Tensor-factorization for higher excitation order Coupled-Cluster amplitudes for calculating ground-state energy. Automatic code generation

我们将发展一种基于耦合团簇振幅张量分解的电子结构理论的新方法。耦合团簇方法已经成为从头计算分子体系能量和性质的标准工具。但是，即使有了有效的局域实现，也几乎不可能计算超出耦合团簇理论水平的大分子，这些理论包括基态能量的单、双和微扰三重修正(CCSD(T))。主要问题是在迭代程序中存储基态振幅所需的大量内存(在局部公式中，即使是微扰三元组也必须迭代计算)，以及高效实施这些方法所需的高工作量。在这个项目中，一种新的张量分解技术将应用于超过CCSD(T)的耦合簇幅度，并对其精度和性能进行测试。由此产生的振幅的数量将相当小，这样这些量就更容易处理了。自动代码生成器对于高效和快速地实施这些方法是必不可少的，并将成为该项目的一部分。新方法将使计算高精度的基态能量成为可能，而基态能量是热化学和化学反应研究中所需要的。

We will develop a new approach to electronic structure theory based on tensor-factorization of coupled-cluster amplitudes. Coupled-cluster methods have become a standard tool for ab-initio calculations of energies and properties of molecular systems. But even with efficient local implementations it is hardly possible to calculate large molecules beyond the level of coupled-cluster theory with singles, doubles and a perturbative triples correction (CCSD(T)) for ground-state energies. The main problems are the large amount of memory needed to store ground-state amplitudes in iterative procedures (in local formulations even the perturbative triples have to be calculated iteratively) and high effort needed for efficient implementation of these methods. In this project a new tensor-factorization technique will be applied on the coupled-cluster amplitudes beyond CCSD(T) and tested for accuracy and performance. The number of resulting amplitudes will be rather small and this quantities are then much easier to handle. An automatic code generator is essential for efficient and fast implementation of these methods and will be a part of this project. The new methods will make it possible to calculate highly accurate ground-state energies, which are needed e.g. in thermochemistry and studies of chemical reactivity.