are available over a range of pres-
sures from 0–3 GPa over a narrow temperature range of 1500–1600°C (Bouchard & Bale, 1995; Dasgupta
& Walker, 2008; Li et al., 2015; Shürmann & Kramer, 1969), the data from which are compiled along with
our 1 GPa data over the same range of temperature (Figure 5). These systems represent a wide range of ele-
ments that may comprise planetary cores, including Fe, Ni, Si, P, S, and C. Over the range of 0–3 GPa, we
observe a slight positive pressure dependence on C solubility in multicomponent Fe‐rich, Si‐bearing molten
alloys with up to ~10 mol% Si (Figure 5); however, the limited data for systems with more than 10 mol% Si do
not exhibit a discernable pressure effect up to 3 GPa (Figure 5). The effects of pressure and temperature are
both subordinate to the effect of alloy composition on CCGS, particularly the abundance of Si.
4.4. Constraints on the Carbon and Silicon Content of Mercury's Core
We do not currently know the bulk carbon content of Mercury nor the bulk composition of its core. However,
we can use several observations to place constraints on these unknowns using data from this study as well as
from previous studies. The presence of a graphite flotation crust on Mercury (Vander Kaaden &
McCubbin, 2015) would likely require that Mercury's core is saturated in carbon (at least at the
core‐mantle boundary). Furthermore, the surface composition measured by the MESSENGER spacecraft
indicates that the planet formed under highly reducing conditions (McCubbin et al., 2012; McCubbin
et al., 2017; Zolotov et al., 2013). The geochemical behavior of elements under these highly reducing condi-
tions will differ from what is observed under more oxidizing conditions, like Earth (Vander Kaaden &
McCubbin, 2016). At conditions as reducing as IW‐3toIW‐7, Si is expected to be a minor or major component
of the core of the planet in the range of <1–25 wt% (Chabot et al., 2014; Hauck II et al., 2013; Margot
et al., 2019; Nittler et al., 2019). Based on the experimental results in the present study, the more Si that parti-
tions into an Fe‐rich planetary core, the less carbon that can be stored in that core. Over this range of Si abun-
dances in the core of Mercury, our results indicate that the carbon content of Mercury's core is in the range of
approximately 0.5–6.4 wt% C, based on a solid inner core radius of 960 ± 430 km (Genova et al., 2019), a core
10.1029/2019JE006239
Journal of Geophysical Research: Planets
VANDER KAADEN ET AL. 9of15
radius of ~2,024 km (Margot et al., 2019), and the presumption that pressure and temperature have minimal
effects on C solubility in the Mercurian core, which is supported by literature data, possibly up to 20 GPa
(Bouchard & Bale, 1995; Dasgupta & Walker, 2008; Deng et al., 2013; Li et al., 2015; Shürmann &
Kramer, 1969; Figure 5). Considering the estimated mass fraction of Mercury that is composed of the core
(i.e., 0.712 to 0.775, computed based on the range of core density from 6,700–7,260 kg/m
3
, range of mantle
density from 2,800–3,600 kg/m
3
, the planet radius of ~2,440 km, and the core radius of ~2,024 km; Margot
et al., 2019), this range in C abundance would require that the bulk planet has 0.3–5.0 wt% C without
including any contributions from a graphite flotation crust. However, the solid‐liquid ratio in the core will
have an effect on this estimate. Specifically, the carbon content of the bulk core at carbon saturation,
hence the minimum bulk planet C abundances, would decrease with increasing solid‐liquid ratios. Our
results indicate the C saturation values in the core would be lower by a factor of 2–4 for a fully solidified core.
To further assess the rationality of the estimated range of carbon contents in a carbon‐saturated Mercurian
core, we use the carbon contents of various chondrite meteorite groups, thought to represent planetary
building blocks (Lodders & Fegley, 1998), to estimate whether Mercury would need superchondritic or sub-
chondritic abundances of carbon to be C‐saturated. In particular, we compute the Si abundance that would
be required in an Fe‐rich metallic melt to saturate the core in carbon assuming bulk Mercury has a chondri-
tic abundance of C. For these computations, we include CI chondrites, which best approximate the compo-
sition of the solar photosphere, as well as enstatite chondrites and CB meteorites given their broad
geochemical similarities with present‐day Mercury (Brown & Elkins‐Tanton, 2009; Nittler et al., 2019;
Taylor & Scott, 2004).
