On the origin of complexity in PKP travel time data

In order to investigate the origin of short spatial scale features in PKP travel time data and to determine whether a complex inner core anisotropy model is required, we have assembled a new global dataset of handpicked absolute PKP(DF) travel times, and completed existing datasets of handpicked relative PKP(AB-DF) and PKP(BC-DF) travel times. We discuss in detail the trends of relative and absolute PKP travel time residuals at the global scale, as well as for a well sampled set of paths between the south Atlantic and Alaska. We discuss the relative merits of several types of models: a) a model of hemispherical anisotropy in the inner core previously proposed to explain PKP(BC-DF) travel time residuals on the global scale; b) a model combining weak constant anisotropy in the inner core with strong heterogeneity in the deep mantle; c) a model involving structure in the outer core associated with the tangent cylinder to the inner core, with axis parallel to the rotation axis, a feature described in magnetohydrodynamical models of the outer core. Because absolute PKP(DF) travel time residuals exhibit the same hemispherical pattern as relative PKP(BC-DF) and PKP(AB-DF) data, when plotted at the location of the bottoming point of DF in the inner core, we infer that the causative structure must at least partly originate in the core. However, the transition between anomalous and normal structure is quite abrupt, and hemispherical inner core anisotropy models fail to reproduce the characteristic ”L shape” of PKP(BC-DF) travel time residuals, when plotted as a function of the angle of the ray in the inner core with the rotation axis (ξ). Models involving mantle heterogeneity compatible with other mantle sensitive data can explain PKP(AB-DF) travel times, but fail to explain 3 sec of average PKP(BC-DF) anomaly observed for paths bottoming in the western hemisphere, for ξ ∼ 20− 30o, even when a model of constant anisotropy in the inner core, compatible with mode splitting data, is also included. On the other hand, models with ∼ 1% faster velocity inside an outer core region roughly delimited by the inner core tangent cylinder allow for rapid transitions, are compatible with rends in absolute PKP(DF) and PKP(BC) times observed in Alaska, and can reproduce the L-shaped feature of the PKP(BC-DF) travel time data. Sustained heterogeneity in the outer core could arise within polar vorteces in and around the tangent cylinder, as suggest by recent dynamical and magnetic investigations. Such models are also compatible with most normal mode splitting data and present less departure from axial symmetry than the hemispherical inner core anisotropy models. When trying to physically explain them, both types of models present challenges, and should be pursued further.