Sintering is an important technological process in materials science and manufacturing, used to produce components from ceramic and/or metal powders. Sintering has been studied for a long time, but we only have accurate models for the simplest systems. Recent work at UQ has resulted in the development of a potentially more accurate model of sintering based on the simulation of atomic movements in the sintering sample. Such simulations are computationally intensive, so the systems which can currently be simulated (even with supercomputers) are orders of magnitude smaller than what is verifiable experimentally by materials scientists. As the size of the simulated system should be close to that of real sintering experiments, we face an overwhelming computational problem.

To tackle this problem, we propose the development of a specialised computer for the simulation of sintering based on solid state atomic movements. This custom computer will implement the algorithm directly in hardware, avoiding the overheads of classical sequential processors. The computer will support highly parallel computations and, by being based on electronically reconfigurable computing chips (FPGAs), will be easily reconfigurable to accommodate changes in the simulation algorithm and/or the simulated system. Our earlier work and other research in reconfigurable computing indicate that performance exceeding that of a supercomputer can be achieved at a fraction of the cost. The reconfigurable nature of the computer will also allow it to be applied to the simulation of other atomic level processes (e.g. creep, solid-state phase transformations and solidification).