A Divide-and-Conquer Implementation of the Discrete Variational DFT Method for Large Molecular and Solid Systems

Surface Structures of SrTiO3 (001): A TiO2-rich Reconstruction with a c(4 × 2) Unit Cell

“…Seeking a qualitative perspective on charge transfer and bonding, local densities of states (LDOS), partial atomic charges, and populations were calculated using a separate linear combination of atomic orbital (LCAO) model. Here, the DFT equations were solved in the local density approximation (LDA, Xα = 0.7 scaled exchange functional) in the framework of the discrete-variational (DV) method. − Owing to the size of the system, a linear-scaling divide-and-conquer (DAC) ansatz was employed − in which every symmetry-unique atom is described by a local cluster of atoms embedded into the effective Kohn−Sham potential of all the other atoms of the slab. A near minimal LCAO basis set of numerical atomic orbitals has been used in the calculations.…”

Solution of the p(2×2) NiO() surface structure using direct methods

“…as``buer atoms'' (see Ref. [29]). With this cut-o, clusters containing 27 atoms describe atoms in bulk NiO.…”

Exploring Sparsity in Three-Dimensional Integration for Density-Functional Calculations

“…This problem is expected to occur in methods performing density matrix weighed prescreening upon differentially updated quantities. The present algorithms are not suited for systems where one-electron occupancies near the Fermi level must be altered to achieve convergence [3,38].…”

“…In the three-dimensional integration approaches based on divide and conquer, low power scaling performance is usually achieved with reducing the size of the partitioned subsystems at the expense of accuracy [28,40,38]. The cost for solving the subsystems, which is in the intermediate size range, is still dominated by the order N 3 integrations.…”