signature=5b1015b92073f0266be741443236bf1e,Impact-induced changes in source depth and volume of magm...

Thermal evolution models

With the convection code GAIA1. HPE are uniformly distributed in the mantle and no primordial crust is assumed. The entire set of parameters used in the simulations is listed in Supplementary Table 2.

During the evolution we track melt (i.e., crustal) production and calculate the source depth of the melt produced as a function of time. We account for the mantle depletion of HPE when extracting melt (see Methods section). Partial melts in Mercury’s mantle are buoyant over the entire mantle depth range, and they always contribute to the building of the crust2O3, SiO2—used to represent major-element mantle chemistry). These melting curves and their parameterizations are reported in Supplementary Fig. 1 and Supplementary Table 3, respectively. Different assumptions lead to variations in the total crust produced, in the timing of crust production, and in the characteristic source depth of the melt associated with convection (Fig. 1a). Other things being equal, a larger amount of HPE in the mantle is associated with a thicker crust since the increased heat production leads to higher temperatures and thus a larger amount of melt (Supplementary Fig. 2a). A megaregolith layer acts as a thermal blanket that prevents efficient heat extraction from the mantle, thus favoring higher temperatures and increased melt and crustal production. Crustal production is concentrated in the early phases of evolution when the mantle is vigorously convecting. The Rayleigh number, a measure of the strength of convection, is inversely proportional to the reference viscosity. Thus a higher reference viscosity corresponds to a less vigorously convecting mantle and to a thinner crust.

Fig. 1

Crustal thickness from convective melting as a function of time. a Each set of data corresponds to different values of the reference viscosity (symbol shape), initial amount of heat-producing elements in the mantle (HPE) and solidus parameterization (symbol size), and presence of a surficial megaregolith layer (indicated by a black contour). Symbol color indicates the average depth of the melt source region according to the color bar. For each model, the rightmost symbol plotted corresponds to the termination of convective-melt production. Models consistent both with the inferred crustal thicknessb The three mantle melting regimes are shown for the baseline model, whose parameters are indicated in the white box. The triangles along the time axis indicate the times of impact for the simulations shown in Figs. 4 and 5

Independent of the choice of parameters, crustal production is very rapid in the early stages of the evolution and is completed between about 500 Myr and 1.5 Gyr. The source depth of the melt is deeper for higher reference viscosities (consistent with classical boundary layer theory, see Methods section) and shallower in the presence of a megaregolith layer. It is roughly independent of the amount of HPE in the mantle (Supplementary Fig. 2a). The source depth of the melt associated with solid-state convection is roughly constant during the entire evolution of the planet, as the limited variation of the color of the symbols for each curve in Fig. 1a illustrates (see also Supplementary Figs. 3 and 4).

Baseline model

A subset of the models plotted in Fig. 1a are compatible both with the inferred thickness of the crust2). The two black boxes in Fig. 2 indicate the location of the lid at the time of the formation of the high-magnesium region (High-Mg), which comprises one of the oldest terrain observed on Mercury2, the model corresponding to a reference viscosity ν

Ref = 1020 Pa s, an abundance of HPE based on the enstatite chondrites4). The total contraction is 9.9 km, of which 4.7 km are accommodated by mantle cooling. The total value is slightly in excess of the 7 km inferred from the analysis of surface contractional features2b and 5).

Fig. 2

Stagnant lid evolution. Temporal evolution of the location of the bottom of the stagnant lid (solid lines) and of the source depth of convective melt (dots) for the models compatible with the thickness of the crust and the duration of volcanic activity (thick black lines in Fig. 1a). The black boxes indicate the inferred thickness of the lid at the time of emplacement of the High-Mg region and the northern volcanic plains (NVP)1b)

Melting regimes

The large expanses of smooth plains are the last events of massive widespread volcanism on Mercury. Their emplacement terminates at about 3.5 Ga1b.

Effects of large impacts on convection

We account for the effects of large impact events on the dynamics of the mantle by using scaling laws to evaluate the thermal anomaly resulting from the release of the impact shock-pressure6). The thermal anomaly induced by an impact warps the isotherms and produces convective currents that enhance melt production and change the vertical extent of the partial melt zone in the volume of the mantle below the impact. We illustrate this effect in Fig. 3, which refers to the baseline model in the case of a Caloris-forming impact occurring at 300 Myr. The dynamical effects of the impact last for about 100 Myr.

