Parsimonious truncated Newton method for time-domain full-waveform inversion based on the Fourier-domain full-scattered-field approximation

To exploit Hessian information in full-waveform inversion (FWI), the matrix-free truncated Newton method can be used. In such a method, Hessian-vector product computation is one of major concerns due to huge memory requirements and demanding computational cost. Using adjoint-state estimated by zero-lag crosscorrelation first-/second-order incident wavefields second-/first-order adjoint wavefields. Different from implementation frequency-domain FWI, construction time domain becomes much more challenging because it not affordable store all time-dependent The widely used wavefield recomputation strategy leads computationally intensive tasks. We have developed an efficient alternative approach computing for time-domain FWI. our discrete Fourier transform applied extract components involved wavefields, which are compute frequency domain. This makes possible avoid reconstructing first- second-order addition, full-scattered-field approximation proposed efficiently simplify computation, enables us refrain repeatedly solving first-order equations (re)computation. With reduced 70% 80% viscous media Gauss-Newton full-Newton construction, respectively. effectiveness also verified frame 2D multiparameter inversion, almost reaches same iterative convergence conventional implementation.