Abstract
The solar cell industries have the prospect of a continued large marked increase in the years to come. However, to achieve this increase, efficient production of silicon wafers for cell production is deemed essential. A key process in this respect is the casting of multicrystalline silicon ingots, which subsequently are cut into wafers. Hence, detailed understanding of the casting process is paramount for the production of high quality wafers. Here, work is presented that was performed at SINTEF to accurately model the heat and melt transport in the SINTEF/NTNU Crystalox DS 250 lab-scale furnace, thereby understanding the modelling details necessary to perform simulations for industrial furnaces with the required predictive accuracy. Dedicated validation experiments have been performed, enabling us to build an axi-symmetric CFD model with predictive power, a model that can be applied with confidence to industrial furnace and process design where validation data are not readily available. Our CFD model was build using the commercial software Fluent 6.3; it includes all major heat transfer mechanisms, i.e. conduction, radiation, and natural convection in the melt. Results are transient field data for temperature, melt velocity, and liquid fraction. A furnace control program was implemented in user defined functions (UDF), ensuring the same regulation of the power input in the model as in the actual lab furnace. The discrepancy can to a large extent be explained by coil losses as the coil is not included in the model. A UDF model has also been implemented for the shutter in the heat gate. The shutter is essentially 3D, but calibrating a 2D model with varying transparency to radiation, we get a good approximation for the angular average of heat leak through the shutter. Also, we need to take into the account the actual distribution on the heating. In particular the residual heating of the bottom plate is important. Taking all into account and using the most a