CI chondrites have an average of 3.45 wt% C (Lodders & Fegley, 1998), but they span a range of 2–5 wt% C
(Grady & Wright, 2003). If we assume the bulk C abundance in Mercury is equivalent to that of average CI
chondrites, based on the mass ratio of 0.712–0.775 for the core of Mercury relative to the total mass of
Figure 6. Bulk C/Si (wt.) versus bulk Fe/Si (wt.) for Earth, Mars, Mercury, and numerous chondrite types. The Mercury
range was computed based on (1) the estimated range in mass fraction of core to bulk planet Mercury (i.e., 0.712–0.775;
computed from data in Margot et al., 2019), (2) the mass ratio of liquid core to solid core (0.658–0.983; computed from
data in Margot et al., 2019 and Genova et al., 2019), (3) our experimental results on the C solubility in molten Fe‐Si
metal and solid Fe‐Si metal with C solubility in Si‐free molten metallic Fe set at 6.5 wt% based on data in Dasgupta and
Walker (2008), (4) an average mantle Si abundance of 25 wt% (after Chabot et al. 2013; Nittler et al., 2019) and an average
mantle Fe abundance of 0.02–1.5 wt% (McCubbin et al., 2017; Nittler et al., 2019), and (5) an Fe/Ni mass ratio in the core
of 17 (Lodders & Fegley, 1998). The dashed line represents the dividing line between non‐CB chondritic and
superchondritic and/or CB‐like Fe/Si ratios. The core Si abundance at the minimum non‐CB chondritic Fe/Si ratio for
bulk Mercury is indicated at 27 wt% Si, and the range in CB chondritic Fe/Si ratios that are allowable for Mercury
are also indicated in the range of 5–18.5 wt% Si. Data for C, Si, and Fe in bulk Earth, Mars, and non‐CB chondrites are
from Grady and Wright (2003) and Lodders and Fegley (1998). Data from CB chondrites are from Ivanova et al. (2008),
Lauretta et al. (2009), Rubin et al. (2003), and Weisberg et al. (1990).
10.1029/2019JE006239
Journal of Geophysical Research: Planets
VANDER KAADEN ET AL. 10 of 15
Mercury, at least 0.3–3 wt% Si would be needed in the core in order to be saturated in C and have chon-
dritic to subchondritic abundances of C. This model represents a lower limit on the bulk Si content of
Mercury's core, given that the Earth is depleted in C relative to CI chondrite by a factor of 47
(McDonough, 2003), and any planetary depletion in C relative to CI would require higher Si abundances
in the core for the core to be saturated in C. However, carbon becomes less volatile under highly redu-
cing conditions owing to the expansion of the graphite stability field and the stability fields of various
carbides (e.g., Wood, 1993), so Mercury could have been more efficient at retaining C than the Earth.
In contrast, the Fe/Si ratio of CI chondrite only matches the bulk composition of Mercury at core Si
abundances in the range of 29–31 wt% Si (Figure 6), which is above the previous upper estimates of
the Si content of Mercury's core based on thermal, density, moment of inertia, and experimental con-
straints (e.g., Chabot et al., 2014; Hauck II et al., 2013).
If Mercury were composed of CB chondrite building blocks, which has been proposed given their reduced
nature and elevated metal abundances (Brown & Elkins‐Tanton, 2009; Taylor & Scott, 2004), and the carbon
contents of these meteorites represented the bulk C of Mercury (0.35–1.21 wt% C; Grady & Wright, 2003), it
would require 17–28 wt% Si in the core of Mercury to saturate the core in C. However, the upper end of this
range is higher than the highest estimates on the Si content of Mercury's core based on thermal, density,
moment of inertia, and experimental constraints (e.g., Chabot et al., 2014; Hauck II et al., 2013).
Interestingly, the estimated bulk Fe, Si, and C abundances of Mercury match CB chondritic proportions
at Si abundances in the core ranging from 5–18.5 wt% (Figure 6). These Si abundances correspond to carbon
abundances in the core of 0.8–4.0 wt% C (Figure 6).
Finally, if Mercury were composed of EH and EL chondrite building blocks, which has been proposed
given their highly reduced nature and chemical similarities to Mercury (e.g., Wasson, 1988), and the car-
bon contents of these meteorites represent the bulk C of Mercury (0.15–0.7 wt% C; Grady &
Wright, 2003), it would require ~16.5–36 wt% Si in the core of Mercury to saturate the core in C.
However, similar to the CB scenario, much of this range is above the previous upper estimates of the
Si content of Mercury's core based on thermal, density, moment of inertia, and experimental constraints
(e.g., Chabot et al., 2014; Hauck II et al., 2013). Furthermore, the estimated bulk Fe, Si, and C abun-
dances of Mercury match EH and EL chondritic proportions at Si abundances in the core ranging from
27–29 wt% and 36–38 wt%, respectively (Figure 6). These Si abundances correspond to carbon abun-
dances in the core of 0.3–0.5 and ~0.20 wt%, respectively (Figure 6).
