An algorithm for minimization of the density-functional energy is described that replaces the diagonalization of the Kohn-Sham Hamiltonian with block diagonalization into explicit occupied and partially occupied (in metals) subspaces and an implicit unoccupied subspace. The progress reported here represents an important step toward the simultaneous goals of linear scaling, controlled accuracy, efficiency, and transferability. The method is specifically designed to deal with localized, nonorthogonal basis sets to maximize transferability and state-by-state iteration to minimize any charge-sloshing instabilities. It allows the treatment of metals, which is important in itself, and also because the dynamics of ‘‘semiconducting’’ systems can result in metallic phases. The computational demands of the algorithm scale as the particle number, permitting applications to problems involving many inequivalent atoms.

• Received 19 August 1993

DOI:https://doi.org/10.1103/PhysRevB.49.10088