Full-waveform analysis of core-mantle boundary structure using adjoint methods

<div>This study presents a new approach for investigating the structure of the core-mantle boundary (CMB)</div><div>topography based on full-waveforms and adjoint methods. We compute intermediate period (10-20 seconds)</div><div>spectral-element seismograms using existing models of core-mantle boundary topography and we analyse the</div><div>sensitivity of relevant seismic phases. Our study adds new information about effects of CMB structure on</div><div>exact synthetics and observable traveltimes of seismic body waves by means of sensitivity kernels. It also</div><div>highlights the difficulty of imaging the boundary due to the strong trade-off between velocity and topography</div><div>variations, addressed by many previous investigators.</div><div> </div><div>Given the significance of CMB and its importance for many disciplines in geophysical research, there have</div><div>been many studies trying to understand and geographically map the variations of topography and velocity</div><div>above this boundary. However, the vast mantle area wherein seismic waves travel before and after they</div><div>interact with the CMB makes the identification of desired seismic phases somehow difficult. In addition, the</div><div>observable traveltimes can be hard to interpret as a result of the boundary’s topography only, due to the</div><div>approximate inverse methods and limited modelling methodologies. Despite considerable progress made the</div><div>past years, there is still a necessity for improving the understanding of effects of core-mantle boundary and</div><div>D″ structure on recorded waveforms.</div><div> </div><div>For our analyses, we perform comparisons between time shifts due to topography made on full-waveform</div><div>synthetics to ray theoretical predictions in order to assess methods usually deployed for imaging CMB.</div><div>Then, we calculate the corresponding sensitivity kernel for time windows around the theoretical arrival of</div><div>each phase. We focus on diffracted, core reflected and refracted <em>P</em> and <em>S</em> waves. The sensitivity kernels</div><div>depict the finite-frequency nature of these waves and possible contributions from other phases unpredictable</div><div>by ray theory. Results show that for most phases ray theory performs acceptably with some accuracy loss,</div><div>however comparisons of the effect of velocity variations to topography on traveltimes are discouraging due</div><div>to the low sensitivity to the latter.</div><div> </div><div>The conclusions drawn by our traveltime and sensitivity analyses are twofold. Firstly, using spectral-</div><div>element waveforms, the seismic phases which are frequently found in literature can be thoroughly investigated</div><div>and better understood, since their traveltime sensitivity through mantle and core is explicitly shown. The</div><div>full-waveform analysis allows us to assess the usability of phases which are informative for core-mantle</div><div>boundary structure and its topography. Secondly, we propose that using the analysed phases simultaneously</div><div>in a full-waveform inversion scheme will improve imaging of the CMB, while also allowing to jointly invert</div><div>for velocity variations along the D″ layer, which is generally poorly understood. From this study, we want</div><div>to promote advanced techniques of full-waveform inversion for improving CMB and lower mantle models.</div>