The previous chondritic models provide some constraints on the Si abundance of a carbon‐saturated
Mercurian core, and these models indicate that at least 0.3 wt% Si is needed in the Mercurian core if
it is saturated in carbon and the bulk planet has chondritic‐subchondritic abundances of C. However,
much higher abundances of Si, in the range of 27–39 wt% Si, are needed in the core of Mercury to main-
tain a non‐CB chondritic bulk planet Fe/Si ratio (Figure 6), and a range of 5–18.5 wt% Si is permitted in
the core for a CB chondritic Fe/Si ratio (Figure 6). These elevated abundances of Si in the core seem dif-
ficult to reconcile with estimates of a bulk silicate portion of Mercury that has 0.94–2.19 wt% FeO
(McCubbin et al., 2017; Vander Kaaden et al., 2017), indicating that either the FeO abundances in
Mercury's mantle are overestimated (cf., Nittler et al., 2019) or the bulk composition of Mercury has a
superchondritic Fe/Si ratio. A superchondritic Fe/Si ratio of Mercury could be consistent with a giant
impact model for Mercury (Benz et al., 1988), which would have enriched bulk Mercury in siderophile
elements relative to chondrite.
With respect to the estimated abundances of carbon in Mercury's core, the chondritic models all assume
a simplified core composition in the Fe‐Ni‐Si‐C system with a chondritic Fe/Ni ratio of 17 (Lodders &
Fegley, 1998); however, the core of Mercury is likely to have additional major or minor components,
in particular, S. Previous studies have demonstrated that the addition of S and Ni to the Fe‐Si‐C system
will further diminish the solubility of C (Dasgupta et al., 2013; Kim et al., 2014; Li et al., 2015; Zhang
et al., 2018). Furthermore, we do not have sufficient evidence to rule out the possibility that Mercury
has a superchondritic abundance of C, which could have been attained contemporaneously with a super-
chondritic Fe/Si resulting from a giant impact. The possible attainment of both superchondritic Fe/Si and
C abundance is supported by C partitioning relationships between silicate melt and Fe‐rich metal (e.g.,
Dasgupta et al., 2013; Li et al., 2015, 2017). Additional experiments in the Fe‐Ni‐S‐C‐Si system under
10.1029/2019JE006239
Journal of Geophysical Research: Planets
VANDER KAADEN ET AL. 11 of 15
Mercurian conditions will be valuable in further constraining the composition, physical properties, and
thermal history of Mercury's core.
4.5. Assessment of Mercury's Core Composition
Our experiments did not include Ni, and our calculations in section 4.4 did not take into account S; however,
some constraints can be placed on these elements based on previous studies. For the computations in
section 4.4, we assume that Mercury has a chondritic bulk composition with respect to the Fe/Ni abundance
ratio of Mercury's core. The Fe/Ni ratio is likely to be chondritic given that Fe and Ni are both highly side-
rophile at the low oxygen fugacity under which the planet differentiated, which is orders of magnitude below
both the IW and nickel‐nickel oxide (NNO) buffers. Consequently, we infer that Mercury's core has an Fe/Ni
weight ratio of 17 (Lodders & Fegley, 1998). Li et al. (2015) showed that the addition of S in an Fe‐rich metal
alloy does not have a considerable effect on carbon solubility in the alloy melt, although Zhang et al. (2018,
and references therein) showed that C solubility diminishes with increasing S content, similar to the results
of Bouchard and Bale (1995). Furthermore, Chabot et al. (2014) showed that even up to 5 wt% S in the core of
Mercury allows for more than 8 wt% Si in the metal. Therefore, given the solubility relationships in metal
between Fe‐Ni‐S‐C‐Si and the presence of a graphite flotation crust on Mercury, we suggest the core of
Mercury is likely composed of Fe‐Ni in chondritic proportions and some combination of S‐Si‐C in lower
abundances (≤5 wt% each for S and C and up to 25 wt% Si). The Si content of the core remains the least con-
strained in the presumed Fe‐Ni‐S‐C‐Si system. The upper end of estimated FeO abundances in the mantle of
Mercury are 1.0–2.2 wt% FeO, which is consistent with a core that has <1 wt% Si and would imply that bulk
Mercury has a superchondritic Fe/Si ratio. In contrast, the core requires ≥27 wt% Si to have a non‐CB chon-
dritic Fe/Si ratio and 5–18.5 wt% Si for a CB chondritic Fe/Si ratio. These Si abundances are consistent with a
Mercurian mantle with ≤0.1 wt% FeO, consistent with bulk mantle FeO abundances estimated by Nittler
et al. (2019). Whether or not Mercury has a chondritic or superchondritic Fe/Si ratio has potentially impor-
tant implications for the origin of Mercury, specifically the building block materials for the planet and
whether or not a portion of the mantle was removed by impact. Given the relatively circular orbit of the
spacecraft and higher resolution of data from the southern hemisphere of the planet, additional constraints
on the C and Si abundance in Mercury from the BepiColombo spacecraft may allow tighter constraints to be
placed on these values in the future (Benkhoff et al., 2010).
