The scales of mantle convection

Seismic, topographic and gravity data show that there are two important scales of mantle convection. These are associated with spherical harmonic degrees ℓ=2 and 6. The ℓ=2 pattern corresponds to the pattern of subduction cooling since the breakup of Pangea. Most hotspots and ridges occur in the half of the globe unaffected by this cooling and all large igneous provinces were generated over this part of the mantle. The ℓ=2 distribution of upwellings and downwellings is likely to be a long-lived feature of the Earth; cold regions of the mantle repeatedly attract continents and subduction reinforces the coldness. An ℓ=1 pattern of convection is probably related to supercontinents and their breakup. Degree 6 convection shows up in the spectrum of hotspots and upper mantle tomography and in the correlation of tomography and topography with the geoid. Cratons, with their deep cold keels, control the ℓ=6 pattern, and may even cause it, by their role in establishing lateral temperature gradients and relief at the top of the convecting mantle. Subduction zones reinforce the craton pattern. Downwellings preferentially occur under cratons; upwellings, and hotspots, occur at complementary locations. Supercontinents, and their associated subduction zones (ℓ=1) constantly assemble and reassemble in the African-Atlantic hemisphere and the continental fragments, in the dispersed state, periodically settle into the polar band of geoid lows (ℓ=2 and 6) that now includes the Americas, Antarctica, Australia and India. If cratons control the ℓ=6 pattern of convection (and many patterns of ℓ=6 are possible; e.g., sectoral, zonal, checker-board) then upper-mantle convection may reorganize roughly every 30 Ma as cratons move about. The hotspot spectrum is probably related to lithospheric extension as well as to broad upwellings. A third and smaller scale of convection, order 400–1000 km in dimension, is just below the resolution of global tomography but shows up in the gravity field (geoid) and topography. This scale is controlled by the depth of an endothermic phase change which tends to stratify mantle convection, and the thickness of the upper mantle low viscosity zone. Convective domains of this dimension are also implied by the scales of chemical homogeneity, lengths of rifts, ridges and seamount chains, fracture zone spacing and mid-ocean ridge segmentation. The temperature, geochemistry and fertility of these upper-mantle scale domains is controlled by their previous history of subduction, continental insulation or refrigeration, and processing by ridges. Fertile, or volatile-rich, or hot, cells can be mistaken for plumes. Hotspot swells are typically of the dimension that we argue is a characteristic upper mantle scale rather than a deep mantle plume scale. Geochemical domains of various sizes exist in the upper mantle. They are broad-scale features, rather than point sources, as in plume theories. Lithospheric dynamics and geometric focusing, not mantle dynamics, control the dimensions of so-called hotspot eruptives