Moderately and slightly siderophile element constraints on the depth and extent of melting in early Mars

The thermal history of Mars during accretion and differentiation is important for understanding some fundamental aspects of its evolution such as crust formation, mantle geochemistry, chronology, volatile loss and interior degassing, and atmospheric development. In light of data from new Martian meteorites and exploration rovers, we have made a new estimate of Martian mantle siderophile element depletions. New high pressure and temperature metal–silicate experimental partitioning data and expressions are also available. Using these new constraints, we consider the conditions under which the Martian mantle may have equilibrated with metallic liquid. The resulting conditions that best satisfy six siderophile elements—Ni, Co, W, Mo, P, and Ga—and are consistent with the solidus and liquidus of the Martian mantle phase diagram are a pressure of 14 ± 3 GPa and temperature of 2100 ± 200 K. The Martian mantle depletions of Cr and V are also consistent with metal–silicate equilibration in this pressure and temperature range if deep mantle silicate phases are also taken into account. The results are not consistent with either metal–silicate equilibrium at the surface or at the current‐day Martian core–mantle boundary. Recent measurements and modeling have concluded that deep (～17 GPa or 1350 km) mantle melting is required to explain isotopic data for Martian meteorites and the nature of differentiation into core, mantle, and crust. This is in general agreement with our estimates of the conditions of Martian core formation based on siderophile elements that result in an intermediate depth magma ocean scenario for metal–silicate equilibrium.     

Thermal-orbital evolution of Io and Europa

Coupled thermal-orbital evolution models of Europa and Io are presented. It is assumed that Io, Europa, and Ganymede evolve in the Laplace resonance and that tidal dissipation of orbital energy is an internal heat source for both Io and Europa. While dissipation in Io occurs in the mantle as in the mantle dissipation model of Segatz et al. (1988, Icarus 75, 187), two models for Europa are considered. In the first model dissipation occurs in the silicate mantle while in the second model dissipation occurs in the ice shell. In the latter model, ice shell melting and variations of the shell thickness above an ocean are explicitly included. The rheology of both the ice and the rock is cast in terms of a viscoelastic Maxwell rheology with viscosity and shear modulus depending on the average temperature of the dissipating layer. Heat transfer by convection is calculated using a parameterization for strongly temperature-dependent viscosity convection. Both models are consistent with the present orbital elements of Io, Europa, and Ganymede. It is shown that there may be phases of quasi-steady evolution with large or small dissipation rates (in comparison with radiogenic heating), phases with runaway heating or cooling and oscillatory phases during which the eccentricity and the tidal heating rate will oscillate. Europa's ice thickness varies between roughly 3 and 70 km (dissipation in the silicate layer) or 10 and 60 km (dissipation in the ice layer), suggesting that Europa's ocean existed for geological timescales. The variation in ice thickness, including both convective and purely conductive phases, may be reflected in the formation of different geological surface features on Europa. Both models suggest that at present Europa's ice thickness is several tens of km thick and is increasing, while the eccentricity decreases, implying that the satellites evolve out of resonance. Including lithospheric growth in the models makes it impossible to match the high heat flux constraint for Io. Other heat transfer processes than conduction through the lithosphere must be important for the present Io.     

Thermal history and evolution of Mars

A theoretical thermal evolution model of Mars is constructed, utilizing as constraints the available geophysical and geological data, including those provided by the Viking missions. The calculation includes conduction and subsolidus mantle convection. Calculated models indicate that Martian evolution can be roughly characterized by four different stages. (1) Core formation and crust differentiation: this stage starts from the planet formation to about 1 by thereafter. During this period, Martian core is separated and the initial crust is differentiated. (2) Heating, expansion, and mantle differentiation: this stage begins after the core separation and extends to about 3 by. First, mantle temperatures rise and reach partial melting. Between 2 and 3 by, extensive melting, differentiation, and outgassing occur. Planetary radius increases and extensional features observed at the surface are most likely generated at this stage. (3) Mature phase: after 3 by, the planet reaches maturity. Between 3 and 4 by slow and sustained evolution continues. Lithosphere thickens and partial melt zone deepens. (4) Cooling period: this stage represents the last phase of Martian history. The planet is cooling slowly. The partial melting zone shrinks and volcanic activity tapers off. At present, Martian lithosphere is about 200 km thick and the mantle is convecting slowly. The models suggest that the core is molten, and the calculated surface heat flux is 35 erg cm^?2 sec^?1.     