5. Conclusion
The current study aimed to examine the role of C in the Mercurian core. Our experimental results, combined
with additional experiments on C solubility in Fe‐rich metallic systems from the literature, suggest the core
of Mercury is likely composed of Fe‐Ni in chondritic proportions with minor abundances of C and S. Si abun-
dances are largely unconstrained but likely in the range of <1–25 wt%. The evidence of a graphite flotation
crust on Mercury (e.g., Klima et al., 2018; Murchie et al., 2015; Peplowski et al., 2016; Vander Kaaden &
McCubbin, 2015) implies that the core is saturated in C, with estimated abundances around 0.4–0.5 wt%
C and 27 wt% Si for a core with EH chondrite Fe/Si and C/Si (Figure 6) or 0.8–4.0 wt% C and 5–18.5 wt%
Si for a core with CB chondrite Fe/Si and C/Si (Figure 6). In contrast, our data indicate that a Si‐free core
yields a superchondritic bulk Mercury Fe/Si and a maximum C abundance in the core of 6.4 wt% C for an
entirely molten Fe core. The highly reducing nature of Mercury as well as its enrichment in volatiles
(Evans et al., 2015; Nittler et al., 2011; Peplowski et al., 2011) has led to an extremely complex and exotic
thermal and magmatic evolution, with a wide array of possible light elements in its core. Understanding
the abundances and distributions of all volatiles present on Mercury is imperative to unlocking its origin
and geologic history.
Appendix A: Piston Cylinder Pressure Calibration
Given the changes that have been made to our 13 mm press since the calibration of Filiberto et al. (2008), an
updated calibration is presented here. Pressure calibration of the barium carbonate cell assembly was con-
ducted using the melting point of optical grade sodium chloride (NaCl) (Clark, 1959; Pistorius, 1966) pur-
chased from Alfa Aesar. The NaCl was ground to a fine‐grained powder and kept in an oven at ~100°C
until the time of each experiment. The melting Tof NaCl was bracketed using a variation of the falling
sphere method. The bulk of the molybdenum capsule was first filled with natural NaCl (ρ≈2.16 g/cm
3
)
10.1029/2019JE006239
Journal of Geophysical Research: Planets
VANDER KAADEN ET AL. 12 of 15
powder. On top of this, two Fo
90
(ρ≈3.31 g/cm
3
) spheres, a few hundreds of microns in diameter, were
placed with careful precision so they were in the center of the capsule and not touching the capsule walls.
The remainder of the capsule was then filled with a thin layer of ground NaCl powder to make sure the
spheres were not touching the molybdenum lid. An experimental assembly identical to those described in
the section 2.2 of this manuscript was employed. For each experiment, the temperature was initially
raised to 800°C at 300°C/min, then to 50°C below the target temperature at 100°C/min, and finally to the
target temperature at 50°C/min. Once at target PT conditions, each experiment was held for ~15 min.
Experiments were quenched at ~60°C/s. Cross sections of the quenched capsules were polished dry using
boron nitride powder as a lubricant. If the spheres were still at the top of the cell assembly after quench,
the NaCl was assumed not to have melted. If the spheres were at the bottom of the assembly at the end of
the experiment, the NaCl was assumed to have melted. The updated calibration is given in Figure A1.
References
Agee, C. B. (1998). Crystal‐liquid density inversions in terrestrial and lunar magmas. Physics of the Earth and Planetary Interiors,107(1–3),
63–74.
Agee, C. B., & Walker, D. (1993). Olivine flotation in mantle melt. Earth and Planetary Science Letters,114(2–3), 315–324.
Ahn, S. H., Kim, Y. H., Shin, J.‐P., & Lee, Y. E. (2014). Thermodynamic assessment of liquid Fe‐Si‐C systems by unified interaction para-
meter model. ISIJ International,54, 750–755.
Ardia, P., Hirschmann, M. M., Withers, A. C., & Stanley, B. D. (2013). Solubility of CH
4
in a synthetic basaltic melt, with applications to
atmosphere‐magma ocean‐core partitioning of volatiles and to the evolution of the Martian atmosphere. Geochimica et Cosmochimica
Acta,114,52–71.
Benkhoff, J., van Casteren, J., Hayakawa, H., Fujimoto, M., Laakso, H., Novara, M., et al. (2010). BepiColombo‐Compre hensive exploration
of Mercury: Mission overview and science goals. Planetary and Space Science,58,2–20.
Benz, W., Slattery, W. L., & Cameron, A. G. W. (1988). Collisional stripping of Mercury's mantle. Icarus,74, 516–528.
Bertka, C. M., & Fei, Y. (1997). Mineralogy of the Martian interior up to the core‐mantle boundary pressures. Journal of Geophysical
Research,102, 5251–5264.
