NERSC Initiative for Scientific Exploration (NISE) 2011 Awards

First Principles Characterization of van der Waals Dispersion Interactions

Deyu Lu, Brookhaven National Lab

NISE project m1288

The proposed NISE will deliver a theoretical tool (modified version of Quantum Espresso) based on first principles, which can be used to understand and characterize material structures at the atomic level. Since the underlying methodology is very general, this tool can be applied to systems with different dimensionalities and to systems exhibiting different level of complexities, such as novel nanomaterials for renewable energy related applications.

Van der Waals dispersion interaction plays an important role in determining the atomic structure of various materials, for example, molecular crystals and organic/inorganic interfaces with promising potential applications for molecular detection and renewable energy generation. Theoretically, it remains a challenging research frontier to develop efficient first principles methods to treat vdW dispersion interactions, since typically the key element, the non-local correlation effect, is not incorporated in many standard approaches, such as density functional theory (DFT) under the local and semi-local approximations.

In this project, we will develop new theoretical models and computer programs based on the adiabatic connection fluctuation dissipation theory, which treat non-local correlation effect rigorously by evaluating the dielectric response function. Currently, the dielectric response function is normally obtained under the random phase approximation (RPA), i.e., at the same level of time-dependent Hartree. Although RPA leads to significant improvement over DFT, the absolute correlation energies in RPA are overestimated substantially, and the cohesive energies are slightly underestimated. Our goal is to address the above issues by introducing a non-local exchange-correlation kernel, which is neglected in RPA.

Previous model exchange-correlation kernels were mostly derived in the context of homogeneous electron gas (HEG). A few extensions have made to treat inhomogeneous systems. However, these methods either are computationally too demanding [1, 2] or violate some physical constrains seriously. Our goal will be achieved in two phases. First, we will generalize previous exchange-correlation models to inhomogeneous systems that can be computed efficiently and also obey proper physical constrains. Here, special attention will be made to design an Ansatz, where real space and reciprocal space variables are fully separated in order to achieve an efficient numerical implementation. Secondly, benchmark calculations will be carried out on bulk silicon and rare gas dimers to evaluate the performance of various model exchange-correlation kernel.

The success of the proposed NISE is strongly tied to the PI's expertise in studying vdW interactions. The PI is a main developer of a first principles RPA program that exhibits the best scaling performance, which is suitable for large-scale parallel computing [4,5]. This program is based on a modified version of Quantum Espresso, and has been used to study vdW interactions in a wide range of materials, including bulk silicon, rare gas dimers, molecular crystals, and self-assembled organic monolayers [5 - 7]. Our new model exchange-correlation kernel has been implemented to the RPA code, and preliminary results are promising. Work is in progress to refine the analytical form of our model to better reproduce the known results for HEG and to perform benchmark calculations for prototypical inhomogeneous systems.