Accelerating em integral equation forward solver for global geomagnetic induction using SVD based matrix compression method

Abstract

We develop an singular value decomposition-based compression of the Green's function matrix of an electromagnetic integral equation forward solver for global geomagnetic induction, on top of an fast Fourier transform reduction of the system to a block-diagonal form. With this approach, the memory usage and CPU time of Krylov subspace iterative solutions are significantly reduced at a very small cost of accuracy, making the accelerated forward solver well suited for 3-D inversions as well as forward simulations with multiple sources. & copy; 2014 The Authors 2014. Published by Oxford University Press on behalf of The Royal Astronomical Society.