Geophysical evidence for an enriched molten silicate layer above Mars's core

Mars consists of an iron-alloy core overlain by a silicate mantle and crust, likely formed via early global magma ocean differentiation. NASA’s InSight mission enabled the first seismic constraints on Mars’s interior, including crustal layering, mantle structure, and core size and composition. Initial seismic detection of the core indicated a radius of ~1,830 ± 40 km (revised to ~1,780–1,810 km), implying a relatively low core density (~6.0–6.2 g cm⁻³) under the assumption of a compositionally homogeneous mantle. Reconciling this low density with cosmochemical and experimental constraints on feasible light-element contents is challenging: sulfur alone would need to be too abundant; adding O, C, and H in amounts previously inferred conflicts with experimental solubilities and partitioning (e.g., H ≤0.15 wt%, C solubility reduced by S, O ≤4 wt% in Fe–S melts). Meanwhile, the S1000a impact event revealed P waves diffracted along the deep mantle (Pdiff), with differential times relative to PP that only a few models can match, requiring a substantial velocity reduction in the lowermost mantle not explained by homogeneous-mantle models. These observations motivate considering a heterogeneous mantle resulting from magma ocean solidification that created a basal mantle layer (BML) enriched in Fe and heat-producing elements. Such a structure could affect thermochemical evolution and seismic observables, potentially resolving inconsistencies in core size/density estimates and deep-mantle wave speeds.

The study builds on InSight seismic and geodetic results that constrained crustal thickness, mantle structure, and core size (e.g., Stähler et al. 2021; Khan et al. 2021; Drilleau et al. 2022), and re-evaluations suggesting a slightly smaller core radius (~1,780–1,810 km). Prior models generally assumed a homogeneous mantle. Cosmochemical and experimental works limit feasible light-element contents in Mars’s core (e.g., sulfur alone insufficient within cosmochemical bounds; oxygen, hydrogen, carbon solubilities constrained), challenging earlier low-density core interpretations. The S1000a impact event provided unique Pdiff observations (Posiolova et al. 2022), whose travel-time differences with PP require lowermost-mantle velocity reductions inconsistent with homogeneous models. Theoretical and geochemical studies of magma ocean crystallization and cumulate overturn suggest basal enrichment in Fe and heat-producing elements, potentially forming a thermochemical boundary layer or BML. Prior assessments of Mars’s tidal dissipation (Phobos tides) indicate a relatively attenuating mantle at long periods, contrasting with weak seismic attenuation at 1 Hz, suggestive of stratification analogous to the Moon’s deep soft layer.

The authors performed a probabilistic inversion of Mars’s interior using seismic, geodetic, and thermochemical constraints. Two inversion sets were explored: (1) non-BML models with a compositionally homogeneous mantle and (2) BML models with a basal mantle layer above the core enriched in Fe and heat-producing elements. The inversion is parameterized by thermochemical evolution variables for a planet with a liquid metal core, silicate mantle (with/without BML), and evolving lithosphere/crust. Mantle viscosity was modeled as a function of temperature T and pressure P using an Arrhenius-type law: η(r) = A_ref η0 exp(E*/(R T) − V* P/(R T)), where E* and V* are effective activation energy and volume, R is the gas constant, η0 is a reference viscosity, and A_ref is a prefactor. Thermal histories produce present-day temperature, density, and seismic velocity profiles, and determine melt fractions in the BML (mushy versus fully/essentially molten). Ray-theoretical modeling and travel-time calculations were used to predict phase arrivals and differential times (e.g., P, S, PP, SS, PPP, SKS, and deep phases). In homogeneous models, ScS reflects at the CMB and Pdiff propagates along the CMB. In BML models, zero Vs in the fully molten portion of the BML causes S-wave reflections at the top of the fully molten region (~150 km above the CMB), and a core-bouncing diffracted phase, PbdiffPcP, that diffracts along the bottom of the depleted mantle above the BML, traverses the molten silicate layer, reflects at the CMB, and returns to the surface, significantly delaying its travel time. Differential travel-time fitting emphasized S1000a constraints such as t_PP − t_Pdiff and other phase-time combinations; probability density functions (PDFs) of model fits were evaluated across the best 1,000 models. Shear attenuation Q profiles were derived from inversion outputs using a frequency-dependent scaling Q(r, ω) ∝ [η(r)/ω]^α with α between 0.1 and 0.3. Apparent deep-mantle Q, Q_ScS, was estimated from amplitude ratios of ScS to S along raypaths (paths computed with TauP), while global tidal Q2 at Phobos’s frequency (~5 h 55 min) provided geodetic constraints. The gradient of apparent Qs with depth in the solid convecting mantle, R_Q, was inferred by combining travel times and amplitudes from events sampling different depths. Core properties (radius, density, CMB temperature) were adjusted to satisfy mass and seismic constraints. Model performance was assessed against multiple observables: travel times of deep reflected/diffracted phases, attenuation at seismic (1 Hz) and tidal periods, and waveform similarity to InSight records.

