Development and Applications of large-scale first-principles DFT calculation methods for complicated systems

Ayako Nakata

ICYS-Namiki researcher

Density functional theory (DFT) is a powerful tool to investigate the atomic and electronic structures of materials independently from experimental data. However, because computational costs of DFT calculations scale cubically to system sizes, the treatable system size are quite limited (in most cases up to about a thousand atoms).

CONQUEST is a linear-scaling (O(N)) DFT code developed jointly by UCL and NIMS. CONQUEST achieves O(N) by using the locality of density matrices with the density matrix minimization method. Local orbitals basis functions are used to express the locality of density matrices. Recently, we have introduced the multi-site method (MSSFs), in which the basis functions are constructed as localized molecular-orbital-like functions. This method enables us not only to reduce the number of the basis functions to be minimal but also to improve the accuracy of the functions as much as required. The accuracy and computational efficiency of the present method are demonstrated by investigating energetic and geometrical properties of several systems.

We have also employed an efficient method to calculate eigenstates in a given energy range proposed by Sakurai and Sugiura with the sparse matrices constructed by CONQUEST. It enables us to obtain band energies and charge densities of very large systems. For example, we can provide theoretical STM images and STS of large and complicated surfaces, containing more than 100,000+ atoms.