Localized orbitals for optimal decomposition of molecular properties

Localized orbitals for optimal decomposition of molecular properties
复制标题

DOI:
10.1002/qua.25798
复制
发表时间:
2019-02-05
影响因子:
2.2
通讯作者:
Bulavin, Leonid A.
Bulavin, Leonid A.
中科院分区:
化学3区
文献类型:
--
作者:
Nikolaienko, Tymofii Y.;Bulavin, Leonid A.

文献摘要

被引文献

相似文献

我们给出了将离域分子轨道转换为定域性质优化轨道(LPO)的过程,该定域性质优化轨道旨在建立一阶约化密度矩阵的Frobenius范数意义下最精确的近似,其形式为定域单原子项和双原子项之和。这样,分子性质的分解成与单个原子和原子对相关的贡献就得到了,分解精度的先验上界是已知的。提出了另一种算法,用于获得每对电子含有一个几乎双占据的定域轨道的“化学家的LPO”(CLPO)集合。CLPO形成了一种理想化的Lewis结构,优化后的结构最接近地再现了源自原始多电子波函数的单电子性质。给出了从一般波函数构造LPO和CLPO的计算算法,并讨论了它们在开源免费软件程序JANPA(/)中的实现。使用在Hartree-Fock和二阶Moller-Plesset理论水平上获得的33432个小分子的密度矩阵的测试集对所提出的方法的性能进行了评估,发现与化学家的Lewis结构图非常吻合。
We present the procedure for transforming delocalized molecular orbitals into the localized property-optimized orbitals (LPOs) designed for building the most accurate, in the Frobenius norm sense, approximation to the first-order reduced density matrix in form of the sum of localized monoatomic and diatomic terms. In this way, a decomposition of molecular properties into contributions associated with individual atoms and the pairs of atoms is obtained with the a priori known upper bound for the decomposition accuracy. Additional algorithm is proposed for obtaining the set of "the Chemist's LPOs" (CLPOs) containing a single localized orbital, with nearly double occupancy, per a pair of electrons. CLPOs form an idealized Lewis structure optimized for the closest possible reproduction of one-electron properties derived from the original many-electron wavefunction. The computational algorithms for constructing LPOs and CLPOs from a general wavefunction are presented and their implementation within the open-source freeware program JANPA (/) is discussed. The performance of the proposed procedures is assessed using the test set of density matrices of 33 432 small molecules obtained at both Hartree-Fock and second-order Moller-Plesset theory levels and excellent agreement with the chemist's Lewis-structure picture is found.