A virtual crystallization furnace for solar silicon

Blocks of silicon for photovoltaic applications are economically crystallized in large casting furnaces. The quality of the material is determined by the velocity of the crystallization front, the flatness of the liquid-solid interface and the thermal gradients in the solid during cooling. The process cycle time, which is determined by the rate of crystallization and cooling, has a large effect on the process economic viability. Traditionally trial and error was used to determine the process control parameters, the success of which depended on the operator's experience and intuition. This paper presents a numerical model, which when completed by a fitted data set, constitutes a virtual model of a real crystallization furnace, the "virtual crystallization furnace" (VCF). The time-temperature distribution during the process cycle is the main output, which includes a display of actual liquid-solid front position. Moreover, solidification velocity, temperature gradients and thermal stresses can be deduced from this output. The time needed to run a simulation on a modern workstation is approximately 1/6 of real process time, thereby allowing the user to make many process variations at very reasonable costs. Therefore the VCF is a powerful tool for optimizing the process in order to reduce cycle time and to increase product quality.