基于局域分子轨道的高效多组态自洽场方法及程序

多组态自恰场(MCSCF)是处理准简并、强相关体系（如化学键的断裂、双自由基反应、过渡金属化合物和光化学反应等）最基本的量子化学方法，但其计算量随着活性轨道、活性电子数目的增长而指数增长。为了解决这一问题，本项目拟发展一种基于定域分子轨道的MCSCF方法，根据研究对象自动选择最优的最小活性空间，从而降低计算量。主要特点有：(1)利用分片定域化方法产生MCSCF初始迭代的局域轨道；(2)利用局域分子轨道进行积分预筛选降低积分变换的计算标度；(3)将空穴粒子对应的图形酉群方法用于单激发组态相互作用计算；(4)在轨道优化过程中采用对一阶约化密度矩阵（或Super-CI B矩阵）对角块最小扰动的块对角化方法，以保证轨道的局域性。该方法将直接产生相互正交且定域到分子片上的非活性轨道、活性轨道和虚轨道，不仅避免了先产生离域轨道再定域化所面临的困难，还可以为后续局域动态相关方法提供很好的参考态和轨道。

Multi-configurational self consistent field (MCSCF) is a basic quantum chemistry method and is widely used in studying quasi-degenerate and strongly correlated electronic systems (e.g. single or multi bond-breaking processes, biradical intermediate in chemical reactions, transition metal compounds and photochemistry reactions, etc.). However, the computational cost of MCSCF scales exponentially with the size of active electrons and active orbitals. This project intends to develop an efficient MCSCF method based on localized molecular orbital. The local MCSCF scheme allows a better control the physical nature of active space, thus enable one to choose a proper active space depending on the problem under study. Some key features of our local MCSCF are (1) the fragment localized molecular orbital (FLMO) approach will be used to generate initial guess orbital in MCSCF; (2) integral prescreening based on localized orbital will be introduced to reduce the scaling of integral transformation; (3) graphic unitary group approach (GUGA) with hole-particle correspondence will be applied in configuration interaction with single excitation (CIS) calculation; (4) block diagonalization of one-body reduce density matrix(or Super-CI B matrix) will be performed in orbital optimization in order to keep the locality of molecular orbitals. The global localized core, active and virtual orbitals, will be obtained directly from the local MCSCF method. The convergence difficulties encountered in standard top-down localization of global canonical orbitals could be avoided, which paves the way for developing the new generation of local multi-reference electronic correlation algorithm.