Earthquake size distribution in subduction zones linked to slab buoyancy

The occurrence of subduction zone earthquakes is primarily controlled by the state of stress on the interface between the subducting and overriding plates. This stress state is influenced by tectonic properties, such as the age of the subducting plate and the rate of plate motion(1-4). It is difficult to directly measure stress on a plate interface. However, the stress state can be inferred using the Gutenberg-Richter relationship's(5) b-value, which characterizes the relative number of small compared to large earthquakes and correlates negatively with differential stress(6-13). That is, a subduction zone characterized by relatively frequent large earthquakes has a low b-value and a high stress state. The b-value for subduction zonesworldwide varies significantly(14,15), but the source of this variance is unclear. Here we use the Advanced National Seismic System earthquake catalogue to estimate b-values for 88 sections in different subduction zones globally and compare the b-values with the age of the subducting plate and plate motions. The b-value correlates positively with subducting plate age, so that large earthquakes occur more frequently in subduction zones with younger slabs, but there is no correlation between b-value and plate motion. Given that younger slabs are warmer and more buoyant, we suggest that slab buoyancy is the primary control on the stress state and earthquake size distribution in subduction zones.