期刊
THEORETICAL CHEMISTRY ACCOUNTS
卷 112, 期 5-6, 页码 442-452出版社
SPRINGER
DOI: 10.1007/s00214-004-0602-3
关键词
fragment molecular orbital; second-order Moller-Plesset perturbation theory; parallelism; integral transformation; density matrix
We propose a parallelized integral-direct algorithm of the second-order Moller-Plesset perturbation theory (MP2) as a size-consistent correlated method. The algorithm is a modification of the recipe by Mochizuki et al. [(1996) Theor Chim Acta 93:211]. There is no need to communicate the bulky data of integrals across worker processes, keeping the formal fifth-power dependence on the number of basis functions. A multiple integral screening procedure is incorporated to reduce the operation costs effectively. An approximate MP2 density matrix can also be directly calculated through the integral contraction with orbital energies. We implement the MP2 code by accepting Kitaura's fragment molecular orbital (FMO) scheme as in the program ABINIT-MP developed by Nakano et al. [(2002) Chem Phys Lett 351:475]. The error in the FMO-MP2 energies is found to be within the order of the chemical accuracy. Timing and parallel acceleration results are shown for test molecules.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据