Nature | Vol 586 | 15 October 2020 | 399(Fig. 1d and Fig. 2b, d) and/or by dripping of lower-crustal material via
heating by the mantle diapirs/plumes.
How a protokeel may cool
The modelling reveals that after the entire oceanic plate is consumed
by subduction, the viscous underplate gradually loses connection
with the downgoing plate, its lateral spreading stagnates and its thick-
ness stabilizes (Fig. 1d, Extended Data Fig. 3b). The modelling thus
reveals that the lateral spreading of underplates is controlled mainly
by the active supply of depleted material from beneath the actively
subducting oceanic plate. After the termination of oceanic subduc-
tion by arc-continent collision, spontaneous spreading of the weak,
isolated underplate has a negligible role (compare Fig. 1c and Fig. 1d).
The underplates can thus preserve their thickness on the timescale of
tens of million years until conductive cooling becomes substantial.
The proposed model of viscous emplacement of protokeels under
subcontinental lithospheric mantle does not immediately create a
pronounced negative temperature anomaly relative to the surround-
ing convecting mantle, as is seen in modern subcratonic mantle keels^3.
However, recent thermal modelling has demonstrated that 3 billion
years of conductive cooling of a melt-depleted subcratonic mantle,
independently of its origin (subduction versus plume-related) leads to
anomalously low temperatures compared to the surrounding convec-
tive mantle^4. The buoyancy of the melt-depleted viscous underplates
produced in our model and their location beneath the continents create
appropriate conditions for their thermal accretion to the bottom of the
continuously cooling continental lithosphere. It can also be assumed
that several subsequent episodes of oceanic subduction and associ-
ated protokeel emplacement might have affected some Precambrian
cratons.
Drivers of keel-forming subduction
It is widely accepted that both the gravitational pull of the cold slabs
and the convective circulation of the mantle powered by slabs and
active hot mantle upwellings (plumes) drive plate motions in the mod-
ern Earth^10 ,^33. The interplay between these two mechanisms as well as
their role during initiation of subduction remains debatable even in
modern plate tectonics^34.
According to our modelling, prescribed plate convergence plays
an important part in stabilizing continued subduction, which is not
interrupted by the frequent slab-breakoff episodes that are charac-
teristic of hot Archaean conditions (for example, refs.^24 ,^25 ). However,
protokeels can also form by viscous underplating in the absence of
forced convergence (Extended Data Fig. 8) as sufficient slab pull is
created by the thicker eclogitized oceanic crust that compensates
for the lower density of the depleted oceanic mantle. Indeed, shal-
low slab breakoff can sometimes preclude the efficient growth of the
protokeels (Extended Data Fig. 8e–h). For this reason, the existence
of layered cratonic keels on Earth could be considered as an indicator
of the important role that mantle convection has as a strong driving
force of plate tectonics in the Early Earth.
Timing of protokeel formation
Figure 4a, b compares changes in the mantle potential temperature
Tp (ref.^8 ) (Fig. 4a) and computed characteristic thickness of oceanic
sublithospheric depleted mantle (Fig. 4b, Methods) through geologi-
cal time. The comparison shows clear maxima for both parameters
(1,500–1,600 °C and 40–100 km, respectively) within the time interval
of approximately 1.8–3.5 Ga, which coincides with major statistical
maxima of cratonic lithosphere ages (Fig. 4c–e). In contrast, the nota-
bly cooler mantle temperature conditions (1,300–1,450 °C; Fig. 4a)
corresponding to modern plate tectonics (taking place 0–1 Ga) are
characterized by the reduced thickness of oceanic sublithospheric
depleted mantle (0–25 km; Fig. 4b) and thus are not favourable for the
formation of viscous underplates and related cratonic keels (Fig. 4c–e).
This comparison could imply that the high temperature of the upper
mantle and the related increased thickness of depleted oceanic sublith-
ospheric mantle are the main parameters that controlled craton forma-
tion on Earth by subduction-induced viscous underplating processes.
The distribution of cratonic mantle lithosphere ages often shows sev-
eral maxima (Fig. 4c–e) corresponding to the formation and reworkingRelative probabilitySublithospheric
depleted layer (km)Tpmantle (ºC)Age (Ga)0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5North Atlantic craton
Whole rocks TRD
n = 50Slave cratonWhole rocks TRD
n = 36Kaapvaal craton
Whole rocks TRD
n = 228Suldes TMA
n = 48Modern
plate
tectonicsTp = 1,300–1,450 ºC Tp = 1,500–1,650 ºCHot plate tectonicsCraton formationabcdeSuldes TMA
n = 710204060801001,3001,4001,5001,600Relative probabilityRelative probabilityFig. 4 | Comparison of estimated mantle potential temperature, thickness
of depleted oceanic sublithospheric ages and cratonic ages. Changes in the
mantle potential temperature Tp of estimated^8 for natural non-arc basalts
(black rectangles) (a) and computed (Methods) characteristic thickness of
oceanic sublithospheric depleted mantle (red circles) (b) compared to
measured cratonic mantle ages^35 (the relative probability density plots of
Re–Os model ages are shown for whole rocks and sulfides by black and grey
lines, respectively) in the Kaapvaal (c), Slave (d) and North Atlantic (e) cratons.
The thickness of the hot (>1,300 °C) sublithospheric depleted (>20% melt
extraction) mantle layer in b is estimated (Methods) for 40-Myr-old oceanic
lithosphere using the mantle potential temperature shown in a. Blue rectangles
in a and b correspond to present-day conditions. TMA and TRD correspond to
two different model age approaches^35 used for the Re-Os isotope system.