An efficient localization procedure for large systems using a sequential transformation strategy

A fast and efficient fragment-based top-down localization scheme is developed to obtain occupied and virtual regionally localized molecular orbitals (RLMOs) for large systems. In the scheme, RLMOs in many fragments of the target system are constructed using a fragment-based sequential transformation strategy based on a series of singular value decomposition operations. The present localization scheme is a non-iterative procedure, which computationally scales cubically with the system size. The locality of RLMOs is quite close to that of localized molecular orbitals obtained with conventional localization schemes. For several large systems, we demonstrate that this approach is efficient for computing both occupied and virtual RLMOs. (c) 2014 AIP Publishing LLC.