Numerical simulations of the differentiation of accreting planetesimals with 26Al and 60Fe as the heat sources

Abstract— Numerical simulations have been performed for the differentiation of planetesimals undergoing linear accretion growth with ²⁶Al and ⁶⁰Fe as the heat sources. Planetesimal accretion was started at chosen times up to 3 Ma after Ca‐Al‐rich inclusions (CAIs) were formed, and was continued for periods of 0.001–1 Ma. The planetesimals were initially porous, unconsolidated bodies at 250 K, but became sintered at around 700 K, ending up as compact bodies whose final radii were 20, 50, 100, or 270 km. With further heating, the planetesimals underwent melting and igneous differentiation. Two approaches to core segregation were tried. In the first, labelled A, the core grew gradually before silicate began to melt, and in the second, labelled B, the core segregated once the silicate had become 40% molten. In A, when the silicate had become 20% molten, the basaltic melt fraction began migrating upward to the surface, carrying ²⁶Al with it. The ⁶⁰Fe partitioned between core and mantle. The results show that the rate and timing of core and crust formation depend mainly on the time after CAIs when planetesimal accretion started. They imply significant melting where accretion was complete before 2 Ma, and a little melting in the deep interiors of planetesimals that accreted as late as 3 Ma. The latest melting would have occurred at <10 Ma. The effect on core and crust formation of the planetesimal's final size, the duration of accretion, and the choice of (⁶⁰Fe/⁵⁶Fe)_initial were also found to be important, particularly where accretion was late. The results are consistent with the isotopic ages of differentiated meteorites, and they suggest that the accretion of chondritic parent bodies began more than 2 or 3 Ma after CAIs.     

Lunar structure and dynamics - results from the apollo passive seismic experiment

Analysis of seismic signals from man-made impacts, moonquakes, and meteoroid impacts has established the presence of a lunar crust, approximately 60 km thick in the region of the Apollo seismic network; an underlying zone of nearly constant seismic velocity extending to a depth of about 1000 km, referred to as the mantle; and a lunar core, beginning at a depth of about 1000 km, in which shear waves are highly attenuated suggesting the presence of appreciable melting. Seismic velocitites in the crust reach 7 km s^–1 beneath the lower-velocity surface zone. This velocity corresponds to that expected for the gabbroic anorthosites found to predominate in the highlands, suggesting that rock of this composition is the major constituent of the lunar crust. The upper mantle velocity of about 8 km s^–1 for compressional waves corresponds to those of terrestrial olivines, pyroxenites and peridotites. The deep zone of melting may simply represent the depth at which solidus temperatures are exceeded in the lower mantle. If a silicate interior is assumed, as seems most plausible, minimum temperatures of between 1450°C and 1600°C at a depth of 1000 km are implied. The generation of deep moonquakes, which appear to be concentrated in a zone between 600 km and 1000 km deep, may now be explained as a consequence of the presence of fluids which facilitate dislocation. The preliminary estimate of meteoroid flux, based upon the statistics of seismic signals recorded from lunar impacts, is between one and three orders of magnitude lower than previous estimates from Earth-based measurements.Paper dedicated to Professor Harold C. Urey on the occasion of his 80th birthday on 29 April, 1973.     

