化学学报 ›› 2004, Vol. 62 ›› Issue (9): 847-853.    下一篇

研究论文

大体系分区密度泛函计算方法

胡向前, 王繁, 黎乐民   

  1. 北京大学化学与分子工程学院,稀土材料化学及应用国家重点实验室,北京,100871
  • 投稿日期:2003-10-16 修回日期:2004-01-16 发布日期:2014-02-17
  • 通讯作者: 黎乐民,E-mail:lilm@pku.edu.cn;Fax:010-62751708 E-mail:lilm@pku.edu.cn
  • 基金资助:
    国家自然科学基金(Nos.29928002,20333020)资助项目.

A Density Functional Computational Method Based on Division of Large Systems

HU Xiang- Qian, WANG Fan, LI Le-Min   

  1. College of Chemistry and Molecular Engineering, State Key Laboratory of Rare Earth Materials Chemistry and Applications, Peking University, Beijing 100871
  • Received:2003-10-16 Revised:2004-01-16 Published:2014-02-17

提出一种新的对大体系进行分区密度泛函计算的方法.将大体系划分为若干较小的区,每个小区是一个相对独立的量子力学子体系,计及其它区势场的影响和电子的Pauli排斥作用,可以进行相对独立的密度泛函计算.对各子体系求解单电子方程:
(FK+FpK)CK=SKCK K=A,B,C,…
式中FKK, CK, SK,εK分别为子体系K的Fock矩阵、轨道系数矩阵、基组重叠矩阵和本征值矩阵,FpK起强制属于不同子体系的占据轨道之间保持正交的作用.得到的轨道是分区定域化的,汇总各区的计算结果得出整个体系的电子结构信息.通过对一些较大分子的计算,考察了几种因素对分区计算精度的影响.结果表明,提出的方法是可行的,通过控制各区基组的大小,可以基本消除基组截断误差,得到精确的计算结果.对于足够大的体系,本方法是一种线性标度算法;和文献报道的相关方法比较,更容易用于对体系的某些区域进行特别的计算研究.

关键词: 密度泛函计算, 大体系, 分区计算方法, 线性标度算法

A new density functional approach for calculations of large systems is presented.The system under consideration is divided into regions which are relatively independent quantum mechanical subsystems and can be calculated individually by means of density functional method, taking into account the effects of the potentials and the Pauli repulsion from the other subsystems.The single particle equation proper to each subsystem is solved,
(FK+FKp)CK=SKCKεK K=A,B,C, …
where FK, CK, SK, εK are the Fock matrix, the coefficient matrix of orbitals, the overlap matrix of the basis set and the energy eigenvalue matrix, respectively, related to subsystem K .The matrix FpK plays the role to constrain the occupied orbitals belonging to different subsystems to be orthogonal to each other.The generated orbitals are regionally localized.Assembling the calculated results from all subsystems yields the electronic structure information of the whole system.The effect of several factors on the calculation accuracy was studied through the calculation of several relatively larger molecules.It is found that the presented method is feasible, the truncation errors of basis sets can be eliminated essentially and accurate results can be generated by controlling the size of the regional basis sets.The presented method is a linear scaling algorithm for sufficiently large systems.Comparing with the existent methods based on the similar strategy, the present method is easier to be used in the study of some regions for special requirements.

Key words: density functional calculation, large system, computational method based on division, linear scaling algorithm