WC grain growth during sintering of cemented carbides : Experiments and simulations

Abstract: Cemented carbides are composite materials consisting of a hard carbide and a ductile binder. They are powdermetallurgically manufactured, where liquid-phase sintering is one of the main steps. The most common cemented carbide consists of WC and Co and it is widely used for cutting tools. Two of the most important factors controlling the mechanical properties are the WC grain size and the grain size distribution and thus it is of great interest to understand the grain growth behavior.In this thesis the grain growth during sintering at 1430 °C is studied both experimentally and through computer simulations. The grain growth behavior in cemented carbides cannot be explained from the classical LSW-theory. The WC grains have a faceted shape necessitating growth by 2-D nucleation of new atomic layers or surface defects. A new model based on 2-D nucleation, long-range diffusion and interface friction is formulated.Three powders having different average sizes are studied and both experiments and simulations show that a fine-grained powder may grow past a coarse-grained powder, indicating that abnormal grain growth has taken place in the fine-grained powder. Fine-grained powders with various fractions of large grains are also studied and it is seen that a faster growth is obtained with increasing fraction of large grains and that an initially slightly bimodal powder can approach the logaritmic normal distribution after long sintering times.The grain size measurements are performed on 2-D sections using image analysis on SEM images or EBSD analysis. Since the growth model is based on 3-D size distributions the 2-D size distributions have to be transformed to 3-D, and a new method, Inverse Saltykov, is proposed. The 2-D size distribution is first represented with kernel estimators and the 3-D size distribution is optimized in an iterative manner. In this way both negative values in the 3-D size distribution and modifications of the raw data are avoided.