Multiscale Simulation of Thermo-Mechanical Properties of Materials
The 21st century is witnessing an unprecedented growth of nanotechnology. The rapid evolution of this new science and its exciting applications have necessitated novel, sophisticated, and physically based approaches for design and performance prediction. The need for a high level of quantitative accuracy in prediction has motivated the development of ab initio (e.g., DFT) and atomistics-based modeling and simulation methods in material science and engineering. Among such methods, perhaps the most well-known is molecular dynamics (MD). Although extremely useful on its natural length and timescales, it represents a tremendous challenge to upscale MD-based simulation to significantly longer time and larger lengthscales. Even in the context of DFT and exascale-based optimism, it is highly unlikely that MD-based simulation will be able to tackle more complex problems involving physical phenomena operating across large ranges of scale. For example, 12 orders of magnitude in timescale as in the modeling of protein folding, or 10 orders of magnitude in spatial scales as in the design of advanced materials.
These limitations have motivated efforts in the research and development of methods which coarse-grain (CG) or upscale atomistic models to longer timescales and larger lengthscales. In particular, CG methods involve the projection of a larger number of ("faster", "shorter") atomic degrees-of-freedom (ADOFs; e.g., atomic displacement, vibrations, forces, energy) onto a smaller number of ("slower", "longer") continuum DOFS (CDOFS; e.g., deformation, stress, energy). A number of multiscale methods have been proposed to upscale MD in this fashion. Even so, development of physically sound and computable methods for the spatial and temporal coarse-graining of ab initio and / or atomistic models for materials as a means of extending these to much longer timescales and much larger lengthscales represents an on-going challenge.
The purpose of the current Ph.D. project is the detailed comparison and further development of two such methods, i.e., the quasi-continuum (QC) method and the hybrid molecular-cluster statistical thermodynamics (HMCST) method. An extension of QC to finite temperature ("hotQC" or "HQC") will be used in this compariseon and after theoretical and analytical investigation of the methods, benchmark computational comparisons of MD, hotQC and HMCST in 1- to 3-D systems based on simple LJ and complex embedded atom potentials are to be provided.
The final step deals with application of these methods to simulation of quasi-static compression, stretching, bending, and nano-indentation.