Formation of mountains on Io: Variable volcanism and thermal stresses

Thermal stresses are potentially important drivers of Io's tectonics and mountain building. It has been hypothesized that sustained local or regional shut down of heat-pipe volcanism on Io could lead to deep crustal heating and large compressive stresses [McKinnon, W.B., Schenk, P.M., Dombard, A.J., 2001. Geology 29, 103-106]. Such large stresses would then be relieved by thrust faulting and uplifting of crustal blocks, producing mountains like those observed on Io. Here we analyze the tectonic consequences of the heat-pipe model in detail, considering both the initial thermal stress state of a basalt or peridotite crust created by heat-pipe volcanism, and relative roles of subsidence stresses (due to burial of preexisting layers) and thermal stresses arising from variable volcanism and changes in crustal (∼lithosphere) thickness. We limit the magnitude of the potential subsidence stresses in our study, because the magnitude of subsidence stresses can be quite large, if not dominant. Results indicate that for a fixed crustal thickness, the region of failure and faulting moves closer to the surface as eruption rate decreases and time increases. When the crust melts at its base as volcanism decreases (as might occur under steady state tidal heating), resulting in crustal thinning, the region of failure is brought even closer to the surface. Naturally, when compressive, subsidence stresses are included, the vertical extent of crust in brittle failure thickens to include most of the lithosphere. In contrast, increases in eruption rate cause the extent of the region in compressional failure to decrease and be driven very deep in the crust (in the absence of sufficient subsidence stress). Therefore, regions of declining volcanism are more likely to produce mountains, whereas regions of extensive or increasing volcanism are less likely to do so. This is consistent with the observation of a global anticorrelation between mountains and volcanic centers on Io. Finally, we find that the choice of crustal composition/rheology (dry basalt vs. dry peridotite) has little effect on our results implying that basalt, peridotite and komatiite are all similarly “stiff” in the Io environment.     

The influence of mantle melting on the evolution of Mars

We present a parameterized convection model of Mars by incorporating a new heat-flow scaling law for stagnant-lid convection, to better understand how the evolution of Mars may be affected by mantle melting. Melting in the mantle during convection leads to the formation of a compositionally buoyant lithosphere, which may also be intrinsically more viscous by dehydration. The consequences of these melting effects on the evolution of terrestrial planets have not been explored before. The temporal evolution of crust and lithospheric mantle is modeled in a self-consistent manner considering mantle melting, convective instability, and the rewetting of dehydrated lithosphere from below by hydrogen diffusion. Though the effect of compositional buoyancy turns out to be minimal, the introduction of viscosity contrast between wet and dry mantle can considerably slow mantle cooling and sometimes lead to non-monotonic core cooling. Furthermore, with or without dehydration stiffening, our model predicts that the martian mantle must have been degassed more extensively (>80%) than previously suggested (<10%); the loss of such a large amount of water from the mantle to surface has significant implications about the role of water in the early surface and climate evolution of Mars.     

Stability of sulfur slopes on Io

The mechanical properties of elemental sulfur are such that the upper crust of Io cannot be primarily sulfur. For heat flows in the range 100–1000 ergs cm^?2, sec^?1, sulfur becomes ductile within several hundred meters of the surface and would prevent the formation of calderas with depths greater than this. However, the one caldera for which precise depth data are available is 2 km deep, and this value may be typical. A study of the mechanical equilibrium of simple slopes shows that the depth to the zone of rapid ductile flow strongly controls the maximum heights for sulfur slopes. Sulfur scarps with heights greater than 1 km will fail for all heat flows greater than 180 ergs cm^?2 sec^?1 and slope angles greater than 22.5°. The observed relief on Io is inconsistent with that anticipated for a predominantly sulfur crust. However, a silicate crust with several percent sulfur included satisfies both the mechanical constraints and the observed presence of sulfur on Io.     

