ALUMINA DISSOLUTION AND CURRENT EFFICIENCY IN HALL-HEROULT CELLS Bjørn Lillebuen 1, Marvin Bugge 1 and Helge Høie 2 1 Hydro Aluminium, P.O. Box 2560, NO-3908, Porsgrunn, Norway 2 Hydro Aluminium Karmøy, NO-4265, Håvik, Norway Keywords: Alumina, Current Efficiency (CE), Sodium, Dissolution Abstract The dissolution and distribution of alumina in cryolite melts can be described as a coupled heat and mass transport process, with intermediate formation of solid cryolite. Current efficiency (CE) can be evaluated by means of the rate equations for the back reaction between dissolved metal and carbon dioxide gas. Solid cryolite may be formed close to the metal pad under certain conditions in the cell. The bath superheat and the mass transfer coefficient at the bath/metal interface will be important parameters for the maximization of current efficiency. In some cells there is a clear correlation between current efficiency and the sodium content in the metal, indicating that mass transfer is the dominant factor. In other cells sodium levels can be quite low even at high current efficiency, which indicates that formation of solid cryolite can play a significant role, making superheat the dominating factor. 1. Alumina Dissolution Introduction When alumina is added to molten bath, it starts to dissolve. The dissolution process is endothermic, so heat needs to be transported to the dissolution interface by a driving temperature gradient. Since the bath temperature normally is just a few degrees above the bath liquidus temperature, the alumina dissolution may lead to local formation of solid cryolite. This solid cryolite will then gradually melt as more heat is tranported to the interface. In this paper we will present some operational data, and discuss correlations and links to the phenomena introduced above. 1. Alumina Quality Alumina Dissolution Alumina quality is often considered to be the main parameter for alumina dissolution control. Modern sandy alumina has quality variations that represent challenges for point feeding control, but more and more we have come to see that these variations may be less critical than the heat and mass transport conditions inside the cells. Successful implementation of point feeding has enabled computer control of the alumina concentration in the bath, and made possible a dramatic reduction in anode effect frequency and duration, in addition to the general performance improvement seen in new smelters. In order to improve from today s situation, a detailed understanding and description of alumina dissolution is required. 2. Dissolution of an Alumina Particle For illustration, a simplistic calculation of the dissolution rate of an alumina particle is shown. The particle is heated to bath temperature and dissolved, and Figure 1 gives an illustration of temperature and concentration gradients in the bath boundary layer ( film ) around the particle, and related to the bath phase diagram. 2. Aluminum Electrolysis The electrolytic deposition of molten aluminum takes place with sodium ions as charge carrier, and discharging of Al-containing species at the metal/bath interface. This sets up a concentration gradient with higher bath ratios at the interface than in the bulk of the bath. Higher bath ratio means higher liquidus temperature, so formation of cryolite may take place also at the cathodic interface between metal and bath. 3. Current Efficiency and Sodium The sodium levels in aluminum are determined by the bath ratio at the cathodic interface according to the equilibrium reaction: Al + 3 NaF = AlF 3 + 3 Na (1) From this same interface metal will dissolve and react with carbon dioxide by the so-called back reaction. The back reaction is assumed to be responsible for the largest part of the current efficiency losses in modern cells. In addition to Al, metals like Na and Li will also dissolve and back react, and dissolved metals may also react with impurities like phosphorous in the melt. Figure 1. Particle Dissolution. The following equations can be written for the dissolution rate, ref. Asbjørnsen and Andersen (1): M = k (c* - c) (2) Q = h (T T*) (3) 389
Q = M (H diss ) (4) From the figure it seems to last about 100 seconds for the temperature to climb back up to the value shown before feeding (T T*)/(c*- cs) = const (5) where M is dissolution rate, Q is heat transport rate, H diss is total dissolution enthalpy, and the concentrations can be interpreted by means of Figure 1. Solving these equations: where: M = k (a / (a+1)) (cs c) (6) a = (h/k) (const/h diss ) (7) Assuming that a is much larger than 1, then: M = k (cs c) (8) This means the normal rate equation as the product between the mass transfer coefficient k and the concentration gradient for the constant bath temperature case. Figure 2. Alumina dissolution rate, from Thonstad et al. (2). The differential equation for dissolution of the particle can be written: dt = const (Adr)/ (MA) (9) where A is area and r is radius. The mass transfer coefficient k can be estimated from the relation: Sh = 0.332 (Re) 1/2 (Sc) 1/3 (10) The dimensionless groups are estimated by inserting relevant physical data, and assuming an interfacial velocity of 5 cm/s. Integrating from t = 0 to t and r = r to 0 gives: t = 26573 (r) 3/2 (11) For a normal alumina particle with diameter 100 micron, r = 0.005 cm, then: t = 9.4 seconds (12) Thonstad et al. (2) have made careful laboratory experiments and measured dissolution rates in the range of 5-10 seconds, which is in good agreement with our very approximate calculation for a constant temperature dissolution. Some of Thonstad s data are reproduced in Figure 2. 3. Dissolution under Point Feeders Figure 3. Temperature drops during dissolution of 1.5 kg alumina doses, taken from Kobbeltvedt et al. (3).. In order to illustrate the thermal effects during dissolution, the approximate volume of bath needed to dissolve 1.5 kg alumina is calculated, assuming perfect mixing in the bath, no exchange with the surroundings, and with ten degrees of bath cooling. With reference to the works done by Bratland and Grjotheim (4) and Holm (5), the bath volume (x) can be written as a function of the temperature drop (dt): dtx = 494 (13) Kobbeltvedt et al. (3) have shown that bath temperatures measured near the feeder will drop several degrees every time the feeder adds 1.5 kg alumina, see Figure 3, taken from his work. or: x = 49 kg bath (14) 390
