The geodynamic origin of melting anomalies found at the surface, often referred to as “hotspots,” is classically attributed to a mantle plume process. The distribution of hotspots along mid-ocean-ridge spreading systems around the globe, however, questions the universal validity of this concept. Here, the preferential association of hotspots with slow- to intermediate-spreading centers and not fast-spreading centers, an observation contrary to the expected effect of ridge suction forces on upwelling mantle plumes, is explained by a new mechanism for producing melting anomalies at shallow (<2.3 GPa) depths. By combining the effects of both chemical and thermal density changes during partial melting of the mantle (using appropriate latent heat and depth-dependent thermal expansivity parameters), we find that mantle residues experience an overall instantaneous increase in density when melting occurs at <2.3 GPa. This controversial finding is due to thermal contraction of material during melting, which outweighs the chemical buoyancy due to melting at shallow pressures (where thermal expansivities are highest). These dense mantle residues are likely to locally sink beneath spreading centers if ridge suction forces are modest, thus driving an increase in the flow of fertile mantle through the melting window and increasing magmatic production. This leads us to question our understanding of sub–spreading center dynamics, where we now suggest a portion of locally inverted mantle flow results in hotspots. Such inverted flow presents an alternative mechanism to upwelling hot mantle plumes for the generation of excess melt at near-ridge hotspots, i.e., dense downwelling of mantle residue locally increasing the flow of fertile mantle through the melting window. Near-ridge hotspots, therefore, may not require the elevated temperatures commonly invoked to account for excess melting. The proposed mechanism also satisfies counterintuitive observations of ridge-bound hotspots at slow- to intermediate-spreading centers, yet not at fast-spreading centers, where large dynamic ridge suction forces likely overwhelm density-driven downwelling.

The lack of observations of such downwelling in numerical modeling studies to date reflects the generally high chemical depletion buoyancy and/or low thermal expansivity parameter values employed in simulations, which we find to be unrepresentative for melting at <2.3 GPa. We therefore invite future studies to review the values used for parameters affecting density changes during melting (e.g., depletion buoyancy, latent heat of melting, specific heat capacity, thermal expansivity), which quite literally have the potential to turn our understanding of mantle dynamics upside down.