一种基于迭代子空间直接求逆算法的高效子空间混合算法

An Efficient Subspace Hybrid Algorithm Based on Direct Inversion in Iterative Subspace Algorithm

Research output: Journal Publications and Reviews (RGC: 21, 22, 62)21_Publication in refereed journalpeer-review

View graph of relations

Author(s)

Related Research Unit(s)

Detail(s)

Original languageChinese (Simplified)
Pages (from-to)418-422
Journal / PublicationJisuan Wuli/Chinese Journal of Computational Physics
Volume38
Issue number4
Publication statusPublished - Jul 2021

Abstract

发展一种基于迭代子空间直接求逆 (DIIS) 算法的可用于快速求解原子和分子 Hartree-Fock 方程的子空间混合算法。该算法利用不同大小的子空间的混合将更接近收敛的 Fock 矩阵的权重增大,混合后的 Fock 矩阵作为新的迭代起点能有效减少自洽场迭代 (SCF) 过程中的迭代次数。与 DIIS 算法进行对比,发现该算法优于两种不同子空间的DIIS算法,并有效减少了计算量。
A hybrid algorithm based on Direct Inversion in Iterative Subspace (DIIS) algorithm is developed for rapid solution of atomic and molecular Hartree-Fock equations. The improved algorithm uses mixing of subspaces with different sizes to increase the weight of Fock matrix closer to convergence, and the mixed Fock matrix is used as a starting point of new iteration. This algorithm reduces effectively the number of iterations in SCF process. Compared with DIIS algorithm before mixing, it is found that this algorithm is superior to the DIIS algorithm with two different subspaces, and the computational cost is effectively reduced.

Research Area(s)

  • 迭代子空间, 自洽场迭代, 混合算法, 加速收敛, DIIS, SCF, hybrid algorithm, accelerated convergence