Io: Energy constraints and plume volcanism

Observational and theoretical considerations, including near-surface energy constraints, suggest a model of Io that features a surface layer of sulfur overlying an active silicate crust. Such a model would imply frequent contact between silicate magma intrusions and the sulfur layer. This contact could produce volcanic plumes driven by high-temperature sulfur vapor. Plumes driven by sulfur vapor meet observationall constraint for a wide range of possible conditions, in contrast to the special conditions required for plume generation by SO₂. Characteristics of the two models are compared, and it is suggested that high-resolution infrared radiometry could identify the driving volatile.     

Mg-rich harzburgites from Vesta: Mantle residua or cumulates from planetary differentiation?

We describe petrographic, electron microprobe, and laser ablation ICP-MS analyses of Mg-rich harzburgite clasts in the Dominion Range 2010 howardites, and conclude that they are xenolithic samples of the vestan mantle. Key chemical and petrologic characteristics of these rocks provide tests for differentiation models. Our results indicate the mantle of Vesta formed through variable degrees of partial melting, which left behind a harzburgite and possibly dunite residuum. The Mg-rich clasts are composed of orthopyroxene and olivine, with minor clinopyroxene, FeNi metal, and distinctive pyroxene–chromite symplectites. We use mineral chemistry to demonstrate the absence of a genetic link between diogenites and the Mg-rich harzburgites. We propose a secondary origin for the formation of symplectites: interaction of silicate and metallic melts during primordial differentiation and core formation. The occurrence of FeNi metal containing ~1.5 wt% Cr within the assemblage indicates a very reducing environment during mantle differentiation (≪IW). Our study suggests that Vesta did not experience complete melting early in its history, and instead supports the formation of a shallow magma ocean.     

Implications from Galileo Observations on the Interior Structure and Chemistry of the Galilean Satellites

Data from the recent gravity measurements by the Galileo mission are used to construct wide ranges of interior structure and composition models for the Galilean satellites of Jupiter. These models show that mantle densities of Io and Europa are consistent with an olivine-dominated mineralogy with the ratios of Mg to Fe components depending on mantle temperature for Io and on ice shell thickness for Europa. The mantle density and composition depend relatively little on core composition. The size of the core is largely determined by the core's composition with core radius increasing with the concentration of a light component such as sulfur. For Io, the range of possible core sizes is between 38 and 53% of the satellite's radius. For Europa, there is also a substantial effect of the thickness of the ice layer which is varied between 120 and 170 km on the core size. Core sizes are between 10 and 45% of Europa's radius. The core size of Ganymede ranges between one-quarter and one-third of the surface radius depending on its sulfur content and the thickness of the ice shell. A subset of the Ganymede models is consistent with an olivine-dominated mantle mineralogy. The thickness of the silicate mantle above the core varies between 900 and 1100 km. The outermost ice shell is about 900 km in thickness and is further subdivided by pressure-induced phase transitions into ice I, ice III, ice V, and ice VI layers. Callisto should be differentiated, albeit incompletely. It is proposed that this satellite was never molten at a large scale but differentiated through the convective gradual unmixing of the ice and the metal/rock component. Bulk iron-to-silicon ratios Fe/Si calculated for the inner pair of satellites, Io and Europa, are less than the CI carbonaceous chondrite value of 1.7±0.1, whereas ratios for the outer pair, Ganymede and Callisto, cover a broad range above the chondritic value. Although the ratios are uncertain, in particular for Ganymede and Callisto, the values are sufficiently distinct to suggest a difference in composition between these two pairs of satellites. This may indicate a difference in iron-silicon fractionation during the formation of both classes of satellites in the protojovian nebula.     

