A diamond-bearing core-mantle boundary on Mercury

MESSENGER spectral data indicate Mercury’s surface darkness is due to widespread graphite, with neutron and gamma-ray spectroscopy suggesting crustal carbon of about 2–4 wt% (possibly <1 wt% in recent reanalysis). Graphite’s association with lower crustal materials supports an endogenous origin. Mercury differentiated under highly reduced conditions (IW−2.6 to IW−7.3), implying a core rich in Si, S, and C and a silicate portion with low FeO and significant S. Carbon likely degassed as CO2, CO, and CH4 and was partly lost to space, but the prevalence of graphite suggests carbon saturation during metal–silicate differentiation and magma ocean crystallization. Because C solubility in reduced silicate melts is extremely low, excess carbon would have formed a graphite flotation crust. Previously, Mercury’s mantle P–T profile was assumed not to intersect the diamond stability field due to low CMB pressure and uncertain magma ocean liquidus, favoring graphite as the stable C phase. New gravity field results suggest a smaller normalized polar moment of inertia (MOI), implying a deeper CMB (ca. 485 ± 20 km vs 436 ± 25 km), potentially increasing CMB pressure and affecting carbon phase stability. This study reassesses carbon speciation, especially graphite vs diamond stability, under revised interior structures and with the effect of sulfur on silicate liquidus temperatures.

Prior work established Mercury’s carbon-rich surface materials and proposed a primordial graphite flotation crust. Under reduced conditions, light elements in the core include Si, S, and C. Earlier models assumed graphite was the sole stable C phase during magma ocean crystallization because Mercury’s CMB pressures were thought too low for diamond stability. MOI estimates from MESSENGER data have recently been refined, suggesting deeper CMB pressures than previously considered. Experimental and thermodynamic studies show sulfur is highly soluble in reduced silicate melts and can significantly depress liquidus temperatures, potentially shifting magma ocean conditions toward the graphite–diamond stability boundary. Studies of core composition and partitioning indicate anticorrelation of Si and C in Fe-rich cores and suggest a solid FeSi inner core, influencing carbon saturation and phase relations during core crystallization. These strands of literature motivate reevaluation of C speciation and the potential for diamond formation either in the magma ocean or via core processes.

• Interior structure and CMB pressure: The authors recomputed present-day CMB pressures using planetary interior structure models constrained by recent MOI estimates and gravity solutions (HgM008). Using Markov Chain Monte Carlo results (439,000 simulations), they obtained CMB pressures of 5.77 ± 0.31 GPa (MOI 0.333 ± 0.005) and 5.38 ± 0.37 GPa (MOI 0.343 ± 0.006), with an upper bound of ~7 GPa.
• High-pressure experiments: To simulate the deepest magma ocean conditions, they conducted multi-anvil experiments at 7 ± 0.5 GPa and 1973–2273 K on Mercury-like silicate compositions derived from EH–EL chondrites, with variable Si partitioning into metal (Mer8 and Mer15) and FeS added to ensure sulfide saturation (3.08–16.23 wt% S in liquids). A graphite capsule ensured carbon saturation. Phase relations and liquidus/solidus were determined; oxygen fugacity was estimated (IW−3.9 to IW−5.1). Analytical methods included BSE imaging, EMPA for phase compositions, and Raman spectroscopy to verify the graphite–diamond transition proximity.
• Thermodynamic modeling of the magma ocean: Using MAGEMin with the THERMOCALC-based model, they calculated S-free silicate liquidus curves as functions of pressure and parametrized the liquidus depression due to S content using their experimental dataset. They derived a pressure-independent, non-linear expression for liquidus depression with sulfur content: (dT/dS)P = 59.2 x^0.75 K per wt% S (x is S wt%), implying up to ~358 K depression at 11 wt% S.
• Graphite–diamond stability: They used the Day (2012) thermodynamic model to evaluate the graphite–diamond transition across relevant P–T conditions, combining the revised CMB pressures with sulfur-depressed liquidus estimates to compute probabilities that magma ocean conditions fell within the diamond stability field at onset and at 50% crystallization (equilibrium crystallization assumption).
• Carbon solubility modeling: Carbon solubility in reduced silicate melts was calculated using two approaches: (1) the experimental regression of Li et al. (2017), log C[ppm] = 0.96 log X_H2O − 0.25 ΔIW + 2.83 (with X_H2O = 0.01 and ΔIW computed from thermodynamic data); and (2) thermodynamic equilibria relating C to fCO and fCO2, C[ppm] = K_CO fCO + K_CO2 fCO2. These constrain the amount of graphite/diamond that could precipitate from the magma ocean.
• Layer thickness calculations: Using densities (magma ocean 3000 kg/m^3; graphite 2100 kg/m^3; diamond 3500 kg/m^3), core radius ~1950 km, and the pressure interval within diamond stability (~4.77–5.77 GPa), they estimated maximum plausible thicknesses of graphite or diamond layers precipitated from the magma ocean.
• Core carbon content and diamond exsolution: They compiled 598 experiments on C-saturated Fe–Si–C systems (filtered to 244 for robust Si and C ranges) to quantify the anticorrelation of Si and C in metal and fit C_Fe–Si [wt%] = 5.48 − 0.45 Si + 0.009 Si^2. Combining this with geophysical MOI-constrained core Si contents, they inferred probable core C concentrations at C saturation. Considering inner core crystallization as FeSi with ~2.1 wt% C and a liquid/solid C partition coefficient of ~0.3, they modeled enrichment of C in the outer core and consequent diamond exsolution at the CMB, estimating present-day diamond layer thickness.
• Consideration of carbide stability: Literature constraints on Fe3C and Fe7C3 stability and melting at 5–7 GPa were used to argue against carbide stability relative to diamond at Mercury’s shallow core conditions.
• Thermal state and stability today: They compared present-day CMB temperatures from thermal evolution models with the graphite–diamond transition at current CMB pressures to evaluate whether the CMB may sit at or within the diamond stability field, potentially buffering temperature.

