What is the best alternative to diagonalization of the Hamiltonian in large scale semiempirical calculations?

Abstract
No abstract available