Moments of inertia and the Chandlerian period for two-and three-layer models of the Galilean satellite Io

We consider two-layer (Fe-FeS core+silicate mantle) and three-layer (Fe-FeS core+silicate mantle+crust) models of the Galilean satellite Io. Two parameters are known from observations for the equilibrium figure of the satellite, the mean density ρ₀ and the Love number k₂. Previously, the Radau-Darwin formula was used to determine the mean moment of inertia. Using formulas of the Figure Theory, we calculated the principal moments of inertia A, B, and C and the mean moment of inertia I for the two-and three-layer models of Io using ρ₀ and k₂ as the boundary conditions. We concluded that when modeling the internal structure of Io, it is better to use the observed value of k₂ than the moment of inertia I derived from k₂ using the Radau-Darwin formula. For the models under consideration, we calculated the Chandlerian wobble periods of Io. For the three-layer model, this period is approximately 460 days.     

Asteroid Differentiation: Pyroclastic Volcanism to Magma Oceans

Abstract— Asteroid differentiation was driven by a complex array of magmatic processes. This paper summarizes theoretical and somewhat speculative research on the physics of these processes. Partial melts in asteroids migrate rapidly, taking < 10⁶ years to reach surface regions. On relatively small (<100 km) asteroids with sufficient volatiles in partial melts (<3000 ppm), explosive volcanism accelerated melts to greater than escape velocity, explaining the apparent lack of basaltic components on the parent asteroids of some differentiated meteorites. Partial melting products include the melts (some eucrites, angrites), residues (lodranites, ureilites), and unfractionated residues (acapulcoites). The high liquidus temperatures of magmatic iron meteorites, the existence of pallasites with only olivine, and the fact that enstatite achondrites formed from ultramafic magmas argue for the existence of magma oceans on some asteroids. Asteroidal magma oceans would have been turbulently convective. This would have prevented crystals nucleated at the upper cooling surface (the only place for crystal nucleation in a low-pressure body) from settling until the magma became choked with crystals. After turbulent convection slowed, crystals and magma would have segregated, leaving a body stratified from center to surface as follows: a metallic core, a small pallasite zone, a dunite region, a feldspathic pyroxenite, and basaltic intrusions and lava flows (if the basaltic components had not been lost by explosive volcanism). The pallasite and dunite zones probably formed from coarse (0.5–1 cm) residual olivine left after formation of the magma ocean at >50% partial melting of the silicate assemblage. Iron cores crystallized dendritically from the outside to the inside. The rapid melt migration rate of silicate melts suggests that ²⁶Al could not be responsible for forming asteroidal magma oceans because it would leave the interior before a sufficient amount of melting occurred. Other heat sources are more likely candidates. Our analysis suggests that if Earth-forming planetesimals had differentiated they were either small (<100 km) and poor in volatiles (<1000 ppm) or they were rich in volatiles and large enough (>300 km) to retain the products of pyroclastic eruptions; if these conditions were not met, Earth would not have a basaltic component.     

Late stage of accumulation and chemical composition of the terrestrial planets

The model in which the differences of chemical composition of the terrestrial planets are determined by special conditions at the later stage accumulation is discussed. Impact heating would rapidly lead to differentiation of Mercury's interiors. Subsequent high-velocity collisions of Mercury with planetesimals of a comparable size would erode away much of the silicate crust and mantle; such silicates would be accumulated by Venus and fall into the Sun. This model is in agreement with the current models of the terrestrial planets internal constitution.     

Highly siderophile element (HSE) abundances in the mantle of Mars are due to core formation at high pressure and temperature

