A Program on Electronic Structure (Materials Research)

这项提案要求支持最近在康奈尔大学开始的“电子结构计划”。该计划的目的是对电子结构计算中使用的算法进行重大改进，减少它们对大系统的计算要求，并随着算法的进步而增加对化学键的理解。这三种标准技术--局域密度泛函方法、格林函数蒙特卡罗方法和量子化学方法--将一起研究，目的是结合每种方法的最佳特征。该计划的一些具体早期项目如下所示。使用新的、非常快速的算法计划对局部密度泛函方法进行广泛而仔细的测试。主要分子序列的研究将没有赝势，也没有特别的基组限制。基于重整化群思想，人们正在寻求一个正式而准确的框架来定义超重原子的赝势。格林函数蒙特卡罗方法中使用的变分波函数正在变得更加通用，以期提高其精度。康奈尔大学已经广泛使用了向量和并行超级计算机，这将有助于实现完全收敛的计算。

This proposal requests support for a "Program on electronic structure" recently started at Cornell. The purpose of the Program is to make major improvements in algorithms used in electronic structure computations, reducing their computational requirements for large systems and building increased understanding of the chemical bond along with the algorithmic advances. The three standard techniques-Local Density functional methods, Green's Function Monte Carlo methods and quantum chemistry approaches - will be studied together with the aim of combining the best features of each. Some specific early projects of the program are as follows. Extensive careful tests of the local density functional method are planned using new, very fast algorithms. Major sequences of molecules will be studied with no pseudopotentials and no ad-hoc basis set limitations. A formal and accurate framework defining pseudopotentials for very heavy atoms is being pursued based on renormalization group ideas. Variational wave functions used in the Green's Function Monte Carlo method are being made more general in hopes of increasing its accuracy. Extensive use of vector and parallel supercomputers already at Cornell will help achieve fully converged computations.