Approximation of solution operators of elliptic partial differential equations by \({\mathcal H}\)- and \({\mathcal H}^2\)-matrices (Q967581)

The author improves in two ways the four-step approach to invert a stiffness matrix \(A\) approximated by an \({\mathcal H}\)-matrix as introduced by \textit{M. Bebendorf} and \textit{W. Hackbusch} in [Numer. Math. 95, No. 1, 1--28 (2003; Zbl 1033.65100)]. First, the solution operator is approximated directly instead of using an integral operator based on Green's function. This eliminates the second step of the original proof and yields error estimates with respect to the natural Sobolev norms instead of the weaker \(L^2\)-norm estimate. Second and more importantly, the \(L^2\)-projections are replaced by Clément-type interpolation operators [\textit{P. Clement}, Rev. Franc. Automat. Inform. Rech. Operat. 9, R-2, 77--84 (1975; Zbl 0368.65008)]. Hence, an approximation of the inverse matrix is constructed directly without a detour via the inverse mass matrix \(M^{-1}\), thus getting local blockwise estimates for the error instead of the global ones developed by \textit{L. Grasedyck} and \textit{W. Hackbusch} [Computing 70, No. 4, 295--334 (2003; Zbl 1030.65033)]. It is proved that this approach can be generalized to cover inverse matrices corresponding to partial differential equations by switching to data-sparse \({\mathcal H}\)- and \({\mathcal H}^2\)-matrices.