Impact-driven subduction on the Hadean Earth

Transcription

1 In the format provided by the authors and unedited. SUPPLEMENTARY INFORMATION DOI: /NGEO3029 Impact-driven subduction on the Hadean Earth C. O Neill, S. Marchi, S. Zhang and W. Bottke NATURE GEOSCIENCE Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

2 Supplementary Information and Methods S1. Mantle convection Table S1. Model parameters Earth Radius R (km) 6371 Core Radius R c (km) 3481 Gravity acceleration g (m/s 2 ) 9.81 Initial CMB temperature T CMB (K) 4400 Mantle thermal conductivity K (W/K) 4.7 Tillotson Equation of State C (km/s) (EOS) parameters S 1.25 S2. Rheology Table S2. Rheological parameters Diffusion creep upper mantle Diffusion creep lower mantle Dislocati on creep Peierls Pre- factor A (Pa - n s - 1 ) Activation energy E (kj/mol) Activation volume V (cm 3 /mol) Stress exponent n Stagnant- lid Episodic Mobile- lid Surface yield τ 0 (MPa) strength Friction coefficient f S3. Core evolution Table S3. Core properties Core density ρ c (kg/m 3 ) Core heat capacity Cp c (J/kg/K) 840 Core thermal expansivity α c (K - 1 ) Core thermal conductivity K c (W/m/K) 40 S4. Impact model The numerical experiments of Pierazzo et al. [44] suggest a and b pressure- decay exponents to be /- 0.17, and / respectively. We have refined our exponents based on temperature decay curves obtained from isale impact simulations (Figure S1), and find a best fit with a=1.68 and b=2.74, within the permissible bounds on Pierazzo et al. [44].

3 The temperature increases are modelled as a radially symmetric perturbation in 3D within the volume of the mantle, originating at a depth according to Equation S13. In 2D local impact temperature perturbations are derived from of a slice through this 3D heating volume. We also note we exclude core- heating effects in our models, which in reality only apply to a small number of impacts few are large enough for shock heating to penetrate to the core. We note previous work [16] has suggested core convection could be shut off by this effect, though this is contingent on core mixing and core- mantle boundary temperature heat flux rates. Figure S1. ΔT vs distance for 1000km, 400km, and 25km impactors (impact velocity 12.7km/s). Results from hydrocode numerical simulations using isale shown in red [25], and our revised temperature- decay curves (S17) using the described exponents are in blue. This temperature decay is assumed radially symmetric within the mantle in our implementation, and variations in impact angle are not considered. Temperature scalings are truncated at the isobaric core in these examples (see 10 for rationale), which in any case are above the solidus cut- off (not shown here, based on Stixrude and Lithgow- Bertelloni [33], and calculated in the convection solver), and do not impact the numerical implementation. Much of the supersolidus material within the isobaric core region is in a vapour phase, which would be efficiently removed from the solid- Earth system, and a temperature truncation is appropriate for this case.

4 S5. Supplementary Impact models and Convection simulations The following section shows a number of evolutionary calculations for varying impact fluxes (including two examples with no impacts). The impact distributions are from Marchi et al. [20], based on Monte Carlo results of inner solar system impacting rates, and are shown in subplot c on each page. Red dots denote impactors larger than 100km radius, blue dots denote impactors proximal to simulation plane (within eight impactor radii). Also shown are heat flux (a, in TW), and magnetic field strength relative to the present day field (b, present day field, and variability, shown as blue line/rectangle). Melt production, relative to a simulation under present- day mantle conditions, is shown in d, and mobility (the ratio of surface to internal RMS velocity) in e (here mobility > 1 denotes a mobile- lid regime). Coefficient of friction has been varied in these cases between (with end- member stagnant cases set at 0.8). These are labelled. Higher frictions coefficient (~0.8) show limited tectonic activity, very low friction coefficients (~0.2) are generally active, or episodically so.

5 Figure S2. Impact distribution 1. Friction coefficient f=0.3.

6 Figure S3. Impact distribution 1. Friction coefficient f=0.2.

7 Figure S4. Impact distribution 1. Friction coefficient f=0.8.

8 Figure S5. Impact distribution 2. Friction coefficient f=0.3.

9 Figure S6. Impact distribution 2. Friction coefficient f=0.2.

10 Figure S7. Impact distribution 3. Friction coefficient f=0.3.

11 Figure S8. Impact distribution 3. Friction coefficient f=0.2.

12 Figure S9. No impacts. Friction coefficient f=0.3.

13 Figure S10. No impacts. Friction coefficient f=0.2.

14 S6. Heat pipe model with impacts Figure S11. Impact model with volcanic heat pipe incorporated. Figures show temperature snapshots (a- e), with crust shown as copper color scale, and mantle depletion in purple. Melt is extracted and placed at the surface, as per Moore and Webb [10]. Extreme melting dominates heat transport through the stagnant lid though the effect is exaggerated as emplacement of melt within the crust is not simulated. The development of an extremely thick cool crust and lithosphere in response to this mechanism results in efficient cooling of the mantle during subduction, and extreme cooling of the core as these slabs accumulate at the core- mantle boundary. The volcanic crust produced is of basaltic composition, however, it transitions to eclogite at depths ~60km and subduction of this material is unimpeded. However, the dominant dynamics at the surface are not dissimilar to our non- heat pipe models. For example, a small impact near the north pole at ca. ~53Myr instigates a subduction event (54Myr). Also shown are surface conductive (blue) and core (red) heat fluxes in f) (note that most the surface heat flux is dominated by volcanic heat transfer, not shown). G) Shows the predicted magnetic field strength (red) relative to present day (blue line), as well as the logged impact heating rate for comparison. h) The impact distribution

15 used in this model. Red indicates impactors larger than 100km radius, blue indicates impacts less than 8 radii from simulation plane. Effective modelling of volcanic pipe heat loss will minimally require an adequate treatment of intrusive:extrusive volcanism, and thus resolved crustal structure something not well constrained for the Hadean.

16 S7. Correlation between tectonics and impacting Correlating these disparate data sets is difficult, as impacting in particular exhibits a strong exponential decay, unlike mobility in these models. As a result, we calculate residual impact rates, in 25 Myr bins, where the exponential term as been removed, to facilitate comparison with average mobility over the same time intervals. The results are shown in Figure 4c. We find a moderate correlation between active tectonics, represented as mobility, and residual impact rate, across all our models that demonstrate variable tectonic activity. In some cases, as noted, there is a direct causality between impacts and tectonics (red vertical bars, Figure 4b). However, the correlation does not capture indirect effects of impacts, such as heating of the mantle and plume invigoration, which may prime it for active tectonics. There is also tectonic activity demonstrated in these models that is not associated with impacts. Nonetheless, there is a correlation between tectonic activity and impact rate, which we suggest may be a feature of terrestrial planet evolution. Figure S12. Mobility (25 Myr bins over all our non- end member simulations) vs residual impacting rate (25 Myr bins, with an exponential trend removed). There is a moderate positive correlation between plate mobility and impact flux in these simulations, with a correlation coefficient of