Thermal evolution of Mercury with a volcanic heat-pipe flux: Reconciling early volcanism, tectonism, and magnetism

A new thermal evolution model for Mercury can explain early volcanism, tectonism, and magnetism in a self-consistent manner.

Abstract

Mercury’s early evolution is enigmatic, marked by widespread volcanism, contractional tectonics, and a magnetic field. Current models cannot reconcile an inferred gradual decrease in the rate of radial contraction beginning at ~3.9 billion years (Ga) with crustal magnetization indicating a dynamo at ~4 to 3.5 Ga and the production of extensive volcanism. Incorporating the strong cooling effects of mantle melting and effusive volcanism into an exhaustive thermal modeling study, here, we show that early, voluminous crustal production can drive a period of strong mantle cooling that both favors an ancient dynamo and explains the contractional history of the planet. We develop the first self-consistent model for Mercury’s early history and, more generally, propose an approach to assess the volcanic control over the evolution of any terrestrial planet or moon.

INTRODUCTION

Mercury’s enigmatic early evolution

Mercury’s geological record is marked by widespread early effusive volcanism (1), a unique global distribution of crustal shortening structures indicative of compressional tectonics (2), and crustal magnetization indicating an ancient dynamo active at ~4.0 billion years (Ga) ago (3). Current models for Mercury’s thermal history assume that planetary-scale cooling is governed by stagnant-lid mantle convection and focus primarily on explaining an inferred 1 to 10 km of radial contraction through effects of secular cooling (2, 4–11). However, model results are inconsistent with a geologically inferred evolution of contraction: Cross-cutting relationships of shortening structures with craters suggest a decline in the contraction rate from 3.9 Ga ago to present day that is not predicted by models (12). Mercury’s ancient dynamo is thought to be “thermally driven” (13), whereby cooling to the overlying mantle drives vigorous core convection and magnetic field generation. An additional puzzle for published models is that such a dynamo is not possible at ~4 Ga ago (4–10, 14). Together, early enhanced global contraction and a dynamo active at ~4.0 Ga ago require an early period of strong mantle cooling that has been challenging to capture in thermal evolution models (4–10, 14).

Extensive volcanic resurfacing events provide an important clue to Mercury’s early evolution: Heat transport through the production and rise of mantle melts in heat-pipes, and their eruption and spreading at the surface (heat-pipe magmatism) can strongly influence the planet’s cooling. The MESSENGER (MErcury Surface, Space ENvironment, GEochemistry, and Ranging) mission to Mercury constrained the record of this volcanic resurfacing and crustal production for the first time (2). The earliest crust is thought to be a graphite flotation crust only a few meters to 1 km thick, which crystallized from an oxygen-poor magma ocean after planetary accretion (15). Afterward, however, basaltic effusive volcanism resurfaced Mercury with repeated low-viscosity lava flows ending at ~4.0 Ga ago, forming a secondary crust called the Intercrater Plains that covers the entire planet (16). From ~4.0 to ~3.5 Ga ago, more localized effusive volcanism continued, creating the Smooth Plains that cover 28% of the planet’s surface (1). In total, a basaltic crust 15 to 60 km thick comprising the Intercrater Plains and Smooth Plains was emplaced by volcanic eruptions (1, 17–21). In this study, we examine the thermal consequence of producing Mercury’s secondary basaltic crust from a partially melted mantle. The importance of heat transport by volcanic advection and resurfacing depends on the melt supply from the mantle. On Jupiter’s moon Io (22, 23) and possibly on the early Earth (24), very high rates of melt migration from the mantle cause volcanic heat transport to be the dominant heat loss mechanism. Previous models for Mercury have either ignored early crustal production (5, 7, 14, 25) or have produced negligible mantle heat loss by this process because they account for only the latent heat flux carried to the surface (6, 8–10) but not the specific heat contribution from the surface accumulation of lavas and consequent downward advection of cold lithosphere into relatively warm mantle (22, 23).

Here, we develop a parameterized modeling strategy to reconcile the three well-constrained elements of Mercury’s early geologic record with the early thermal evolution of the planet (Fig. 1). Our thermal modeling approach includes four new considerations (Materials and Methods) with consequences that are directly testable with existing observations: First, we include a volcanic heat flux in the heat balance that increases early mantle cooling and the early rate of global contraction via the production and removal of melt to generate a volcanic crust. Second, we address the partitioning of water and radioactive elements into these melts, which, in turn, modulates the evolution of melting and the resulting inventory of radioactive elements in the crust. Third, we quantify ohmic dissipation in the core, which affects the likelihood of a dynamo for evolving rates of core cooling. Fourth, we include the effect of elastic membrane stresses in the lithosphere that act to modulate and ultimately reduce the total radial contraction of the planet. To identify and understand conditions consistent with observations, we map an extensive parameter space (tables S1 to S3 and Materials and Methods).

