SH velocity and compositional models near the 660-km discontinuity beneath South America and northeast Asia

[1] We constrain SH wave velocity structures near the 660-km discontinuity beneath South America and northeast Asia, using triplicated phases near the discontinuity recorded in the epicentral distance range of 10 –35 for three deep events. We then explore mineralogical and compositional models appropriate for explaining the inferred seismic structures between the two regions. SH velocity structures near the 660-km discontinuity are found to be different in the two regions. Beneath South America, the velocity gradient above the 660-km discontinuity is larger than that of Preliminary Earth Reference Model (PREM), while the velocity jump across the discontinuity is the same as PREM. Beneath northeast Asia, the velocity gradient above the 660-km discontinuity is the same as that of PREM, while the velocity jump across the discontinuity is larger than PREM. Both regions are characterized by a large velocity gradient extending about 80 km deep below the 660-km discontinuity. The different velocity structures require different mineralogical models in the transition zone in the two regions. The larger velocity gradient above the 660-km discontinuity beneath South America requires existence of the ilmenite phase in the bottom of transition zone, while that beneath northeast Asia can be explained by the temperature and pressure dependence of elastic properties of the major mantle mineral assemblages. The observed large velocity gradients in the top of the lower mantle can be explained by gradual transformation of garnet to perovskite persisting to greater depths. The velocity jump across the 660-km discontinuity beneath South America, in the presence of the ilmenite phase in the bottom of the transition zone, requires a bulk composition of more garnet than the pyrolite model, while a larger velocity jump across the discontinuity beneath northeast Asia requires a larger fraction of garnet transforming to perovskite across the discontinuity than what is required for explaining PREM. These different mineralogical models can be caused by different mantle temperature or composition, especially the aluminum content in mantle composition. The presence of garnet 80 km below the 660-km discontinuity in the two regions may be explained by a uniform composition in the lower mantle with an aluminum content of 3.4%. The existence of ilmenite in the bottom of the transition zone beneath South America and the absence of ilmenite beneath northeast Asia can be explained by either a difference in mantle temperature of about 100 C (with that beneath South America being lower) between the two regions assuming a uniform mantle composition or, alternatively, a difference in aluminum content of about 1% (with that beneath South America being lower) between the two regions without invoking a temperature difference between the two regions. We also discuss conditions of mantle composition and temperature that double discontinuities may occur near the 660-km depth, as well as the depth separation and velocity jumps of the double discontinuities. For the inferred mantle temperature and composition beneath South America and northeast Asia, the maximum separation of the double discontinuities is 20 km and cannot be resolved by the SH wave data.