Bouchard, D., & Bale, C. (1995). Simultaneous optimization of thermochemica l data for liquid iron alloys containing C, N, Ti, Si, Mn, S, and
P. Metallurgical and Materials Transactions B: Process Metallurgy and Materials Processing Science,26, 467–484.
Brown, S. M., & Elkins‐Tanton, L. T. (2009). Compositions of Mercury's earliest crust from magma ocean models. Earth and Planetary
Science Letters,286, 446–455.
Chabot, N. L., Campbell, A. J., Jones, J. H., Humayun, M., & Vern Lauer, H. Jr. (2006). The influence of carbon on trace element parti-
tioning behavior. Geochimica et Cosmochimica Acta,70, 1322–1335.
Chabot, N. L., Wollack, E. A., Klima, R. L., & Minitti, M. E. (2014). Experimental constraints on Mercury's core composition. Earth and
Planetary Science Letters,390, 199–208.
Chabot, N. L., Wollack, E. A., McDonough, W. F., Ash, R. D., & Saslow, S. A. (2017). Experimental determination of partitioning in the
Fe‐Ni system for applications to modeling meteoritic metals. Meteoritics & Planetary Science,52(6), 1133–1145. https://doi.org/10.1111/
maps.12864
Chen, B., Li, J., & Hauck, S. A. (2008). Non‐ideal liquidus curve in the Fe‐S system and Mercury's snowing core. Geophysical Research
Letters,35(7), L07201.
Chen, P. X., Chu, S. J., & Han, P. W. (2015). The solubility of carbon in molten iron‐nickel‐silicon alloy at 1773 K and 1823 K (pp. 528–535).
Theory: The Fourteenth International Ferroalloys Congress Fundamentals.
Clark, J. S. P. (1959). Effect of pressure on the melting points of eight alkali halides. The Journal of Chemical Physics,31, 1526–1531.
Figure A1. Updated calibration for the 13 mm piston cylinder (Q) at JSC constructed using the melting point of NaCl.
10.1029/2019JE006239
Journal of Geophysical Research: Planets
VANDER KAADEN ET AL. 13 of 15
Acknowledgments
All data used in this manuscript,
including individual EPMA analyses
and BSE images of run products, are
publicly available in Harvard Dataverse
(dataverse.harvard.edu at https://doi.
org/10.7910/DVN/KAURFB). The
manuscript itself will be made publicly
available at NASA PubSpace (https://
www.ncbi.nlm.nih.gov/pmc/funder/
nasa/). The authors thank Lisa
Danielson, Kellye Pando, and Jenny
Rapp for helpful discussions and
assisting in laboratory procedures as
well as Megan Mouser and Justin
Reppart for helping with calibration of
the piston cylinder. We thank Ian
Szumila, Zia Rahman, and Lindsay
Keller for helping with synthesizing
and characterizing the SiC standard.
We also thank the MESSENGER
Science Team, with special thanks to
the MESSENGER Geochemistry
Discipline Group, specifically Nancy
Chabot, for fruitful discussions
regarding the interpretation of
MESSENGER data and run products.
The authors are appreciative of
numerous anonymous reviewers for
their helpful comments and suggestions
on previous versions of this manuscript
as well as the efforts of the Editors for
their review and handling of this
manuscript. Support for this research
was provided in part by NASA's
planetary science research program as
well as by NASA's Solar System
Workings Program. This work was also
supported by NASA Headquarters
under the NASA Earth and Space
Science Fellowship Program‐Grant
NNX15AQ80H awarded to K. E. V. K.
Dasgupta, R., Chi, H., Shimizu, N., Buono, A. S., & Walker, D. (2013). Carbon solution and partitioning between metallic and silicate melts
in a shallow magma ocean: Implications for the origin and distribution of terrestrial carbon. Geochimica et Cosmochimica Acta,102,
191–212.
Dasgupta, R., & Walker, D. (2008). Carbon solubility in core melts in a shallow magma ocean environment and distribution of carbon
between the Earth's core and the mantle. Geochimica et Cosmochimica Acta,72, 4627–4641.
Deng, L. W., Fei, Y. W., Liu, X., Gong, Z. Z., & Shahar, A. (2013). Effect of carbon, sulfur and silicon on iron melting at high pressure:
Implications for composition and evolution of the planetary terrestrial cores. Geochimica et Cosmochimica Acta,114,
220–233.
Elardo, S. M., & Shahar, A. (2017). Non‐chondritic iron isotope ratios in planetary mantles as a result of core formation. Nature Geoscience,
10, 317–323.
Evans, L. G., Peplowski, P. N., McCubbin, F. M., McCoy, T. J., Nittler, L. R., Zolotov, M. Y., et al. (2015). Chlorine on the surface of Mercury:
MESSENGER gamma‐ray measurements and implications for the planet's formation and evolution. Icarus,257, 417–427.
Filiberto, J., Treiman, A. H., & Le, L. (2008). Crystallization experiments on a Gusev Adirondack basalt composition. Meteoritics &
Planetary Science,43,1137–1146.