Fig. 1. Mercury’s thermal and geological evolution.

(A) Schematic model for Mercury at 4.5 to 3.5 Ga ago when effusive volcanism was extensive. q_m, q_c, and H_m are the mantle convective heat flux, the core heat flux entering the mantle, and the radioactive heat production (W/kg). Buoyant magma rises from the partial melt layer in the mantle and enters overlying channels (i.e., heat-pipes) at an average volumetric flux v controlled by Darcy’s law. This rapid advection of latent heat to the surface, combined with the downward advection of the cold lithosphere into relatively warm mantle, comprise the volcanic heat flux (q_vol) that can help sustain a thermally driven dynamo. (B) A period 3.5 Ga ago: Volcanism has ceased and radial contraction has resulted in high compressional stresses (σ_θ) to form Mercury’s shortening structures. (C) Timeline of major events. Duration and relative rate of shortening structure formation (brown) and effusive volcanism that produced Mercury’s crust (red). Constraints for the timing of a magnetic field (yellow). Crustal magnetic fields confirm an active dynamo sometime between 4.0 and 3.5 Ga ago. The global dipolar field results from a modern-day dynamo. Whether a magnetic field was active between these times is unknown.

The goal of this study is to satisfy the following constraints: (i) a thermally driven dynamo until at least 4.0 Ga ago, (ii) a total of 1 to 10 km of radial contraction, (iii) volcanism and crustal production that ended ~3.5 Ga ago, and (iv) a total crustal thickness of 15 to 60 km. A thermally driven dynamo at 4.0 Ga ago condition also permits Mercury’s modern-day dynamo to be driven by processes including more recent inner core growth (7, 10, 14), which we do not investigate. In addition, although we show that a decline in contraction rate occurs, consistent with observations, we do not require successful models to exhibit this behavior because this evolution depends on the stratigraphic timeline, about which there is still debate (26, 27). We proceed as follows. We first address how volcanic crustal production influences the secular cooling rate of the planet and how this process affects the likelihood of an early thermally driven dynamo while satisfying the inferred 1 to 10 km of radial contraction. We next discuss how the partitioning of H₂O and radioactive heat-producing elements into rising mantle melts modulates crustal production. Notably, we show that the extent of H₂O and radioactive heat-producing element partitioning is ultimately critical for producing models that predict an average crustal thickness of 15 to 60 km and a cessation of effusive volcanism at 3.5 Ga.

RESULTS

Melt-free solid-state versus volcanic heat-pipe models: Consequences for dynamo generation and radial contraction

We first show a comparison of the predicted thermal evolution for a typical heat-pipe model and a typical solid-state model in which mantle melting and volcanism are ignored (Fig. 2). As discussed in more detail below, the new aspects of our model affect the evolution of heat flow out of the core and mantle, the total amount and evolution of crustal production and planetary contraction, the total concentration of heat-producing elements in the crust, and the duration of an early thermal dynamo. Over all our model runs, we find that the heat-pipe models are more likely to predict an early thermally driven dynamo duration and total radial contraction that are consistent with those inferred from observations (Fig. 3). Further, as discussed more below, we find that these models permit a less-restrictive range of model parameters.

Fig. 2. Comparison of heat-pipe versus solid-state evolution.

Thermal evolution for representative heat-pipe and solid-state models with the same initial conditions (mantle temperature T_m = 1800 K, core temperature T_c = 1900 K, 200 ppm H₂O in olivine, thermal conductivity k_m = 3 W/m K, carbonaceous chondrite abundances of heat-producing elements, and for the heat-pipe model D_mantle/melt = 0.1). (A) q_m for heat-pipe (pink) and solid state (blue) and q_vol (red). (B) T_m (solid line) and T_c (dashed line) for each model. (C) Depth extent of melting (red) and crustal thickness (pink) for the heat-pipe model. (D) q_c for each model. (E) Radial contraction, ∆R, and (F) the net core entropy (E_net). (G and H) Evolution of H₂O and H_m in the mantle (blue and pink curves). The dark red curve in (G) indicates the evolution of K (right y axis). In the heat- pipe model, H₂O concentrations decline with time due to volcanic resurfacing. The decline in H_m with time in the solid-state model results only from the radioactive decay of U, K, and T, while H_m declines at an accelerated rate because of crustal extraction in the heat-pipe model. The heat-pipe model predicts less than 10 km of contraction and, unlike the steady-state model, allows a core dynamo to be sustained until 4.0 Ga ago, produces a 37-km-thick crust by 3.5 Ga ago, and exhibits an initially faster d(∆R)/dt that slows after ~3.5 Ga.

