Molecular dynamics simulations of brittle fracture in amorphous silica

Abstract

Fracture in brittle materials under a macroscopic load, results from the propagation of atomic-scale defects/cracks under the influence of a local stress field. These local stress fields are significantly higher than the macroscopic stress applied, causing local rearrangement of atoms around the crack tip and a consequent straining of atomic bonds that ultimately break, leading to separation of the material. The brittle fracture process has been a subject of many simulations and experiments, but the exact nature of atomic rearrangement that occurs in the regions of high stress has not yet been clearly identified. Thus, a primary objective was to accurately characterize the atomic restructuring in these critical regions. The method of molecular dynamics (MD), a widely used atomistic computation tool, was used to study the atomic-scale dynamics that take place during fracture of a typical brittle material---amorphous silica (a-SiO2). The interatomic interactions were represented by potential functions derived from first-principles. The effects of charge-transfer and temperature on the fracture process of a-SiO2 samples of different densities were investigated as a function of uniaxial strain-rates (0.1/ps-0.005/ps). A mechanism involving growth and coalescence of voids previously identified to underlie the process of brittle fracture was studied in detail in this thesis as a function of interatomic potential function, charge transfer and temperature. The regions surrounding these voids were found to be characterized by edge-sharing silica tetrahedra, while the rest of the material retained the bulk structure (corner-sharing tetrahedra) of silica glass. A secondary objective of this research work was to develop multiscale methodologies capable of modeling typical 'materials' problems like fracture. A global representation of the fracture process needs a seamless coupling of techniques capable of modeling different length and time scales. Specifically, far away from the critical regions, where the system is in elastic conditions, it is computationally prudent as well as scientifically elegant to use continuum-level simulation schemes like finite elements (FE) and finite difference time domain methods (FDTD) rather than atomistics, and only use atomistic simulations to model the highly strained regions. In this work, a continuum-FDTD region was coupled to an atomistic-MD region to study the propagation characteristics of a stress wave with broadband spectral features. The 'mismatch' in the coupling was quantified by analyzing the amount of reflection of the probing wave from the FDTD-MD interface. The above described work forms the basis for future fracture studies.