Because continental crust is so light – an average density of 2700 kg m^-3 compared with the mantles’ value of 3300 – it has been widely believed that continents cannot be subducted en masse. Yet it is conceivable that sial can be ‘shaved’ from below during subduction and from above by erosion and added to subductable sediment on the ocean floor. Certainly, there is overwhelming evidence for the net growth of continents through time and plenty for periods of increased and dwindling growth in the past. In some ancient orogens there are substantial slabs of continental composition whose mineralogy bears witness to ultra-high pressure metamorphism at depths greater than that of the base of continents. These slabs had been caught-up in subduction but never reached sufficiently high density to be retained by the mantle; they eventually ‘bobbed up’ again. On the other hand, if early continents were less silica rich through incorporation of substantial proportions of rock with basaltic composition parts of them could founder if subjected to high-pressure, low-temperature metamorphism. But not all crustal recycling to the mantle is through subduction. Some abnormally highly elevated parts of the continents that rose quickly in geological terms, such as the Tibetan Plateau, may have formed by lower crustal slabs becoming detached or delaminated from their base. Again modelling can help assess the past magnitude of continental recycling (Chowdhury, P. et al. 2017. Emergence of silicic continents as the lower crust peels off on a hot plate-tectonic Earth. Nature Geoscience, v. 10, p. 698-703; DOI: 10.1038/NGEO3010).

Various lines of evidence suggest that between 65 to 70% of the present continental volume existed by 3 billion years ago, yet that does not manifest itself in the rock record; perhaps a sign that some has returned to the mantle. It is also widely suggested that plate tectonics in the modern style began at about that time. Pryadarshi Chowdhury and colleagues simulate what may happen at depth in continent-continent collision zones – the classic site of orogenies –at different times in the past. Under the hotter conditions in the early Archaean mantle delamination would have been more likely than it has been during the Phanerozoic; i.e. the peeling off and sinking of the denser, more mafic lower crust and the attached upper mantle. The authors show that increased mantle temperature further back in time increases the likelihood and extent of such delamination. It also encourages partial melting of the descending continental material so creating rising bodies of more silicic magma that add to the remaining continent at the surface. Together with the lower crust’s attachment of to a mantle slab, this ensures that the peeled off material is able to descend under its own load. Once below a depth of 250 km felsic rocks are doomed to further descent. Waning of radiogenic mantle heat production encourages descending slabs to fail and break from the connection with lithosphere at higher levels so that a smaller proportion of the lower crust becomes detached and recycled. This evolution suggests that less and less continental crust is recycled with time. This broadly fits with current geochemical ideas based on the record of radiogenic Nd-, Sr- and Pb-isotopes in rocks ranging from early Archaean to Phanerozoic age.