Recent Development of Divide-and-conquer Linear Scaling Method
نویسندگان
چکیده
This paper addresses the recent development of a series of linear-scaling electronic structure calculation methods [1-9], which are based on the divide-and-conquer (DC) method by Yang et al. [11,12]. The DC method used to be applied mainly to pure density functional theory (DFT) or semi-empirical molecular orbital (MO) calculations. We have applied the DC method to the Hartree-Fock (HF) and hybrid HF/DFT calculations [1-3]. For fast self-consistent-field (SCF) convergence, we have modified the direct inversion in the iterative subspace (DIIS) [1] and fraction occupation number (FON) techniques [3]. Furthermore, we have extended the DC method to the correlated calculations such as the second-order Møller-Plesset perturbation (MP2) [4], coupled-cluster with singles and doubles (CCSD) [5], and triples correction (CCST(T)) [6]. In the DC-based correlated calculations, the energy density analysis (EDA) scheme [13] plays an important role to estimate the non-redundant correlation energies of the individual subsystems. In addition to the energies, molecular properties such as dynamical polarizabilities can be obtained by the DC techniques [8]. The computational code for a series of DC calculations has been implemented into the official version of GAMESS program package (2009 Jan R1) [9].
منابع مشابه
A divide-and-conquer linear scaling three-dimensional fragment method for large scale electronic structure calculations
We present a new linear scaling ab initio total energy electronic structure calculation method based on a divide-and-conquer strategy. This method is simple to implement, easy to parallelize, and produces accurate results when compared with direct ab initio methods. The new method has been tested on nanosystems with up to 15 000 atoms using up to 8000 processors. (Some figures in this article a...
متن کاملFree Vibration Analysis of Repetitive Structures using Decomposition, and Divide-Conquer Methods
This paper consists of three sections. In the first section an efficient method is used for decomposition of the canonical matrices associated with repetitive structures. to this end, cylindrical coordinate system, as well as a special numbering scheme were employed. In the second section, divide and conquer method have been used for eigensolution of these structures, where the matrices are in ...
متن کاملOn Using Divide and Conquer in Modeling Natural Systems
In recent years, we have been studying approaches to the realistic modeling of natural systems, especially biological systems. We have tested several of these in a project devoted to modeling pancreatic organogenesis, a complex system that dynamically promotes structural and molecular development. Here, we describe one of these approaches—a kind of ‘divide and conquer’ technique, in which the s...
متن کاملSecond-order Moller-Plesset perturbation (MP2) theory at finite temperature: relation with Surjan's density matrix MP2 and its application to linear-scaling divide-and-conquer method
متن کامل
Evolutionary Divide and Conquer (II) for the TSP
Results presented in recent papers demonstrate that it is possible to produce high quality solutions to TSP instances of up to several hundred cities using simple greedy heuristics when a Genetic Algorithm (GA) is used to perturb the city coordinates. The present paper extends the earlier studies to larger problems and a divide and conquer algorithm in the style of Richard Karp. Using a GA to p...
متن کامل