• CMB pressure: Interior models yield present-day CMB pressures of 5.77 ± 0.31 GPa (MOI 0.333 ± 0.005) and 5.38 ± 0.37 GPa (MOI 0.343 ± 0.006); maximum plausible CMB pressure ~7 GPa.
• Experimental liquidi/solidi at 7 GPa: For sulfur-bearing EH–EL-derived compositions, Mer8 liquidus 2188 ± 15 K; solidus 2023 ± 50 K. Mer15 liquidus 2213 ± 10 K; solidus 2113 ± 35 K. Orthopyroxene and olivine are primary phases; sulfide present in all runs. The graphite–diamond transition at 7 GPa lies near 2356 K (consistent with Raman observations in capsules).
• Sulfur depresses magma ocean liquidus strongly: 1 wt% S lowers liquidus by ~59 K; up to ~358 K depression at 11 wt% S. The effect is largely pressure-independent under reduced conditions. Carbon has negligible effect (<1 K) on the liquidus due to extremely low solubility.
• Diamond stability in the magma ocean is unlikely: For a sulfur-free melt, all CMB pressure estimates are in graphite stability. With 7 wt% S, 0.6% of onset conditions cross into diamond stability; with 11 wt% S, 8.9%. At 50% equilibrium crystallization, diamond stability probabilities rise to 2.5% (7 wt% S) and 20.6% (11 wt% S). Thus, graphite precipitation and a primordial graphite flotation crust are favored.
• Amount of diamond/graphite from magma ocean: Given low C solubility in reduced melts (Li et al. model ~7–15 ppm C at IW−5 to IW−6; thermodynamic model 0.06–0.2 ppm), diamond produced near the CMB during magma ocean crystallization would form only a 0.1–200 m layer; graphite layer thickness from magma ocean could be ~2–2000 m depending on fO2 and solubility model.
• Core carbon content at C saturation: Using the Si–C anticorrelation in Fe–Si–C and MOI-constrained core Si, inferred core C is 3.4 ± 1.0 wt% (MOI 0.333 ± 0.005) or 1.67 ± 1.0 wt% (MOI 0.343 ± 0.006). The solid inner core is inferred to be FeSi with ~2.1 wt% C.
• Diamond from core crystallization and present-day layer: As the inner core grows (C-poor FeSi), the outer core enriches in carbon. At C saturation under Mercury’s low-P core conditions, diamond exsolves and floats to the CMB. Modeled present-day diamond layer thickness at the CMB is 14.9 ± 10.6 km (MOI 0.333 ± 0.005) and 18.3 ± 10.6 km (MOI 0.343 ± 0.006).
• Carbide stability disfavored: At ~5–7 GPa, Fe3C melts peritectically to liquid + diamond near 1650–1688 K; thus carbides are unlikely to be stable compared to diamond under Mercury’s shallow core conditions.
• Present-day CMB near graphite–diamond transition: Thermal evolution suggests CMB temperatures today are close to the graphite–diamond transition at the relevant pressures, potentially buffering the CMB temperature. A diamond layer could coexist with or lie adjacent to a hypothesized FeS layer if present, depending on FeS physical state.
• Implications: A highly conductive diamond layer at the CMB may influence heat transfer, thermal stratification at the top of the core, and potentially Mercury’s dynamo behavior.