Fig. 3

Impact-induced convective currents and melt production. The images pertain to the baseline model (Fig. 1b) in the event of an impact happening at t = 300 Myr and producing a Caloris-sized basin. a–e: Temporal evolution of the temperature field (background) and of the magnitude of the velocity field (arrows). f–j: Temporal evolution of the melt production, which is expressed in equivalent crustal thickness. The outermost layer refers to the column-integrated value. For the simulation shown here, the impactor hits the target with a velocity of 42.5 km/s and an impact angle of 45°. The impact event corresponds to 0 Myr. The thermal anomaly warps the isotherms and convective currents, which generate melt, develop in the region close to the impact but with a certain delay with respect to the impact event. Enhanced melt production persists below the impact location for >100 Myr. The thermal anomaly is completely absorbed after about 180 Myr (not shown)

For the baseline model, we simulate an impact at five different times, representative of the three regimes of melting defined above and shown in Fig. 1b. These times are 300 and 500 Myr (vigorous melting), 750 and 1,000 Myr (waning melting), and 1197 Myr (absent melting). For each of these times, we simulate two events corresponding to impacts occurring over a downwelling and an upwelling. We perform two sets of simulations using impact parameters compatible with the creation of a Rembrandt-sized basin (impactor diameter 37 km, impactor velocity 42.5 km/s, basin diameter 716 km) and a Caloris-sized basin (impactor diameter 92 km, impactor velocity 42.5 km/s, basin diameter 1550 km). For these two basins, estimates of the time of emplacement and volume of the interior volcanic plains are available4 and 5).

Fig. 4

Melt production following an impact forming a Rembrandt-sized basin. a Results using the baseline model (Fig. 1b) for the impact occurring at 300, 500, 750, 1000, and 1197 Myr. The impactor has a diameter of 37 km and hits the target with a velocity of 42.5 km/s at an impact angle of 45°. For each epoch, the cases for the impact occurring over an upwelling and a downwelling are plotted. The impact over the downwelling (black line) initially extracts shallower material with respect to the impact over the upwelling (white line). Differences in the total amount of melt for impacts over a downwelling vs. impacts over an upwelling depend on the size of the basin with respect to the size of the convection cells in the mantle (Fig. 6). The gray background indicates the melting regime as in Fig. 1b, which are illustrated in d–f at the bottom of this figure. b, c Zoom on the first few tens of Myr after the impact events at 750 and 1197 Myr. For impacts happening in the vigorous melting regime and early in the waning melting regime, the depth of the source region for the postimpact melt is rapidly controlled by the contribution of the convective melting (d, e below). For the impact at 1197 Myr the contribution of convection melting is almost absent, and the melt is the result of partial melting in the shallow mantle (f, below). The postimpact melt thickness (y-axis) represents an upper bound for the melt sheet thickness within the basin. d–f For the three melting regimes, the cartoons illustrate the temperature field in the mantle (background red/blue field), the impact-induced thermal anomaly (spherical red shape), and the melting contribution both from convection (azure arrows) and from postimpact melting (violet arrows). The white lines represent a cold (solid) and a hot (dashed) isotherm. The area of the arrows qualitatively indicates the amount of melt produced. More melt production is associated with upwellings (see also Supplementary Fig. 3) The thickness and source depth of the final melt sheet depends on the relative contribution of the two sources of melt and is represented by the boxes labeled “Final melt sheet”

Fig. 5

Melt production following an impact forming a Caloris-sized basin. a As in Fig. 4a but for an impactor with a diameter of 92 km, compatible with the formation of a Caloris-sized basin. b, c Zoom on the first few tens of Myr after the impact events at 750 and 1197 Myr. The larger impactor induces melting at slightly higher depths (colors) with respect to the case of Rembrandt (Fig. 4b, c)