Fig. 3. Time of dynamo cessation and total radial contraction.

(A) Time at which a thermally driven dynamo ends for all model runs for solid-state (blue) and heat-pipe (pink) cooling modes without ohmic dissipation (E_Φ = 0). (B) Same as (A) but with ohmic dissipation included, assuming B_core is 10,000 nT and L_crit is 280 m [i.e., E_Φ ~13 MW/K (fig. S5)]. (C and D) Histograms of the final ΔR, for model runs from (A) and (B), respectively, that can have a thermally driven dynamo at 4.0 Ga ago.

In marked contrast to previous results, we find that mantle melt production and the resultant heat-pipe–controlled crustal production profoundly affect the early thermal evolution of the planet (Fig. 2). Compared to typical solid-state models, rapid mantle cooling with q_vol as the dominant mode of heat transport occurs (Fig. 2, A and B) when melt production and extraction rates are high (Fig. 2C), and this balances the high rate of radioactive heat production expected over the first few hundred million years of evolution (Fig. 2H). This crucial difference between heat-pipe and solid-state models has consequences for sustaining an ancient dynamo. In heat-pipe models, early rapid mantle cooling drives high rates of core heat loss (Fig. 2D), which intensifies entropy production in the core, significantly increasing the likelihood for a core dynamo at 4.0 to 3.5 Ga ago compared to typical solid-state models (Fig. 2F and Fig. 3, A and B). An additional key result is that the enhanced core and mantle cooling from heat- pipe magmatism are consistent with the geologically inferred onset of shortening structure formation at ~3.9 Ga and a decline in the rate of shortening through time (Fig. 2G) (12). In contrast, solid-state models typically predict a relatively constant rate of radial contraction (Fig. 2E).

The duration of volcanism and the final crustal thicknesses in the example heat-pipe model in Fig. 2 are also compatible with observations (Fig. 2C). A 37-km-thick crust was produced 3.5 Ga ago, consistent with production of the intercrater and Smooth Plains (16) and with crustal thickness estimates (17). Critically, melt production and extraction in the heat-pipe model remove significant concentrations of H₂O and radioactive heat-producing elements (U, K, and Th) from Mercury’s mantle. In this example, the initial mantle H₂O concentration of 200 parts per million (ppm) fell to 90 ppm 3.5 Ga ago (Fig. 2G) and accelerated the decrease of radiogenic heat production H_m compared to the solid-state model (Fig. 2H). However, this model also predicts crustal K concentrations that are ~1000 ppm higher than the maximum estimates inferred spectroscopically at present day (Fig. 2G) (28, 29). The decrease in mantle H₂O and radioactive elements in the heat-pipe model also reduces the likelihood that mantle convection can persist to the present day: In the model in Fig. 2, the mantle was cooled by conduction alone after 1 Ga ago.

Overall, our results for all models from our parameter search show that 36% of all heat-pipe models can sustain a thermally driven dynamo until 4.0 Ga ago with 1 to 10 km of radial contraction without ohmic dissipation and 19% can satisfy these conditions when ohmic dissipation is included in the calculations (Fig. 3, C and D). This result is in sharp contrast with solid-state models for which only 2% of models without ohmic dissipation and no model runs with ohmic dissipation were successful (Fig. 3B). Therefore, only heat-pipe models can explain early dynamo and the radial contraction history of Mercury.