The study revisits the long-standing assumption that graphite is the only stable carbon phase in Mercury’s interior. Incorporating updated geodetic constraints (lower MOI, deeper CMB) and the strong sulfur-induced depression of silicate liquidus temperatures, it shows that while diamond formation in the magma ocean was possible, it was statistically unlikely. The low carbon solubility under reduced conditions means any diamond formed then would be minor. In contrast, inner core crystallization provides an efficient mechanism to exsolve diamond from a carbon-saturated, Si-bearing liquid outer core, with diamonds floating to the CMB and accumulating over time. Modeled thicknesses of a present-day diamond layer (order 10–20 km with large uncertainty) indicate a potentially significant C reservoir at the CMB. The present CMB temperature likely lies near the graphite–diamond boundary, which may buffer the CMB’s thermal state. Redistribution of carbon phases in the mantle is considered limited after strong lower mantle convection ceased (~3.7 Ga), and much of the diamond (or graphite precursor) layer likely accumulated after this epoch. The presence of a diamond layer has implications for Mercury’s thermal evolution: high thermal conductivity may promote thermal stratification at the top of the core, affecting dynamo generation. Possible coexistence or interaction with a FeS layer is discussed, with stability and layering dependent on viscosity and density contrasts. The findings also inform broader questions about carbon cycling and storage in reduced terrestrial planets and potential exoplanetary analogs.

By integrating revised interior models, high-pressure experiments, and thermodynamic calculations, the paper concludes that Mercury likely developed a diamond-bearing layer at the core–mantle boundary primarily through diamond exsolution during inner core crystallization in a carbon-saturated Fe–Si core. Diamond formation directly from the magma ocean was possible but statistically improbable and minor in volume, whereas core processes could yield a CMB diamond layer averaging roughly 15–18 km thick (±10.6 km). Present-day CMB conditions may lie near the graphite–diamond transition, potentially buffering temperatures. These results reshape the understanding of Mercury’s deep carbon cycle and suggest the diamond layer could influence core–mantle heat transfer and magnetic field generation. Future work should aim to quantify how a km-scale diamond layer affects Mercury’s thermal evolution and dynamo, seek indirect geophysical or geochemical signatures compatible with such a layer, investigate interactions with any FeS layer, and explore pathways for bringing diamond to the surface (e.g., via deep-source magmatism in the High-Mg province, to be probed by BepiColombo).

• Interior model uncertainties: MOI estimates and interior structure inversions yield ranges in CMB pressure and core size, leading to large uncertainties in diamond layer thickness.
• Thermodynamic and experimental assumptions: Liquidus depression by sulfur is parametrized from experiments at 7 GPa and assumed pressure-insensitive; carbon’s negligible impact on liquidus is inferred by analogy. Oxygen fugacity calculations assume ideal activities (γFeO = γFe = 1).
• Carbon solubility models: Different approaches (empirical vs thermodynamic) give order-of-magnitude discrepancies (0.06–15 ppm), propagating into estimates of graphite/diamond thickness from the magma ocean.
• Equilibrium crystallization assumption: Probabilities of diamond stability at 50% crystallization assume equilibrium crystallization, which may not strictly apply in a dynamic magma ocean.
• Carbide stability extrapolation: Inferences rely on literature data near 5–7 GPa; exact phase relations in Mercury’s core alloy may vary with composition and temperature.
• Present-day confirmation: A diamond layer cannot be unambiguously detected with current interior models and Love number constraints given uncertainties in mantle and core properties. The extent of graphite-to-diamond transformation and any early redistribution remains uncertain. The existence, thickness, and state (solid/liquid) of a proposed FeS layer at the CMB are also uncertain.