Self-consistent finite-difference electronic structure calculations for large systems

An efficient and flexible self-consistent method of solving the Schrödinger equation for large systems is presented. This uses a finite-difference method, with the atomic cores replaced by an embedding potential. The resulting Hamiltonian matrix is sparse and can be diagonalised using the Lanczos algorithm, with computer time proportional to the system size. This all-electron method uses a small muffin tin radius and allows for the full potential outside the muffin tin. Within the self-consistent local density functional framework, Poisson's equation is solved using the multigrid method. Results for fcc copper are shown.