Genova, A., Goossens, S., Mazarico, E., Lemoine, F. G., Neumann, G. A., Kuang, W., et al. (2019). Geodetic evidence that Mercury has a
solid inner core. Geophysical Research Letters,46,3625–3633.
Grady, M. M., & Wright, I. P. (2003). Elemental and isotopic abundances of carbon and nitrogen in meteorites. Space Science Reviews,106,
231–248.
Greiner, E. S., Marsh, J. S., & Stoughton, B. (1933). The alloys of iron and silicon. New York: McGraw‐Hill Book Co.
Hansen, M. (1958). Consitution of binary alloys. McGraw‐Hill Book Company, New York,1305.
Hauck, S. A., & Johnson, C. L. (2019). Mercury: Inside the iron planet. Elements,15,21–26.
Hauck, S. A. II, Margot, J. L., Solomon, S. C., Phillips, R. J., Johnson, C. L., Lemoine, F. G., et al. (2013). The curious case of Mercury's
internal structure. Journal of Geophysical Research ‐Planets,118, 1204–1220.
Hauck, S. A. II, Solomon, S. C., & Smith, D. A. (2007). Predicted recovery of Mercury's internal structure by MESSENGER. Geophysical
Research Letters,34, L18201.
Hirschmann, M. M., Withers, A. C., Ardia, P., & Foley, N. T. (2012). Solubility of molecular hydrogen in silicate melts and consequences for
volatile evolution of terrestrial planets. Earth and Planetary Science Letters,345,38–48.
Hume‐Rothery, W. (1966). The structures of alloys of iron. An elementary introduction. In H. M. Finniston, D. W. Hopkins, & W. S. Owene
(Eds.), Pergamon. eBook ISBN: 9781483137285.
Ivanova, M. A., Kononkova, N. N., Krot, A. N., Greenwood, R. C., Franch, I. A., Verchovsky, A. B., et al. (2008). The Isheyevo meteorite:
Mineralogy, petrology, bulk chemistry, oxygen, nitrogen, carbon isotopic compositions, and Ar‐40‐Ar‐39 ages. Meteoritics & Planetary
Science,43, 915–940.
Kawanishi, S., Yoshikawa, T., & Tanaka, T. (2009). Equilibrium phase relationship between SiC and a liquid phase in the Fe‐Si‐C system at
1523–1723K. Materials Transactions,50,806–813.
Kim, D.‐H., Paek, M.‐K., Kim, T.‐J., Won, S.‐Y., & Pak, J.‐J. (2014). Carbon solubility in liquid iron containing V, Mo, and Ni. Materials
Transactions,55,610–615.
Klima, R. L., Denevi, B. W., Ernst, C. M., Murchie, S. L., & Peplowski, P. N. (2018). Global distribution and spectral properties of
low‐reflectance material on Mercury. Geophysical Research Letters,45, 2945–2953.
Knibbe, J. S., & van Westrenen, W. (2018). The thermal evolution of Mercury's Fe‐Si core. Earth and Planetary Science Letters,482, 147–159.
Lauretta, D. S., Goreva, J. S., Hill, D. H., Killgore, M., La Blue, A. R., Campbell, A., et al. (2009). The Fountain Hills unique CB chondrite:
Insights into thermal processes on the CB parent body. Meteoritics & Planetary Science,44, 823–838.
Li, Y., Dasgupta, R., & Tsuno, K. (2015). The effects of sulfur, silicon, water, and oxygen fugacity on carbon solubility and partitioning in
Fe‐rich alloy and silicate melt systems at 3 GPa and 1600°C: Implications for core‐mantle differentiation and degassing of magma oceans
and reduced planetary mantles. Earth and Planetary Science Letters,415,54–66.
Li, Y., Dasgupta, R., & Tsuno, K. (2017). Carbon contents in reduced basalts at graphite saturation: Implications for the degassing of Mars,
Mercury, and the Moon. Journal of Geophysical Research: Planets,122, 1300–1320.
Li, Y., Dasgupta, R., Tsuno, K., Monteleone, B., & Shimizu, N. (2016). Carbon and sulfur budget of the silicate Earth explained by accretion
of differentiated planetary embryos. Nature Geoscience,9, 781–788.
Lodders, K., & Fegley, B. (1998). The planetary scientist's companion. New York: Oxford University Press.
Malavergne, V., Cordier, P., Righter, K., Brunet, F., Zanda, B., Addad, A., et al. (2014). How Mercury can be the most reduced terrestrial
planet and still store iron in its mantle. Earth and Planetary Science Letters,394, 186–197.
Malavergne, V., Toplis, M. J., Berthet, S., & Jones, J. (2010). Highly reducing conditions during core formation on Mercury: Implications for
internal structure and the origin of a magnetic field. Icarus,206(1), 199–209.