Mercury’s surface composition has the lowest oxidation state among the terrestrial planets, consistent with accretion from anhydrous enstatite chondrites, implying a water-poor mantle (30). Furthermore, the water storage capacity of olivine at Mercury’s mean mantle pressure of ~3 GPa and temperature is limited to less than ~100 ppm (with a maximum of <200 ppm), implying mean concentrations in the mantle of around 200 ppm (with a maximum of 400 ppm) (31). Mantle viscosity and thermal expansion are governed by the bulk composition, water content, and initial mantle temperature. Thermal evolution calculations that include heat-pipe magmatism produce an early thermally driven dynamo and less than 10 km of radial contraction for a wide range of parameter combinations (Fig. 4), implying that our results are not sensitive to mantle and core physical properties and mantle rheological parameters that are highly uncertain. In contrast, successful solid-state models require a reference mantle viscosity that implies an unrealistically high water content (Supplementary Materials) (31), an initially superheated core, an initial mantle temperature (T_m) of 1800 to 1900 K that exceeds the mantle solidus over nearly the full mantle depth, a low core thermal conductivity and thermal expansivity, and no ohmic dissipation in the core (Fig. 4). A superheated core is contentious (32) and physically unrealistic because the excess temperature would be quickly consumed through the latent heat cost of mantle melting at and above the core-mantle boundary. The requirement in solid-state models that the core thermal conductivity k_c = 20 W/m K is difficult to reconcile with experimental values of ~30 W/m K for pure iron (33) at Mercury’s core pressures, although previous studies suggest that substantial concentrations of Si can lower the thermal conductivity (34, 35). Furthermore, in contrast to heat-pipe models that capture the requisite 1 to 10 km of global contraction for a range of reasonable core thermal expansivities, successful solid-state models require a core thermal expansivity that is ~33% (i.e., α_c = 4 × 10⁻⁵ 1/K) lower than recent estimates for Mercury (36). Overall, solid-state models thus require highly restrictive, and arguably unphysical, parameter combinations to produce the strong core cooling needed to sustain a dynamo until 4.0 Ga ago and 1 to 10 km of radial contraction (Fig. 4).

Fig. 4. Model parameters with an early dynamo and 1 to 10 km of radial contraction.

Histograms of model parameters for solid-state (blue) and heat-pipe (pink) models for which a thermally driven dynamo persisted until at least 4.0 Ga ago when ohmic dissipation was not included and that predicted 1 to 10 km of total global contraction. (A) Initial T_m, (B) initial excess core temperature T_c − T_m, (C) enstatite or carbonaceous chondrite initial concentration of radioactive elements, (D) mantle thermal conductivity (k_m), (E) water concentration in olivine, and (F) the net entropy (E_s − E_k) at 4.0 Ga ago or, equivalently, the maximum ohmic dissipation (E_Φ) allowed to sustain a dynamo. The black bar shows an estimate of E_Φ assuming B_core = 10,000 nT. (G) Core thermal conductivities k_c, (H) volumetric thermal expansions for the core α_c, and (I) the effect of the mantle-melt partition coefficient, D_mantle/melt, for H₂O and radioactive heat-producing elements.

DISCUSSION

Crustal production: Importance of partitioning of H₂O and radioactive elements

Our study highlights a sensitivity of model success to mantle-melt partitioning of H₂O and heat-producing radioactive elements and, thus, to the value of the partition coefficient D_mantle/melt. Using D_mantle/melt of 0.1, which is inferred from surface observations assuming an enstatite chondrite initial abundance of heat-producing elements, produced the most successful models with a 15- to 60-km-thick crust produced 3.5 Ga ago, between 1 and 10 km of radial contraction and an ancient dynamo 4.0 Ga ago with ohmic dissipation (see Fig. 2 and fig. S8). However, when the partitioning of radioactive heat-producing elements is neglected, our results are inconsistent with observations: Although a dynamo can be active more recently than 4.0 Ga ago and ∆R is between 1 and 10 km, effusive volcanism continues well after 3.5 Ga ago and produces an average crustal thickness of 100 to 170 km (fig. S8). In contrast, using a lower bound for D_mantle/melt of 0.01 considerably decreases the amount and duration of mantle melt production. In this case, although a dynamo can be active more recently than 4.0 Ga ago and ∆R is 1 to 10 km, the final crustal thickness produced was, on average, 13 km (fig. S8A), comparable to the minimum inferred crustal thickness of 15 km and volcanism largely ceased before ~4.0 Ga (fig. S8B). Furthermore, we note that our results rule out a chondritic radioactive element composition entirely for Mercury’s mantle: A chondritic composition consistently produces surface K concentrations that are too high compared with observations (fig. S8C).

