Computer simulation of nanoparticle sintering

Molecular Dynamics techniques were used to simulate Cu nanoparticle arrays at different temperatures. New potential was modified to provide the reliable representation of Cu nanoparticle systems. Different sets of simulation arrays were used to study the different stages of nanoscale sintering. The results are summarized as follows: (1) Three new mechanisms were found to contribute significantly to the early stage sintering. They are plastic deformation, mechanical rotation and amorphisation-recrystallization. Mechanical rotation, caused by large atomic forces relative to the particle masses, appeared to be almost independent of temperature. Plastic deformation was found to involved via dislocation generation and transmission and appeared almost athermal as well. Amorphisation of sub-critical grains was found to significantly increase the diffusion rates in the affected regions and dramatically accelerate the sintering kinetics and grain boundary kinetics. A high-speed plastic deformation, twinning, was found in the nanosphere sintering and resulted in sintered clusters with multiply twinned particles. (2) Only two of the six classical mechanisms (surface diffusion and grain boundary diffusion) were found to be important for nanoscale sintering. However, the processes were considerably accelerated by strong interatomic attractions, and depended only weakly on temperature. These “force-driven” diffusions are different from the standard “random walk” ones, and need to be treated quite differently. (3) Large array simulations are made possible by using pressure clamp technique. The effect of grain size distribution, temperature and pressure were investigated. Residual pores were found playing important role in fully densification. (4) Classical sintering laws, such as Herring\u27s law, were found invalid to describe nanoscale sintering due to the facts that a series of mechanisms co-exists and come intro play in an overlapping manner in a short period of time. (5) Sintering Diagrams, estimated from the simulation results, appear to be differently from the ones that are predicted by the classical theory.