Abstract:
In the Hall-Heroult process for aluminium production, alumina is dissolved in a cryolitebased electrolyte, and direct current is passed from sacrificial carbon anodes through the electrolyte to the molten aluminium pad which acts as the cathode. The aluminium produced is deposited at the bottom of the cell, whilst bubbles of the carbon oxides are released from beneath the anodes into the electrolyte. The addition and subsequent electrolytic removal of alumina to and from the electrolyte contribute to process variation. Although the reaction in which alumina is consumed proceeds continually, alumina addition is generally a batch process. Further, alumma dissolution in cryolite is endothermic. The chemical composition and temperature of the electrolyte therefore change as alumina is added, dissolves and is consumed. The behaviour of alumina in Hall-Heroult cells, in particular the variation associated with its feeding, dissolution, dispersion and reaction, has been the subject of this investigation. Process variation has been studied in cells of three designs and with three different feeding technologies (breaker bar, point feed and continuous point feed), and mathematical models have been developed to predict aspects of alumina behaviour. Measurements of alumina concentration variation over feed cycles have shown that, when alumina additions are large and sufficient heat for dissolution cannot be supplied by the electrolyte, up to seventy percent of alumina does not dissolve upon addition. Alumina is gradually fed into the electrolyte from reserves of sludge (a slurry of undissolved alumina and electrolyte) beneath the metal pad, and from the alumina-based cell cover, at rates of up to half the cell's alumina consumption rate. If constraints to dissolution are alleviated by feeding alumina gradually into the electrolyte, a high percentage of the feed added is found to dissolve. Consequently, poor control of alumina feedrate can Iead to greatly increased variation in alumina concentration. Anode effects, which Iead temporarily to increased temperatures and turbulence in the electrolyte, also contribute to alumina concentration variation. On a breaker bar fed cell, the percentage of alumina fed which dissolves is higher during anode effect termination than during a normal feeding cycle. Alumina concentration m the electrolyte often increases as a result of an anode effect. Laboratory-scale alumina dissolution has been modelled as the heat-transfer controlled dissolution of an alumina-cryolite agglomerate, assuming one-dimensional conduction within the particle and convection at the particle boundary and using fmιte difference methods. The model successfully predicts the observed trends in thermal transient in the electrolyte as superheat and turbulence change. Variation in alumina concentration often leads to changes in electrolyte temperature and aluminium fluoride concentration. As alumina concentration falls during short-term nonfeed periods, measured changes in aluminium fluoride concentration show that liquid electrolyte mass can decrease by around five percent over periods of one or two hours, through the freezing of cryolite-rich ledge. The mass of liquid electrolyte in the cell is believed to vary by fifty percent or more in the medium-term, as electrolyte freezes and melts with changes in the cell heat balance. Electrolyte mixing, measured using a strontium carbonate tracer, has been found to occur relatively slowly. Even distribution throughout the electrolyte takes up to seventy minutes for the cell designs studied, and this contributes to the spatial variation in electrolyte temperature and composition observed within the cell. Flow pattems are not symmetrical about the cell axes. A model for electrolyte mixing, in which the two dimensional diffusion equations are solved, allows mixing rates in the cell to be characterised by an effective diffusivity parameter. This enables the number of point feeders required to distribute alumina satisfactorily to be estimated for different cell designs. A model has been developed which predicts anode current distribution, by considering the anodes and corresponding inter-electrode gaps as parallel resistors and solving the resultant set of simultaneous algebraic equations. Differences in anode age and in interelectrode distance are predicted to be the main factors affecting anode current distribution, and hence relative alumina consumption rates, in the cell.