Implications for future studies of Mercury and terrestrial planets

Our results and approach provide promising advances for future studies of Mercury’s evolution. In particular, by demonstrating a strong likelihood that Mercury’s early dynamo could have been thermally driven when volcanic resurfacing is considered, our results permit late inner core solidification and a modern-day dynamo driven by compositional buoyancy from inner core growth, consistent with numerical dynamo simulations (37–39). However, dynamo models often require a large inner core to explain the anomalously weak magnetic field strength (37–39), which would predict an additional compositionally controlled contribution to radial contraction that is not considered here. Furthermore, we examine the sensitivity of successful model runs to the core field B_core, which governs ohmic dissipation. For a lower bound on B_core of 1000 nT (Supplementary Materials) (40), ohmic dissipation is negligible, and inclusion of this effect has no effect on model success (fig. S5). In contrast, for B_core > 30,000 nT, ohmic dissipation rates are sufficiently high that no heat-pipe model is successful. These results suggest an upper bound on the ancient field strength in the core that can guide future dynamo simulations.

The plausible suites of volcanic heat-pipe models that reconcile the tectonic, volcanic, and early dynamo evolutions on Mercury also highlight some challenges. First, we do not consider intrusive magmatism, and the proclivity for magma storage will increase as the growing crust warms over the volcanically active period (41, 42). Evolving conditions favoring storage will reduce the volume fraction of mantle melts delivered to the surface to drive maximal cooling and will potentially affect the viscosity structure of the lithosphere and the character of solid-state convection (43). Second, the extrusion efficiency will also be modified by the changing global crustal stress field: As the compressional stress field increases in the crust from global contraction, it will inhibit the vertical ascent of magma to the surface, promoting intrusive rather than extrusive volcanism (1, 42). Third, our upper-bound value for D_mantle/melt = 0.1 and lower bound value of D_mantle/melt = 0.01 predict surface K concentrations that are 3 and 10 times larger than the maximum inferred values, respectively (fig. S8C). If the enstatite chondrite compositional model for heat production in Mercury’s mantle is accurate and the average surface composition of high-K lavas is well understood, then the following considerations could modify the value of D_mantle/melt and reconcile our predicted surface K concentration with observations. The bulk K concentration in the crust could be different from the spectroscopic constraint that provides information only about the top few tens of centimeters (44). Furthermore, the dearth of thermodynamic models for mantle-melt partitioning behavior appropriate for Mercury’s low-oxygen mantle means that this process, and the resulting crustal composition, is incompletely understood (45, 46). Last, partitioning of some K into the iron core from the mantle during its formation under highly reducing conditions (45) would reduce the mantle K concentration. Although contentious, such a picture has been explored parametrically in studies of Mercury’s dynamo energetics (14). Inclusion of such additional core heating could also enhance the likelihood of dynamo generation in heat-pipe models (47) and expand the parameter space of successful models.

Our modeling approach provides a powerful and simple methodology that can be applied to other terrestrial planets. In particular, Mars and Venus both have complex and different histories of volcanism, magnetic field generation, and surface tectonics. A general conclusion of the present study is that mantle heat loss particularly at early stages of planetary thermal evolution when rates of internal radiogenic heat production are maximized is not always controlled primarily by solid-state mantle convection. When this heat production exceeds the rate of surface heat loss by solid-state mantle convection, the production, rise, and eruption of mantle melts can play an essential role in planetary cooling (24).

MATERIALS AND METHODS

A new approach to Mercury’s early thermal evolution

The details of our approach including equations, parameters, and variables explored are given in the Supplementary Materials. Here, we highlight key novel features of our method. To augment existing approaches to modeling Mercury’s thermal history, we first add a volcanic heat flux (q_vol) to the well-established heat balance of typical parameterized stagnant-lid mantle models (48, 49). The volcanic heat flux is controlled by the rate at which an imbalance between radiogenic heat production and stagnant-lid convective heat loss produces excess thermal energy available to overcome the latent heat of melting. The resulting melt is delivered to the surface at an average rate v through upward porous media flow from the mantle melting region through the crust in “heat-pipes,” where it accumulates above initially cold lithosphere that, in turn, subsides into the warmer mantle (22–24). The mantle temperature (T_m) and the temperature of the core-mantle boundary (T_c) evolve depending on the changing rates of internal radiogenic heat production and mantle convective overturn (Fig. 1A) (48, 49). The evolution of v through the crust, the porosity in the melting mantle, and the permeability for a typical heat-pipe model are given in fig. S4.