Margot, J.‐L., Hauck II, S.A., Mazarico, E., Padovan, S., Peale, S.J., 2019. Mercury's internal structure. In “Mercury: The view after
MESSENGER”edited by Solomon, Nittler, and Anderson, 85‐113.
McCubbin, F. M., & Barnes, J. J. (2019). Origin and abundances of H
2
O in the terrestrial planets, Moon, and asteroids. Earth and Planetary
Science Letters,526, 115,771.
McCubbin, F. M., Kaaden, K. E. V., Peplowski, P. N., Bell, A. S., Nittler, L. R., Boyce, J. W., et al. (2017). A low O/Si ratio on the surface of
Mercury: Evidence for silicon smelting? Journal of Geophysical Research: Planets,122, 2053–2076.
McCubbin, F. M., Riner, M. A., Vander Kaaden, K. E., & Burkemper, L. K. (2012). Is Mercury a volatile‐rich planet? Geophysical Research
Letters,39.
McCubbin, F. M., Vander Kaaden, K. E., Tartese, R., Boyce, J. W., Mikhail, S., Whitson, E. S., et al. (2015). Experimental investigation of F,
Cl, and OH partitioning between apatite and Fe‐rich basaltic melt at 1.0–1.2 GPa and 950–1000°C. American Mineralogist,100,
1790–1802.
McDonough, W.F., 2003. Compositional Model for the Earth's core, in: Holland, H.D., Turekian, K.K. (Eds.), Treatise on Geochemistry
Volume 2: The Mantle and Core. Elsevier, pp. 547–568.
Misawa, K. (2004). The Yamato 980459 olivine‐phyric shergottite consortium. Antarct Meteorite Res.,17,1–12.
Murchie, S. L., Klima, R. L., Denevi, B. W., Ernst, C. M., Keller, M. R., Domingue, D. L., et al. (2015). Orbital multispectral mapping of
Mercury using the MESSENGER Mercury dual imaging system: Evidence for the origins of plains units and low‐reflectance material.
Icarus,254, 287–305.
Murthy, V. M., van Westrenen, W., & Fei, Y. W. (2003). Experimental evidence that potassium is a substantial radioactive heat source in
planetary cores. Nature,423(6936), 163–165. https://doi.org/10.1038/nature01560
10.1029/2019JE006239
Journal of Geophysical Research: Planets
VANDER KAADEN ET AL. 14 of 15
Nakajima, Y., Takahashi, E., Suzuki, T., & Funakoshi, K. (2009). “Carbon in the core”revisited. Physics of the Earth and Planetary Interiors,
174, 202–211.
Nittler, L.R., Chabot, N.L., Grove, T.L., Peplowski, P.N., 2019. The chemical composition of Mercury. In “Mercury: The view after
MESSENGER”edited by Solomon, Nittler, and Anderson, 30‐51.
Nittler, L. R., Starr, R. D., Weider, S. Z., McCoy, T. J., Boynton, W. V., Ebel, D. S., et al. (2011). The major‐element composition of Mercury's
surface from MESSENGER X‐ray spectrometry. Science,333,1847–1850.
Peplowski, P. N., Evans, L. G., Hauck, S. A., McCoy, T. J., Boynton, W. V., Gillis‐Davis, J. J., et al. (2011). Radioactive elements on Mercury's
surface from MESSENGER: Implications for the Planet's formation and evolution. Science,333, 1850–1852.
Peplowski, P. N., Klima, R. L., Lawrence, D. J., Ernst, C. M., Denevi, B. W., Frank, E. A., et al. (2016). Remote sensing evidence for an
ancient carbon‐bearing crust on Mercury. Nature Geoscience,9, 273–276.
Peplowski, P. N., Lawrence, D. J., Evans, L. G., Klima, R. L., Blewett, D. T., Goldsten, J. O., et al. (2015). Constraints on the abundance of
carbon in near‐surface materials on Mercury: Results from the MESSENGER gamma‐ray spectrometer. Planetary and Space Science,
108,98–107.
Pistorius, C. W. F. T. (1966). Effect of pressure on the melting points of the sodium halides. The Journal of Chemical Physics,45, 3513–3519.
Predel, B. (1995). Fe‐Si (iron‐silicon). In O. Madelung (Ed.), Dy‐Er‐Fr‐Mo. Landolt‐Börnstein—Group IV physical chemistry 5e.
Righter, K., & Drake, M. J. (2000). Metal/silicate equilibrium in the early Earth—New constraints from the volatile moderately siderophile
elements Ga, Cu, P, and Sn. Geochimica et Cosmochimica Acta,64,3581–3597.
Righter, K., Go, B. M., Pando, K. A., Danielson, L. R., Ross, D. K., Rahman, Z., & Keller, L. P. (2017). Phase equilibria of a low S and C lunar
core: Implications for an early lunar dynamo and physical state of the current core. Earth and Planetary Science Letters,463, 323–332.
