The essence of the Ringwood megalith model is that while ocean lithosphere subduction is initially thermally driven because the downgoing slab is cold [the basaltic (eclogite) layer is 5% more dense than the surrounding mantle, and compensates for the depleted harzburgite which is 2% lighter – see Fig. 13], the compositional buoyancy difference between the two becomes significant at 700km depth, resulting in the mechanical separation of basaltic from harzburgite/dunite components (see Fig. 15). However, recent modelling by Gaherty & Hager (1994), using a range of viscosity contrasts for eclogite vs harzburgite, shows that the two are unlikely to separate. The slab buckles and folds as it reaches the 700 km discontinuity, but there is no obvious separation of eclogitic and harzburgitic components. The compositional buoyancy differences are subordinate to the overall thermal buoyancy.
While it is generally known that the convecting Upper Mantle (above the 670km discontinuity) is chemically depleted in lithophile elements because of the progressive growth and extraction of the continental crust from it, it has been commonly thought that the Lower Mantle is largely undepleted. However, Kerr et al. (1995) have proposed that the Lower Mantle is also depeleted, in part because of the return of subducting slab material right through the 670km discontinuity into the lower mantle: see also van der Hilst & Seno (1993). This implies that there is much more interchange between Upper and Lower mantle than was first thought. The Lower Mantle feeds into the upper mantle in the form of large hot plumes (see later lecture). Figure 23 below shows how cool subducted material may go right into lower mantle, or get stuck termporarily at the 670km discontinuity and then 'drip' into the lower mantle:
Fig. 23 (after Kerr et al. 1995) shows 2-layer convection of the mantle, as subducting plates lodge at the 670km discontinuity, or get convected back into the upper mantle; but with periodic interchange between the two as cold plates avalanche down into the lower mantle, and deep mantle plumes are displaced and rise to form major oceanic plateaus and continental large igneous provinces ("LIPs"). There may be composition differences between upper mantle and lower mantle as a result of such processes through Earth history.
Larson and Kincaid (1996) then go on to argue that the breakup of major continents, as occurred with the Gondwana supercontinent in the Mesozoic (ca. 130Ma), leads to more rapid subduction of old cold ocean crust. These cold slabs then avalanche down and penetrate the 670km thermal boundary layer into the lower mantle. One effect is to raise the 670km TBL; another is to displace material from the deep lower mantle (D") which appears as major mantle plumes during the mid-Cretaceous magnetic superchron (120 Ma - 80 Ma). See later notes on mantle plumes.