CAREER: CDS&E: Nonlocal and Periodic Density Embedding

点击翻译按钮获取中文摘要

Michele Pavanello of Rutgers University, Newark is supported by an award from the Chemical Theory, Models and Computational Methods program to develop a novel computational chemistry methods and software. The goal of modern theoretical chemistry is to be able to predict properties of materials and molecular reactivity ahead of costly experiments. This has led to a reformulation of quantum mechanics called Density Functional Theory or DFT. Although current implementations of DFT find much broader applicability than any other quantum chemistry method, their application to realistically sized model systems is still problematic. This project tackles these problems head-on and explores an alternative to DFT called subsystem DFT. Subsystem DFT is a reformulation of DFT potentially capable of approaching much larger system sizes with no appreciable loss of accuracy of the simulations. There are three main goals of this project: 1) develop new subsystem DFT computer software, 2) apply the software to fundamental systems such as water, photovoltaic cells, and catalysis systems, 3) train students from underrepresented backgrounds in the fields of theoretical chemistry and computer coding. The impacts are to advance the accuracy and efficiency of theoretical chemistry methods and to use the new methods to better understand fundamental real-world systems while training the future workforce. The project centers on partitioning the electron density into subsystem contributions leading to a subsystem formulation of DFT. This results in a new computational framework based on a plane waves basis set capable of modeling semiconductors, conductors and bulk systems. Additional project activities include developing functionals of the kinetic energy and the exchange-correlation energy aimed at making subsystem DFT predictive and quantitative beyond semilocal and hybrid Kohn-Sham DFT. Specifically, the scientific impacts are computational / algorithmic advances with the goal of outputting a massively parallel code capable of exploiting the locality of electronic structures with an unprecedented efficiency and for a wide class of model systems.