• A basal mantle layer (BML) with fully and partially molten silicate above the core simultaneously explains: (1) deep reflected and diffracted seismic phases, (2) weak shear attenuation at seismic frequencies, and (3) strong tidal dissipation at Phobos periods.
• In BML models, S-wave reflections occur ~150 km above the CMB at the transition to fully molten silicate (Vs ≈ 0), not at the CMB itself. A core-bouncing diffracted phase, PbdiffPcP, traverses the molten silicate layer, substantially increasing travel times and matching S1000a differential times that homogeneous models miss.
• Core radius and density: BML models yield Rc = 1,650 ± 20 km and average core density ρc = 6,470 ± 60 kg m⁻³ (~6.47 g cm⁻³), 5–8% higher density than previous seismic estimates, reducing the required light-element inventory in the core. The apparent core radius (core radius plus fully molten silicate thickness) is Ra = 1,780 ± 20 km, close to recent re-estimates from larger datasets.
• Thermal state: The BML acts as both heat buffer and source, causing core heating from ~2,160 K to ~2,840 K and an overall present-day mantle that is ~300 K colder than in non-BML models. BML models start ~70 K hotter on average and have ~30× lower viscosities initially; effective activation parameters differ (smaller E*, larger V* than non-BML), yielding comparable present-day upper-mantle viscosities (~10²¹ Pa s) but a more viscous deep convecting mantle.
• Seismic and tidal attenuation: Inferred Q_ScS ≈ 1,500 ± 500 at 1 Hz indicates weak seismic attenuation in the solid mantle; global tidal Q2 = 95 ± 10 indicates strong long-period dissipation. BML models satisfy both constraints more readily because the partially molten layer accommodates tidal dissipation, while many homogeneous models struggle to match both across frequencies. Models matching both R_Q (1.9 ± 0.6) and Q_ScS tend to have α ≈ 0.2–0.25.
• Composition: The denser, smaller core is consistent with cosmochemical bounds using fewer/lower-abundance light elements (e.g., ~17 wt% S and ~2.9 wt% O, or smaller proportions of S, O, H, C in combination) than previously inferred.
• Implications: The layered mantle implies that crustal magnetic signatures likely require external dynamo-driving mechanisms, as the BML suppresses core heat loss necessary for a long-lived thermally driven dynamo.

The findings address the core-size/composition paradox and deep-mantle seismic anomalies by introducing a thermochemical stratification with a molten silicate BML above the core. This structure naturally shifts the S-wave reflection point above the CMB and slows deep P-diffracted paths, reconciling observed differential times from the S1000a event that homogeneous models fail to match. The revised smaller, denser core alleviates tensions with cosmochemical and experimental limits on light elements, requiring fewer and less abundant alloying components to explain the core density. The BML framework also unifies attenuation observations across disparate frequencies: strong tidal dissipation arises from the partially molten layer, while the solid mantle exhibits weak seismic attenuation at 1 Hz, consistent with observed Q_ScS and inferred R_Q trends. Additionally, the model offers consistency with core nutation constraints. However, by buffering and heating the core, the BML likely impedes a thermally driven core dynamo, implying that Mars’s ancient magnetic crustal signatures may stem from external or transient mechanisms (e.g., early core superheating, overturn-induced CMB heat-flux spikes, giant impacts, or satellite-driven instabilities). Overall, a heterogeneous mantle with a BML provides a coherent explanation of multiple geophysical datasets and reframes interpretations of Mars’s interior evolution.

This study proposes and tests a Mars interior model featuring a basal mantle layer enriched in Fe and heat-producing elements that is partially to fully molten above the metallic core. Probabilistic inversion against seismic travel times and attenuation, along with geodetic tidal constraints, indicates that BML models outperform homogeneous-mantle models: they explain deep reflected/diffracted phases, align with weak seismic attenuation yet strong tidal dissipation, and yield a smaller (Rc ≈ 1,650 ± 20 km), denser (ρc ≈ 6.47 g cm⁻³) core consistent with cosmochemical bounds on light elements. The work implies S-wave reflections occur above the CMB, identifies a core-bouncing diffracted phase (PbdiffPcP), and suggests Mars’s ancient magnetism likely required external or episodic dynamo drivers. Future research should re-examine InSight’s deep-phase windows for additional BML-interacting phases, better constrain the frequency dependence of Qs in Mars’s solid mantle, refine melt fraction and rheology parameterizations within the BML, and integrate additional seismic, geodetic, and laboratory constraints to narrow model uncertainties.

• Frequency dependence of attenuation is imperfectly constrained (α ~0.1–0.3), and results favor α ≈ 0.2–0.25; uncertainty in α affects Q inferences and model comparisons across timescales.
• The seismic modeling employs spherically symmetric (1-D) structures for raypath and travel-time calculations; lateral heterogeneity, if significant, could influence deep-phase interpretations.
• Identification and modeling of PbdiffPcP and S-reflection interfaces rely on assumed melt-fraction thresholds and rheological transitions within the BML, which carry petrological and rheological uncertainties.
• Thermochemical evolution and viscosity parameterizations (E*, V*, η0) influence present-day structure and inferred BML properties; alternative parameter choices could shift model outcomes within uncertainty bounds.
• Some constraints (e.g., waveform similarity, nutation compatibility) reference supplementary analyses not detailed in the excerpt, indicating dependencies on additional modeling choices and datasets.