New Density Functional Solution for Nondynamic and Strong Correlation

中田纳西州立大学（MTSU）的孔静教授获得了化学学部化学理论、模型和计算方法项目的奖励，为化学过程的计算机模拟开发了更精确的理论模型。计算机快速增长的能力使得通过模拟来预测化学反应变得更加可行。模拟与实验相结合，可以加速科学发现，但这只有在基础理论模型足够精确的情况下才有可能实现。目前流行的电子相互作用模型对化学的许多领域是准确的，但不是全部。它们不能准确预测的物理效应对于准确描述催化、光学材料、半导体和超导系统是重要的。该项目的目标是在基础层面建立更准确的理论模型。该项目的成功完成扩大了基于量子力学的模拟的范围，提高了化学和其他物理科学研究的质量和生产力。这项工作是在MTSU一个新的计算科学项目的背景下进行的，这是一个跨学科项目，为招收非化学背景的学生从事计算化学研究提供了一个独特的机会。MTSU的学生群体中，本科生、少数族裔和第一代大学生所占比例很高。本科生参与计算，对分子的活动进行基准测试，或对部分方法进行编码。目前正在开发一门新的高等本科生分子模拟课程，本项目的成果可用于新课程和其他本科课程工作。强相关问题是Kohn-Sham密度泛函理论（KS-DFT）的最后一个前沿问题。它是分子和材料系统DFT失效的主要原因。在这项工作中，基于单行列式KS-DFT的非动态/强相关性的精确处理正在发展。通用泛函处理强相关性的准确度与当前主流泛函处理弱相关分子和扩展系统的准确度相同。它的设计使其能够以足够的效率实现常规化学应用，准确处理动态和非动态相关性。为了实现这一目标，基于单行列式KS方案的非动态相关的一般框架正在开发中，但要符合确切的条件。这些条件是特定于包含退化的非动态相关，并作为开发新模型泛函的指导方针。与非动态和强相关的广泛基准数据库的开发以及有效实现的新算法一起，该函数显着拓宽了DFT在涉及所有强度电子相关的化学和材料科学问题中的应用。本项目所产生的软件和相应的功能通过开源程序和开源发布站点提供，以便与其他程序一起使用。源代码的可用性也有助于其他人开发新的DFT方法。这项研究预计将对MTSU新生的计算科学博士项目产生重大影响，该项目是全国为数不多的项目之一。

Professor Jing Kong of Middle Tennessee State University (MTSU) is supported by an award from the Chemical Theory, Models and Computational Methods Program in the Chemistry Division to develop more accurate theoretical models for computer simulations of chemical processes. The rapidly increasing power of computers is making it ever more feasible to predict chemical reactivity through simulations. Simulations, in concert with experiment, can accelerate scientific discovery, but this is only possible if the underlying theoretical model is sufficiently accurate. Current prevailing models for electronic interactions are accurate for many, but not all, areas of chemistry. The physical effects they fail to predict accurately are important for accurate descriptions of catalysis, optical materials, and semiconducting and superconducting systems. The goal of this project is to build more accurate theoretical models at the fundamental level. The successful completion of this project expands the scope of quantum-mechanical based simulations and improves the quality and productivity of research in chemical and other physical sciences. The work is being done in the context of a new Computational Science program at MTSU, which is an interdisciplinary program that provides a unique opportunity to recruit students from non-chemistry backgrounds to pursue research in computational chemistry. MTSU has a student body with high percentages of undergraduate, minority and first-generation college students. Undergraduate students participate in computations that benchmark the activity of molecules or by coding parts of the methods. A new molecular modeling course for upper-level undergraduates is being developed and the results of this project can be used for the new course and other undergraduate course work.The strong correlation problem is the last frontier of Kohn-Sham Density Functional Theory (KS-DFT). It accounts for most of the failures of DFT for molecular and material systems. In this work, an accurate treatment of nondynamic/strong correlation is being developed based on single-determinant KS-DFT. The general-purpose functional treats strong correlation to the same degree of accuracy as current mainstream functionals treat weakly correlated molecular and extended systems. It is designed so that it can be implemented with sufficient efficiency for routine chemical applications with accurate treatment of both dynamic and nondynamic correlations. To achieve this goal, a general framework for nondynamic correlation is being developed based on the single-determinant KS scheme, subject to exact conditions. These conditions are specific for nondynamic correlation with degeneracy included and serve as guidelines for the development of a new model functional. Together with the development of an extensive benchmark database for nondynamic and strong correlation, and a new algorithm for efficient implementation, this functional significantly broadens the application of DFT to chemical and materials science problems involving electron correlation of all strengths. The software and the corresponding features produced in this project are made available through open source programs and open source distribution sites such that it can be used with other programs. The availability of the source code also facilitates others to develop new DFT methods. This research is expected to have a major impact on the nascent Computational Science PhD Program at MTSU, one of a handful in the country.