Postimpact eruptions

Following a large impact event, two processes favor the subsequent eruption of magma at the surface at very high ratios of extrusive-to-intrusive volcanism. First, any preexisting crust is largely removed and only partly melted and mixed with upper mantle material in the melt pool5). For basins smaller than the typical convection cell, the influence of impact location is more pronounced, as in the case of Rembrandt (Fig. 4a). This observation can explain the differences in the postimpact melt thickness associated with impacts over upwellings and downwellings in Figs. 4 and 5. This conclusion is not affected by the cylindrical rescaling adopted6. The temporal delay between the basin formation and the last volcanic material extrusion also depends upon the interplay between the impact anomaly and the underlying mantle dynamics, as the impact event in the absent melting regime illustrates (Fig. 4c). An impact occurring over a downwelling, a region where cold and negatively buoyant material descends into the mantle, results in an almost immediate melting event that extracts material from very shallow depths. When the impact is localized over an upwelling, the thermal anomaly revives the upwelling, which was not producing melt at the time of the impact, and generates additional melt for an interval of time of about 70 Myr.

Fig. 6

Convection pattern and basin size. In each panel, the vertical solid black line indicates the impact location, which correspond to a downwelling (a) and an upwelling (b). The background red/blue field represents the temperature field in the mantle. The white lines represent a cold (solid) and a hot (dashed) isotherm. Green arrows represent convection melting, and their area qualitatively indicates the amount of melt produced, which is larger for upwellings, where hot positively buoyant material rises, and almost absent for downwellings, where cold negatively buoyant material sinks (see also Supplementary Fig. 3). For a basin with a diameter comparable to or larger than the size of a convective cell (e.g., Caloris), the location of the impact has a minor influence on the total amount of postimpact melt produced, since the basin encompasses more than a single convective cell (vertical black dashed lines). For a smaller basins (e.g., Rembrandt), more melt is associated with the impact happening over an upwelling. Compare upwelling and downwelling cases in Figs. 4 and 5

Observational signatures of impacts as a function of time

The impact-induced thermal anomaly generates melt in the shallow mantle and subsequently modifies the melt production in the mantle (Fig. 3). For early impacts, when the mantle is in the vigorous melting regime and the melt production rate (i.e., the slope of the curve in Fig. 1b) is high, this shallow melt is rapidly masked by large volumes of convection-related melt that is generated at depth in the mantle, mostly in association with hot plumes (Figs. 4 and 5 and Supplementary Fig. 3). In terms of source depth, the melt erupted in the basin rapidly loses any signature of the shallow, impact-induced melt, to resemble melt erupted at the same epoch in regions unaffected by the impact. Indeed, the source depth for the postimpact melt in the vigorous melting cases in Figs. 4 and 5 is basically undistinguishable from the convective source depth (Supplementary Fig. 3). Based on the large amount of melt produced in the early phases of the evolution, old basins should correspond to only minor, if any, crustal thinning. Crustal thickness maps corresponding to the old large basins Matisse-Repin (location −24.4 °N, 285.1 °E, diameter 887 km) and Calder-Hodgkins (location 17.1 °N, 21.7 °E, diameter 1460 km) are consistent with this expectation4 and 5). The total amount of melt produced decreases by about an order of magnitude and the crustal thinning associated with the formation of the basins is likely preserved. Indeed, the interior of Caloris is associated with the thinnest crust on Mercury4c, f and 5c).

A model for the melt sheets of Caloris and Rembrandt

The emplacement of the Caloris interior plains postdates the Caloris impact event, dated at about 3.7–3.8 Ga5b). From the basin stratigraphy, the volume of the Caloris interior plains has been inferred to be in the range (3.2–5.2) × 106 km3 (ref. 5b). Thus we can match the observed thickness for extrusive-to-intrusive ratios in the range 50–82%. The Rembrandt basin (diameter 716 km) has a history similar to Caloris, forming at about 3.8 Ga and having its interior covered by volcanic plains that postdate the basin formation by 100–200 Myr4b). Similarly to the case of Caloris, a high extrusive-to-intrusive ratio, in the range 28–64%, can explain the observed thickness of 360–520 m inferred by analysis of the crater statistics of the Rembrandt plains

Our model predicts prolonged volcanic activity, from few tens to about a hundred million years, following the Caloris and Rembrandt impact events. This result is consistent with the inferred delay between the formation of the basins and the emplacement of the interior melt sheets. However, the exact timing is difficult to measure with high accuracy, since the error associated with estimates of this kind of delays is close to the value of the delay itself6). For this end-member scenario, volcanic activity within the basin lasts for 75–110 Myr and an extrusive-to-intrusive ratio in the range 36–66% can explain the observed melt sheet thickness.