A second new contribution is our treatment of the requisite conditions for a thermally driven core dynamo (Supplementary Materials). In contrast to simple critical core-mantle boundary heat flux values, we augment established handlings of entropy sources and sinks that require heat loss from the core to lead to positive rates of entropy production (47), to include a more restrictive explicit inclusion of ohmic dissipation. Ohmic dissipation E_Φ ∝ (B_core/L_crit)², where B_core is the magnetic field strength in the core and L_crit is the physical length scale over which the frictional processes leading to this dissipation occurs. For thermally driven core convection that is predominantly in a quasi-geostrophic limit L_crit~H_coreε_Ω^1/3 (50, 51), where ε_Ω is the Ekman number and H_core is the radial extent of the dynamo region. We find that an upper bound for L_crit for Mercury is 280 m (Supplementary Materials) and could be a factor of a few smaller, in particular for modern dynamos in which the dynamo region may be restricted to only part of the outer core. The ancient magnetic field strength is considerably more uncertain. We initially take B_core to be 10,000 nT, which is one order of magnitude greater than Mercury’s present-day large-scale magnetic field at the core-mantle boundary (40) but allows for plausible contributions from small-scale poloidal fields and toroidal components that are undetectable at and above the planet’s surface [e.g., (49)]. The sensitivity of ohmic dissipation to B_core is shown in fig. S5.

A third addition to existing models is the inclusion of elastic membrane stresses that can retard radial contraction with planetary cooling (52) and that arise from the curvature of the elastic part of Mercury’s lithosphere. To this end, we represent Mercury as a thick elastic spherical shell overlying a viscous interior, in contrast to previous studies that assume a fluid planet in hydrostatic equilibrium (8, 9).

A fourth contribution is to include effects related to the preferential partitioning of H₂O and radioactive heat-producing elements (K, U, and Th) into mantle melts that are extracted and erupted at the surface, a process that has been explored in models of Earth (53, 54) and Mars (55–57) evolution. Mantle-melt partitioning expressed through the partition coefficient D_mantle/melt is expected to be similar for each radioactive element (U, K, and Th) and H₂O (Supplementary Materials). From K concentrations inferred from erupted lavas at Mercury’s surface and assumed carbonaceous or enstatite chondrite abundances of heat-producing elements for the planet (table S3), we explore values for D_mantle/melt of 0.1 and 0.01 (45, 58) and compare the results to calculations in which partitioning is ignored. Crucially, volcanic losses of H₂O and radioactive heat production from the mantle increase the melting temperature (59) and viscosity (60) of Mercury’s mantle, which, in turn, have significant consequences for the longevity and intensity of mantle melting and for the likelihood and character of solid-state mantle convection. We explore varying initial H₂O concentrations from 0 ppm (dry mantle) to 400 ppm, which is the maximum water storage capacity of olivine at Mercury’s mean mantle pressures (31).

The concentration of radioactive heat-producing elements in Mercury’s mantle at any time controls the rates of both secular cooling and melt production (Supplementary Materials). A planet’s total inventory of radioactive elements is determined from the material from which it is accreted. Several studies note that Mercury’s high surface concentrations of radioactive elements together with the low Fe and high S surface composition compared to carbonaceous chondrites (the nominal “chondritic” bulk compositional model used commonly as a reference case) are a strong evidence of a highly reducing enstatite chondrite bulk composition (28, 44). Accordingly, we explore initial concentrations of U, K, and Th in the mantle, assuming either an enstatite or a comparatively depleted carbonaceous chondrite bulk composition (table S3). Furthermore, surface concentrations of K inferred in the northern hemisphere range between 300 and 2500 ppm. The extent to which models satisfying the early crustal production, contractional, and magnetic histories of the planet that also satisfy Mercury’s present maximum crustal K concentrations helps to constrain the likelihood for a predominantly enstatite chondrite or carbonaceous chondrite bulk composition.

A practical aim of the methodology developed in our work is ultimately to provide restrictive sets of conditions that can guide a next generation of computationally intensive two-dimensional (2D) and 3D numerical simulations of Mercury’s thermal history. A powerful aspect of our modeling approach over such more complex and computationally expensive simulations is, however, that we can examine sensitivities to an extensive parameter space arising partly from uncertainties in key physical properties and initial starting conditions. Consequently, we investigate 16,200 heat-pipe models and explore 2700 solid-state models with no melting (table S2), even where mantle temperatures exceed the solidus. Although this situation is physically unrealistic, it is consistent with some previous models for Mercury (4, 5, 14, 25). We run such models to show how heat-pipe volcanism alters the planet’s thermal evolution.

Supplementary Materials

This PDF file includes:

Supplementary Text

Figs. S1 to S8

Tables S1 to S3

References