Wilhelm Jan, Seewald Patrick, Del Ben Mauro, Hutter Jürg
Department of Chemistry and National Centre for Computational Design and Discovery of Novel Materials (MARVEL), University of Zurich , 8057 Zurich, Switzerland.
Computational Research Division, Lawrence Berkeley National Laboratory , Berkeley, California 94720, United States.
J Chem Theory Comput. 2016 Dec 13;12(12):5851-5859. doi: 10.1021/acs.jctc.6b00840. Epub 2016 Nov 4.
We present an algorithm for computing the correlation energy in the random phase approximation (RPA) in a Gaussian basis requiring [Formula: see text] operations and [Formula: see text] memory. The method is based on the resolution of the identity (RI) with the overlap metric, a reformulation of RI-RPA in the Gaussian basis, imaginary time, and imaginary frequency integration techniques, and the use of sparse linear algebra. Additional memory reduction without extra computations can be achieved by an iterative scheme that overcomes the memory bottleneck of canonical RPA implementations. We report a massively parallel implementation that is the key for the application to large systems. Finally, cubic-scaling RPA is applied to a thousand water molecules using a correlation-consistent triple-ζ quality basis.
我们提出了一种用于在高斯基组中计算随机相位近似（RPA）相关能的算法，该算法需要[公式：见正文]次运算和[公式：见正文]的内存。该方法基于使用重叠度量的单位分解（RI）、高斯基组中RI - RPA的重新表述、虚时间和虚频率积分技术以及稀疏线性代数的运用。通过一种克服规范RPA实现内存瓶颈的迭代方案，可以在不进行额外计算的情况下进一步减少内存。我们报告了一种大规模并行实现，这是应用于大型系统的关键。最后，使用相关一致的三重ζ质量基组将立方标度RPA应用于一千个水分子。
