Seismic tomography has widely been used to determine 3D velocity structures in the Earth's interior based on measurements of travel-time anomalies and waveform differences, and provides inferences to tectonic and geodynamic processes of the crust and mantle. Traditionally, ray-theory, a zeroth-order approach, is used to interpret high-frequency seismic travel time arrivals. However, to produce Fréchet kernels that describe the sensitivity of travel-time measurements or other observables made on broad-band seismograms, the Born approximation must be invoked. This approximation quantifies the scattered/residual wavefields generated by 3D heterogeneities as a first-order approach. Under the Born approximation, calculations of sensitivity kernels involve the interaction of a forward field and a so-called adjoint field, which can both be accurately obtained through numerical simulations. By taking advantage of the symmetry of 1D reference models, the number of numerical simulations and the amount of storage required for forward or adjoint strain fields are significantly reduced, and sensitivity kernels for any source-receiver geometry can be obtained by relatively simple interactions of the two wavefields.
In this talk, I will discuss the implementation of this technique and explore some of the resulting sensitivity kernels that are produced. The final database will enable fast and accurate interpretation of global seismic phases and will be applied to global tomographic inversions in the near future.