Highly siderophile elements (HSEs) can be used to understand accretion and core formation in differentiated bodies, due to their strong affinity for FeNi metal and sulfides. Coupling experimental studies of metal–silicate partitioning with analyses of HSE contents of Martian meteorites can thus offer important constraints on the early history of Mars. Here, we report new metal–silicate partitioning data for the PGEs and Au and Re across a wide range of pressure and temperature space, with three series designed to complement existing experimental data sets for HSE. The first series examines temperature effects for D(HSE) in two metallic liquid compositions—C‐bearing and C‐free. The second series examines temperature effects for D(Re) in FeO‐bearing silicate melts and FeNi‐rich alloys. The third series presents the first systematic study of high pressure and temperature effects for D(Au). We then combine our data with previously published partitioning data to derive predictive expressions for metal–silicate partitioning of the HSE, which are subsequently used to calculate HSE concentrations of the Martian mantle during continuous accretion of Mars. Our results show that at midmantle depths in an early magma ocean (equivalent to approximately 14 GPa, 2100 °C), the HSE contents of the silicate fraction are similar to those observed in the Martian meteorite suite. This is in concert with previous studies on moderately siderophile elements. We then consider model calculations that examine the role of melting, fractional crystallization, and sulfide saturation/undersaturation in establishing the range of HSE contents in Martian meteorites derived from melting of the postcore formation mantle. The core formation modeling indicates that the HSE contents can be established by metal–silicate equilibrium early in the history of Mars, thus obviating the need for a late veneer for HSE, and by extension volatile siderophile elements, or volatiles in general.     

Models, figures and gravitational moments of Jupiter’s satellite Io: Effects of the second order approximation

The effects of the equilibrium figure theory to within terms of the second order in a small parameter α on figure parameters and gravitational moments of the Galilean satellite Io have been considered. Integro-differential equations of the theory of figure to second order have been first solved numerically. Relations between the low-order coefficients of the gravitational field for satellites in hydrostatic equilibrium are generalized according to the second order theory. To show the effects of the second approximation, two three-layer trial models of Io are used. The considered models of the Io’s interiors differ by the size and density of the core, while having the same thickness and density of the crust, and the mantle density difference is only 20 kg/m³. The corrections of second order in smallness to the gravitational moments J₂ and C₂₂ decrease the third decimal digit of model gravitational moments by two units. As the effects of third and forth harmonics are determined mostly by outer layers of Io, to distinguish between model mantle density, the gravitational moments J₄, C₄₂ and C₄₄ should be determined to accuracy with three or four decimal digits. The second order corrections mostly effect the semi-axis a, and less the semi-axes b and c.     

Modeling the early evolution of Vesta

The early evolution of the asteroid Vesta has been extensively studied because of the availability of relevant data, especially important new studies of HED meteorites which originated from Vesta and the Dawn mission to Vesta in 2011–2012. These studies have concluded that an early melting episode led to the differentiation of Vesta into crust, mantle, and core. This melting episode is attributed to the decay of ²⁶Al, which has a half‐life of 7.17 × 10⁵ yr. This heating produced a global magma ocean. Surface cooling of this magma ocean will produce a solid crust. In this paper, we propose a convective heat‐transfer mechanism that effectively cools the asteroid when the degree of melting reaches about 50%. We propose that a cool solid surface crust, which is gravitationally unstable, will founder into the solid–liquid mix beneath and will very effectively transfer heat that prevents further melting of the interior. In this paper, we quantify this process. If Vesta had a very early formation, melting would commence at an age of about 1,30,000 yr, and solidification would occur at an age of about 10 Myr. If Vesta formed with a time delay greater than about 2 Myr, no melting would have occurred. An important result of our model is that the early melting episode is restricted to the first 10 Myr. This result is in good agreement with the radiometric ages of the HED meteorites.     

The heating history of Vesta and the onset of differentiation

In this work, we study the link between the evolution of the internal structure of Vesta and thermal heating due to ²⁶Al and ⁶⁰Fe and long‐lived radionuclides, taking into account the chemical differentiation of the body and the affinity of ²⁶Al with silicates. We explored several thermal and structural scenarios differing in the available strength of energy due to the radiogenic heating and in the postsintering macroporosity. By comparing them with the data supplied by the HEDs and the Dawn NASA mission, we use our results to constrain the accretion and differentiation time as well as the physical properties of the core. Differentiation takes place in all scenarios in which Vesta completes its accretion in <1.4 Ma after the injection of ²⁶Al into the solar nebula. In all those scenarios where Vesta completes its formation in <1 Ma from the injection of ²⁶Al, the degree of silicate melting reaches 100 vol% throughout the whole asteroid. If Vesta completed its formation between 1 and 1.4 Ma after ²⁶Al injection, the degree of silicate melting exceeds 50 vol% over the whole asteroid, but reaches 100 vol% only in the hottest, outermost part of the mantle in all scenarios where the porosity is lower than 5 vol%. If the formation of Vesta occurred later than 1.5 Ma after the injection of ²⁶Al, the degree of silicate melting is always lower than 50 vol% and is limited only to a small region of the asteroid. The radiation at the surface dominates the evolution of the crust, which ranges in thickness from 8 to about 30 km after 5 Ma: a layer about 3–20 km thick is composed of primitive unmelted chondritic material, while a layer of about 5–10 km is eucritic.     

