Nonholonomic Ab Initio Molecular Dynamics for Computational Materials Design

E. Fahrenthold
University of Texas at Austin,
United States

Keywords: novel modeling methods, novel theory, computational materials design


Nanocomposites research attempts to design new materials with unique or optimized properties, by forming composites at the nanoscale. Previous work has demonstrated the ability to significantly modify the mechanical, electrical, magnetic, thermal, or chemical properties of various bulk materials by adding relatively small amounts of nanoparticles, nanorods, or nanotubes. Despite this success, the realization of dramatic improvements in engineered materials remains a difficult task. Understanding the thermal, mechanical, and chemical dynamics of fabrication processes is critical if the aforementioned potential is to be realized in macroscale devices. Example defense applications of current research interest include the development of new energetic materials, the development of high ampacity nanocomposite electrical conductions, and the development of sensitive and selective nanocomposite explosive sensors. The majority of published research on nanocomposites has taken an experimental approach. Simulation can serve as a valuable adjunct to experiment, reducing the time and cost of new materials development efforts. In recent work the author and co-workers have developed the first unified computational approach to the multiscale (molecular to macro) materials design problem. This paper extends the aforementioned work, formulating a new model of ab initio molecular dynamics. It differs from previous work in two important respects: (1) the quantum, molecular, and macro (process environment) scales are coupled using a nonholonomic Hamiltonian modeling methodology, and (2) the formulation rigorously satisfies the first and second laws of thermodynamics under conditions of general thermal and mechanical interaction with the external environment. The current work builds upon previous research, conducted by the author and co-workers, formulating and validating models for a variety of complex problems (at various scales and in various reference frames) using a nonholonomic Hamiltonian modeling methodology. Nonholonomic methods offer a canonical Hamiltonian approach to two fundamental modeling issues of wide interest: (1) systematic linking of multiple scales, and (2) systematic linking of multiple energy domains. Alternative holonomic approaches which appear in the literature are either limited in scope or introduce new stability or accuracy concerns. Examples include: (1) the introduction of a constant system energy assumption, used to link the quantum and molecular scales, (2) the introduction of fictitious momentum states, also used to link the quantum and molecular scales, and (3) the use of augmented Lagrangians as a means of describing environmental mechanical, thermal, or electromagnetic loading on the particle ensemble. The extended use of nonholonomic modeling methods, which have seen very limited application in computational chemistry, offers new and important research opportunities in computational materials design. This work has been supported by the Defense Threat Reduction Agency (grant number HDTRA1-13-0023) and the Office of Naval Research (grant number N00014-15-1-2693).