Text S1
Termination of Deep Earthquakes at the Base of the Upper Mantle
It has long been known that the termination of deep earthquakes in subducting slabs
correlates with the depth of the seismic discontinuity that defines the boundary between the
upper and lower mantles. There are only three basic concepts that can potentially explain that
correlation: (i) Slabs lose their strength as they enter the lower mantle; (ii) the stresses in slabs
drop abruptly as they enter the lower mantle; (iii) the mechanism that enables deep earthquakes
to occur cannot operate in the lower mantle. The first of these (slab weakening) is relevant to the
subject of this paper because if slabs do weaken as they enter the lower mantle, that
automatically offers one possible explanation for the cessation of deep earthquakes.
The
arguments for and against such weakening are provided in the main text and will not be
addressed further here.
The second explanation is improbable because (i) there is no obvious reason for an abrupt
drop of tectonic stresses; (ii) should there be such a drop of tectonic stresses, one consequence
would be that in slabs that enter the mantle directly, the deepest earthquakes should not show the
observed down-dip compressive focal mechanisms. Moreover, Brudzinski and Chen (2005)
demonstrated that tectonic stresses are not required for intermediate and deep earthquakes – even
horizontal slabs display earthquakes (which have randomly oriented focal mechanisms reflecting
a spatially varying stress field). Thus, even if the tectonic stresses were reduced abruptly in the
uppermost lower mantle, the stresses associated with the volume decrease associated with
ringwoodite breakdown could still trigger earthquakes unless the instability responsible for those
earthquakes cannot be activated by that reaction.
The third explanation – that the mechanism of deep earthquakes cannot operate in the
lower mantle (transformation-induced faulting) was proposed 20 years ago and the evidence for
that explanation has grown steadily since that time. A brief summary is presented here only to
make clear that termination of deep earthquakes is not an independent argument one can make to
argue that slabs weaken as they enter the lower mantle.
It was suggested in the 1970s that subducting slabs could be so cold that olivine might be
carried metastably into the mantle transition zone (MTZ) (Sung and Burns, 1976). Experimental
data later provided observations that such could be the case and that deep earthquakes in
subducting slabs could be triggered by a shearing instability accompanying the phase
transformations from olivine to its high-pressure polymorphs [e.g. Green and Burnley, 1989;
Green et al., 1990; Green and Houston, 1995; Green, 2007]. That instability, however, cannot
operate during the endothermic ringwoodite break-down reaction that defines the bottom of the
upper mantle (Green and Zhou, 1996; Gleason and Green, 2009). The mineral reactions from
olivine to a modified spinel structure (wadsleyite) or to a true spinel (ringwoodite) are
exothermic polymorphic transformations – one phase changes to another phase of the same
composition but different structure with release of heat – whereas the breakdown of ringwoodite
at the base of the MTZ involves endothermic decomposition of a single parent phase into two
daughter phases, one with a perovskite structure and the other an oxide.
(Mg,Fe)2SiO4 → (Mg,Fe)SiO3 + (Mg,Fe)O
spinel
perovskite
(S1)
oxide
Because the phase transformations from olivine to wadsleyite or ringwoodite are
exothermic and polymorphic [Vaughan and Coe, 1981; Green and Houston, 1995; Green, 2007],
under certain conditions they can lead to latent-heat-driven runaway nucleation of the stable
phase and consequently, through a complicated process, develop a shearing instability that in
subduction zones could provide the explanation for deep earthquakes [Green and Burnley, 1989;
Green et al., 1990; Green and Houston, 1995; Kirby et al., 1996]. In contrast, the breakdown
reaction (S1) yields two daughter phases; such reactions cannot support such a reaction-triggered
instability whether or not it is exothermic [Gleason and Green, 2009] and therefore could be the
reason for termination of earthquakes at the base of the MTZ.
Metastable olivine has now been observed seismically in four deep slabs [Iidaka and
Suetsugu, 1992; Chen and Brudzinski, 2001; Kaneshima et al., 2007; Jiang et al., 2008;
Kawakatsu and Yoshioka, 2011], providing strong support for this deep earthquake mechanism
and implying that slabs are very cold and essentially dry below 400 km [Green et al., 2010;
Kawakatsu and Yoshioka, 2011]; such observations also suggest that cold slabs are strong.
Runaway shear heating has also been suggested as a mechanism for deep earthquakes
[e.g. Karato et al., 2001] but that mechanism in and of itself cannot explain cessation of deep
earthquakes at the base of the upper mantle because there is no intrinsic reason for that
mechanism to be active in the transition zone but impossible in the lower mantle. Moreover, the
discovery of significant amounts of metastable olivine in multiple cold slabs during the last 10
years undermines many of the assumptions upon which the hypothesis of earthquakes by
runaway shear heating has been based (see also Green, 2003).
If follows from this discussion that cessation of earthquakes cannot be used as an
argument for slab weakening during entry into the lower mantle. The arguments for strength of
slabs must stand on their own merits.