396 | Nature | Vol 586 | 15 October 2020
Article
of portions of the lithosphere that originally formed at mid-ocean
ridges^20 ,^21. This scenario satisfactorily explains (1) the considerable
thickness of the keels^3 ,^6 ; (2) their considerable melt-depletion corre-
sponding to the ancient oceanic lithosphere^1 ; and (3) the occurrence
of eclogite xenoliths in kimberlites, some of which have geochemical
features of oceanic crust^6. There are, however, serious obstacles to the
application of this model caused by inconsistency between the rela-
tively high amounts of ancient oceanic crust in the oceanic lithosphere
compared to the rarity of eclogites among mantle xenoliths in kim-
berlites^1 ,^3 ,^19. In addition, the ancient slabs with relatively thick oceanic
crust would probably be affected by slab break-off and/or slab retreat
rather than by multiple stacking^22. Therefore, an alternative approach
is required to resolve this long-standing geological problem and here
thermo-mechanical modelling may provide insights.
Previous thermo-mechanical modelling of oceanic subduction
at active plate margins and intraoceanic tectonic settings has dem-
onstrated that plate convergence on top of the fertile upper mantle
does not produce self-sustaining one-sided subduction (even when
plate convergence was prescribed) at upper-mantle temperatures
more than 175 °C above the present (before Neoarchaean time, about
3.0–2.5 Ga) because the plates are weakened by intense percolation
of melts derived from the fertile mantle^22 ,^23. However, plates moving
over the top of a melt-depleted hot mantle that has already produced
an oceanic crust in spreading ridges may have been more coherent and
prone to episodic one-sided subduction^24. Therefore, we designed a
regional two-dimensional high-resolution thermo-mechanical model
(Extended Data Fig. 1, Extended Data Table 1) that simulates oceanic sub-
duction followed by arc-continent collision (see Methods). The model
takes into account initial melt-depletion and the resulting composi-
tional buoyancy of the mantle^1 (see discussion in Methods), hydration/
dehydration processes and melt extraction from the mantle as well as
rheological weakening of the lithosphere subjected to melt percola-
tion (see Methods). With this model, we are able to demonstrate how
cratonic keels can be built by large-scale viscous flows of hot depleted
sub-lithospheric mantle triggered by oceanic subduction at elevated
mantle temperatures.Making keels by Archaean subduction
A typical subduction model with a mantle potential temperature 200 °C
hotter than the present-day value (that is, ΔT = 200 °C, mantle potential
temperature Tp = 1,500 °C) and a prescribed convergence rate shows
continued descent of the oceanic plate without any stacking of the slab
under the overriding plate^20 ,^21. In spite of the high mantle temperature,
hot depleted sub-lithospheric oceanic mantle remains melt-free. As
a result, no melt-induced weakening or buckling of the oceanic plate
occurs (Fig. 1 ), such as that observed in previous Archaean subduction
models that assume a fertile oceanic mantle^22 ,^23. The melt-depleted
sub-lithospheric oceanic mantle is less dense than the underlying fertile
ambient mantle (Methods, Extended Data Fig. 2) and has low viscosity
because of high temperature (>1,300 °C). Therefore, this mantle doesSedimentsMantleSerpentinized/
hydrated mantle
Partially molten
dry/hydrated mantleInitial Newly formed or modi ed
Air/water Mantle strdepleted (>20%)onglyContinental crustContinental crustContinental crustMantle
transition
zoneStagnant slabsSlab
breakoff5 ɫm yr–1MantleContinental crust Oceanic crustOceanic crustOceanic crust5 ɫm yr–1Oceanic crust5 ɫm yr–1Mantle weak zoneContinental crust
(upper/middle)Restite of melting
Upper oceanic crust of hydrated mantleLower crust continental and oceanic2,000 2,400 2,500 2,600 2,700 2,8002,900
Distance (km)2,300200600400050030010020003001003002000100abdΔT = 200 ºC
4.9 Myr8.4 Myr13.8 Myr21.4 Myr5 ɫm yr–1 Mantle
protokeel
20003001002,100 2,200Viscous ow of
depleted mantleViscous underplate source mantleMantle
protokeelDepleted mantleU (kg m–3)3,3003,3503,400
3,45070
90
110
130
150
170
190
210
230
250
2702,200 km
2,450 kmDepth (km)Depth (km)Depth (km)cDepth (km)
Depth (km)Fig. 1 | Development of subduction/collision-induced mantle keel at
elevated mantle temperature ΔT = 200 °C (Tp = 1,500 °C). a, Subduction of
the oceanic plate with highly depleted mantle (at 4.8 Myr from the beginning of
the experiment). b, Detachment of the low-density, hot, ductile and depleted
sublithospheric mantle from the downgoing slab and accumulation in the
hinge of the slab. Thinned slab breaks off (at 8.4 Myr from the beginning of the
experiment). c, Viscous underplating of the low-density, melt-depleted
sublithospheric oceanic mantle derived from beneath the subducting oceanic
plate and accumulation of fragments of the stagnant slab in the mantle
transition zone (at 13.8 Myr from the beginning of the experiment).
d, Stabilization of the viscous underplating mantle and major disconnection of
the protokeel volume from the downgoing plate (at 21.4 Myr from the
beginning of the experiment). Density ρ profiles of the mantle at distances of
2,200 km (orange line) and 2,450 km (magenta line) are shown as the inset in d.
Dotted yellow lines in d indicate upper and lower boundaries of the mantle
transition zone. Arrows show direction of plate motion and viscous f low of the
depleted mantle. The colour key is shown at the bottom of the figure. Viscous
underplate source mantle (T >1,300 °C, melt depletion >20%) is outlined in
magenta for better visibility. Viscosity of the mantle is intrinsically dependent
on pressure, temperature, degree of depletion and presence of f luid/melt
(see Methods). The rheological transition between the rigid lithospheric and
low-viscosity asthenospheric (sublithospheric) mantle approximately
corresponds to the 1,300 °C isotherm.