The gas fluxing of aluminum: Mathematical modeling and experimental investigations

Chlorine fluxing is an essential purification step in aluminum refining in which impurities such as Ca, Na, Li, and Mg are removed by bubbling a mixture of chlorine and argon gas through molten aluminum. The gas is injected into the fluxing vessel through a rotating shaft and impeller which simultaneously agitates the melt, while breaking up and dispersing gas bubbles through the liquid phase. The efficiency of impurity removal and control of toxic chlorine and chloride emissions are dependent upon the extent of gas dispersion or mixing, residence time of the bubbles, and surface area of the bubbles. Clearly the gas injection and distribution within the liquid metal cannot be directly observed and such operations are often poorly controlled and not well understood. Problems arise when the injection gas, i.e. chlorine, is not completely consumed by reaction with impurities and the excess is reported as emissions of chlorides such as toxic HCl. The intention is to improve the technology to eliminate this waste (saving on the energy entailed in the chlorine production and reducing pollution) by better dispersion of the injected gas throughout the metal. Previous experimental investigations using a capacitance probe, capable of immersion in liquid aluminum for several hours, have been carried out to detect bubbles in an industrial fluxing unit at the Alcoa Technical Center. Bubble frequency data have shown the bubbles to be fairly well dispersed in the areas of the fluxing unit, decreasing in observed bubble frequency with increasing distance from the impeller (source of gas injection). To gain further insight and add to our experimental findings, two computational models have been developed to simulate the complex two-phase fluid dynamics of a rotary gas injection system. The results of these two modeling approaches are presented and analyzed and compared to experimental bubble measurements gathered using the capacitance probe. Bubble size distributions and residence times from the discrete phase model were incorporated in an external demagging reaction model to predict chlorine utilization efficiency. This simplified model included several assumptions regarding the kinetics and reaction path, however the model showed reasonable agreement to prior experimental magnesium removal data and provides valuable information related to the interplay of reaction progress in a fluxing unit and the fluid dynamics, in terms of bubble size, trajectory and resulting bubble residence time.