Monte Carlo configuration interaction
Monte Carlo configuration interaction
复制标题
DOI：
10.1006/jcph.1998.5953
复制
发表时间：
1998-10-10
影响因子：
4.1
通讯作者：
Greer, JC
中科院分区：
文献类型：
--
作者：
Greer, JC
提出了一种求解量子多体问题的方法，该方法具有适合于并行计算机体系结构的性质。基本方法是使用多体展开函数的线性变分原理的应用，称为位形相互作用或位形叠加方法。通过使用用于组态生成的蒙特卡罗技术重复生成展开向量，可以实现变分能量的顺序改进。通过在不同处理器上同时执行扩展空间的独立采样，可以在变分计算之后将结果组合以形成改进的扩展向量。重复该步骤序列，直到在波函数或能量中达到期望的收敛水平。在并行环境下对该方法进行了分析，讨论了算法的效率、可伸缩性和两层并行方法。(C)1998年学术出版社。
A procedure for solving quantum many-body problems is presented and is shown to have properties which make it well suited for parallel computer architectures. The underlying method is an application of the linear variational principle using many-body expansion functions and is known as the configuration interaction or superposition of configurations method. By repeatedly generating expansion vectors using a Monte Carlo technique for configuration generation, a sequential improvement in the variational energy can be achieved. By performing independent samples of the expansion space concurrently on different processors, the results may be combined after a variational calculation to form an improved expansion vector. This sequence of steps is repeated until a desired level of convergence in the wavefunctions or energies is achieved. Analysis of the method is given within a parallel environment: efficiency, scaling, and a two-tiered approach to parallelism with the algorithm are discussed. (C) 1998 Academic Press.