Electronic structure theory: Merging wave function theory, density functional theory, and machine learning.

点击翻译按钮获取中文摘要

Electronic structure theory provides the basis for our understanding of much of chemistry, biochemistry, material science, condensed matter physics, etc. It also provides the tools for computational modelling in the various areas mentioned. The proposed research aims at advancing electronic structure theory. In particular, we focus on density functional theory (DFT), where we address the exchange-correlation energy among electrons. This is the crucial component of Kohn-Sham DFT and it accounts for the complicated interactions between electrons. Recently, there has been enormous progress in this area which made Kohn-Sham DFT one of the most used tools in chemistry and the most cited topic in the history of the physical sciences. However, there remain numerous systems and properties for which the existing approximations and computational modelling tools are not reliable. Building on and continuing our previous work, our research program aims at deepening the understanding of the exchange-correlation energy and at improving upon existing exchange-correlation functionals. In the last decade, artificial intelligence and machine learning emerged as powerful tools that can be of use in virtually all areas of science and technology. The design of approximations and improved computational modelling tools in electronic structure theory is very challenging and we develop and adapt machine learning algorithms to perform an increasing share of the work. With machine learning, we can solve problems in the construction of approximations that are insurmountable for humans. Combining our experience in electronic structure theory with rapidly evolving artificial intelligence, we will provide improved computational tools, further simplifying the design of new materials and compounds. Furthermore, we intend to continue our work in the area of molecular electronics where molecules are connected to contacts to serve as electronic components and the current that passes through the molecules is measured as a function of the applied voltage. In the past, we developed simple models for molecular electronic devices (MEDs) that enabled us to predict new phenomena, some of which have been verified experimentally. In the proposed research we want to improve and extend our theories, in particular, we propose to develop theories that can describe the impact of electron-electron interaction on MEDs. While electronic structure theory has made enormous progress, most of its tools are not suitable for molecular conductors and with our work, we want to improve the description of exchange and correlation effects among electrons in MEDs.