Riner, M. A., Bina, C. R., Robinson, M. S., & Desch, S. J. (2008). Internal structure of Mercury: Implications of a molten core. Journal of
Geophysical Research‐Planets,113.
Robaut, F., Crisci, A., Durand‐Charre, M., & Jouanne, D. (2006). Practical aspects of carbon content determination in carburized steels by
EPMA. Microscopy and Microanalysis,12(4), 331–334. https://doi.org/10.1017/S1431927606060466
Rubin, A. E., Kallemeyn, G. W., Wasson, J. T., Clayton, R. N., Mayeda, T. K., Grady, M., et al. (2003). Formation of metal and silicate
globules in Gujba: A new Bencubbin‐like meteorite fall. Geochimica et Cosmochimica Acta,67, 3283–3298.
Shürmann, E., & Kramer, D. (1969). Giessereiforschung,21,33–47.
Siebert, J., Corgne, A., & Ryerson, F. J. (2011). Systematics of metal‐silicate partitioning for many siderophile elements applied to Earth's
core formation. Geochimica et Cosmochimica Acta,75, 1451–1489.
Stevenson, D. J., Spohn, T., & Schubert, G. (1983). Magnetism and thermal evolution of the terrestrial planets. Icarus,54, 466–489.
Taylor, G. J., & Scott, E. R. D. (2004). Mercury. In A. M. Davis (Ed.), Treatise on geochemistry: Meteorites, (pp. 477–485). Amsterdam, The
Netherlands: Comets and Planets. Elsevier.
Tibbetts, G. G. (1980). Diffusivity of carbon in iron and steels at high temperatures. Journal of Applied Physics,51, 4813.
Vander Kaaden, K. E., Agee, C. B., & McCubbin, F. M. (2015). Density and compressibility of the molten lunar picritic glasses: Impli cations
for the roles of Ti and Fe in the structures of silicate melts. Geochimica et Cosmochimica Acta,149,1–20.
Vander Kaaden, K. E., & McCubbin, F. M. (2015). Exotic crust formation on Mercury: Consequences of a shallow, FeO‐poor mantle.
Journal of Geophysical Research: Planets,120, 195–209.
Vander Kaaden, K. E., & McCubbin, F. M. (2016). The origin of boninites on Mercury: An experimental study of the northern volcanic
plains lavas. Geochimica et Cosmochimica Acta,173, 246–263.
Vander Kaaden, K. E., McCubbin, F. M., Nittler, L. R., Peplowski, P. N., Weider, S. Z., Frank, E. A., & McCoy, T. J. (2017). Geochemistry,
mineralogy, and petrology of boninitic and komatiitic rocks on the Mercurian surface: Insights into the Mercurian mantle. Icarus,285,
155–168.
Wade, J., & Wood, B. J. (2005). Core formation and the oxidation state of the Earth. Earth and Planetary Science Letters,236,78–95.
Warren, P. H. et al. (1996). Compositional‐petrologic investigations of Eucrites and the QUE94201 Shergottie, paper presented at Proc.
NIPR Sym. Antarctic Meteorites, Nat. Inst. Polar Res., Tokyo.
Wasson, J. T. (1988). The building stones of the planets. In C. R. C. Vilas & M. S. Matthews (Eds.), F, (pp. 622–650). Mercury: University of
Arizona Press, Tucson.
Weider, S. Z., Nittler, L. R., Starr, R. D., McCoy, T. J., Stockstill‐Cahill, K. R., Byrne, P. K., et al. (2012). Chemical heterogeneity on Mercury's
surface revealed by the MESSENGER X‐Ray Spectrometer. Journal of Geophysical Research: Planets,117.
Weisberg, M. K., Prinz, M., & Nehru, C. E. (1990). The Bencubbin chondrite breccia and its relationship to CR chondrites and the
ALH85085 chondrite. Meteoritics,25, 269–279.
Wert, C. A. (1950). Diffusion coefficient of C in alpha‐Iron. Physical Review,79, 601–605.
Wood, B. J. (1993). Carbon in the core. Earth and Planetary Science Letters,117, 593–607.
Zhang, Z., Hastings, P., Von der Handt, A., & Hirschmann, M. M. (2018 ). Experimental determination of carbon solubility in Fe‐Ni‐S melts.
Geochimica et Cosmochimica Acta,225,66–79.
Zolotov, M. Y., Sprague, A. L., Hauck, S. A., Nittler, L. R., Solomon, S. C., & Weider, S. Z. (2013). The redox state, FeO content, and origin of
sulfur‐rich magmas on Mercury. Journal of Geophysical Research: Planets,118.
10.1029/2019JE006239
Journal of Geophysical Research: Planets
VANDER KAADEN ET AL. 15 of 15