Partial melting of the Indarch (EH4) meteorite: A textural,chemical, and phase relations view of melting and melt migration

Abstract— To test whether aubrites can be formed by melting of enstatite chondrites and to understand igneous processes at very low O fugacities, we have conducted partial melting experiments on the Indarch (EH4) chondrite at 1000–1500 °C. Silicate melting begins at 1000 °C, and Indarch is completely melted by 1500 °C. The metal-sulfide component melts completely at 1000 °C. Substantial melt migration occurs at 1300–1400 °C, and metal migrates out of the silicate charge at 1450 °C and ～50% silicate partial melting. As a group, our experiments contain three immiscible metallic melts (Si-, P-, and C-rich), two immiscible sulfide melts (Fe- and FeMgMnCa-rich), and silicate melt. Our partial melting experiments on the Indarch (EH4) enstatite chondrite suggest that igneous processes at low fO₂ exhibit several unique features. The complete melting of sulfides at 1000 °C suggests that aubritic sulfides are not relics. Aubritic oldhamite may have crystallized from Ca and S complexed in the silicate melt. Significant metal-sulfide melt migration might occur at relatively low degrees of silicate partial melting. Substantial elemental exchange occurred between different melts (e.g., S between sulfide and silicate, Si between silicate and metal), a feature not observed during experiments at higher fO₂. This exchange may help explain the formation of aubrites from known enstatite chondrites.     

Crustal recycling, mantle dehydration, and the thermal evolution of Mars

We have reinvestigated the coupled thermal and crustal evolution of Mars taking new laboratory data concerning the flow behavior of iron-rich olivine into account. The low mantle viscosities associated with the relatively higher iron content of the martian mantle as well as the observed high concentrations of heat producing elements in a crust with a reduced thermal conductivity were found to promote phases of crustal recycling in many models. As crustal recycling is incompatible with an early separation of geochemical reservoirs, models were required to show no episodes of crustal recycling. Furthermore, admissible models were required to reproduce the martian crust formation history, to allow for the formation of partial melt under present day mantle conditions and to reproduce the measured concentrations of potassium and thorium on the martian surface. Taking dehydration stiffening of the mantle viscosity by the extraction of water from the mantle into account, we found that admissible models have low initial upper mantle temperatures around 1650 K, preferably a primordial crustal thickness of 30 km, and an initially wet mantle rheology. The crust formation process on Mars would then be driven by the extraction of a primordial crust after core formation, cooling the mantle to temperatures close to the peridotite solidus. According to this scenario, the second stage of global crust formation took place over a more extended period of time, waning at around 3500 Myr b.p., and was driven by heat produced by the decay of radioactive elements. Present-day volcanism would then be driven by mantle plumes originating at the core-mantle boundary under regions of locally thickened, thermally insulating crust. Water extraction from the mantle was found to be relatively efficient and close to 40% of the total inventory was lost from the mantle in most models. Assuming an initial mantle water content of 100 ppm and that 10% of the extracted water is supplied to the surface, this amount is equivalent to a 14 m thick global surface layer, suggesting that volcanic outgassing of H₂O could have significantly influenced the early martian climate and increased the planet’s habitability.