Although molecular dynamics (MD) simulation has been widely used in the study of physical properties of nanomaterials, it suffers from a well-known time scale limitation where it cannot, in the absence of special high-performance computing environments, simulate processes that take longer than about hundreds of nanoseconds. However, many important process in physics and materials science take place on time scales that cannot be reached by MD, and thus a key limitation of atomistic simulations is the ability to study the mechanical deformation of nanostructures at experimentally-relevant time scales. In this thesis, a new computational technique is proposed to overcome this issue. Specifically, a generic history-penalized self-learning metabasin escape (SLME) algorithm is developed which demonstrated high computational efficiency in potential energy surface exploration. The SLME method is then coupled with mechanical deformation to elucidate the strain rate and temperature effects on (1) the yield stresses of amorphous solids, (2) shear transformation zone (STZ) characteristics in two-dimensional amorphous solids, (3) shear transformation zones that nucleate at the free surfaces of nanostructured amorphous solids. The SLME method has also been utilized to study force-induced unfolding of the protein ubiquitin. In doing so, the long time scale atomistic simulations shed new insight into the role of amino acid sequence in protein unfolding dynamics. Specifically, it is found that the functional (binding) sites of protein can be responsible for protein unfolding dynamics, instead of only performing biological functions as previously thought. |