In words: 1.5 kg alumina will dissolve with a cooling effect corresponding to ten degrees cooling of 49 kg bath. Two comments should be made. Firstly, the calculation above was made for a bath with 2.3 wt% alumina. The dissolution enthalpy is strongly dependent upon alumina concentration as shown in Figure 4, taken from Holm s early work (5). Cryolite formation is therefore favored if the cell operates at low alumina concentrations in the bath. Secondly, it is conceivable that some of the alumina/cryolite mix can float on the metal surface, thus influencing on-going processes at the interface. This will be discussed in the following with a view on current efficiency and the sodium content in Al. where r is back reaction rate = metal dissolution rate, k is mass transfer coefficient for dissolved metal, A is metal/bath interface area, c Al is metal saturation concentration at the interface metal/bath, and f1 and f2 are functions of mass transfer parameters and solubility parameters for the metal/bath and for the gas/bath interfaces. If the gas bubble area (A b ) is very small, the following approximation can be written: r Al = k Al A b c Al (16) If instead the bubble area is very much larger than the metal area, then: r Al = k Al A Al c Al (17) For practical application in plant cells we may write finally: r Al = k Al A Al c Al (1 f ) (18) Here f can be treated as a cell calibration constant, and: CE = 100% (r r Al ) /r (19) when r is metal production for CE = 100%. It is possible to describe also the effect of bath impurities on current efficiency in a similar context. If the bath contains an oxidized impurity i n+ which reacts with dissolved metal in the bath boundary layer close to the metal surface, then: i n+ + (n / 3) Al = i + (n / 3) Al 3+ (20) This will enhance the metal dissolution rate, thereby reducing current efficiency, with an enhancement factor (e), given by: e = 1 + (n/ 3) (D i / D Al ) (c i /c Al ) (21) Figure 4. Partial Dissolution Enthalpy, from Holm (5).. Current Efficiency Current efficiency is reduced in a number of ways, for instance by electronic conduction, and by the presence of impurities like phosphorous. However, it is expected that current efficiency reductions mainly occur as a result of the back reaction between dissolved metal and carbon dioxide gas. Applying boundary layer diffusion theory (classical film theory) it is easy to show that: r Al = k Al A Al c Al (1 f1/f2) (15) Here D is diffusivity for i n+ and for dissolved Al, respectively, c i is concentration of i n+ in the bulk of the bath, and c Al is metal solubility in the bath at the metal/bath interface. Equation (21) can help explain why some multivalent impurities like phosporous can reduce the current efficiency. 1. Sodium in Al Sodium and Phosphorous in Al The well-known shift in bath composition passing from bulk bath through the boundary layer and on to the metal/bath interface, can be described by (6): C AlF3 (bulk) C AlF3 (interface) = I / nfk (22) where F is Faraday constant, I is the current density, k is the mass transfer coefficient, and C is the concentration. The high bath ratio at the interface will determine sodium levels in Al according to: 391
Al + 3NaF = 3Na + AlF 3 (23) In this way, sodium in Al will depend on mass transfer and current density, and less on bulk bath ratio, see Tabereaux (7). Older pots show lower sodium content, displayed in Figure 7, which is in line with the normal reduction of current efficiency with increasing age, shown in Figure 9. In Figure 5 the sodium in Al is plotted along with current efficiency for one side-by-side potline, as time series over several years. In spite of many different operational upsets, the covariation between sodium and current efficiency is obvious. Sodium will sometimes change before current efficiency, as expected since current efficiency numbers are based on aluminum tapping weights. No similar covariation has been found for older end-to-end and Soederberg lines. Sodium levels are normally lower in these older lines, even though current efficiency may sometimes be higher. Na ppm 120 100 80 60 40 Na Month CE 97 95 93 91 89 CE % Figure 7. Sodium in Al as a function of cell age (days). Excellent current efficiency values can be achieved with low noise levels, as indicated in Figure 8. Figure 5. Current efficiency and sodium content, monthly values. Results of a comprehensive study in one of our potlines are displayed in Figures 6-8. All cell numbers were averaged over one whole year, and a regression analysis was performed: Na (ppm) = -1271 + 14.4 (%CE) (24) This explained 84% of the data variation displayed in Figure 6.. Figure 8. Current efficiency and cell noise. In Figure 9 current efficiency is plotted versus cell age. These data have been collected some years later than data in Figures 6-8, but confirm the trend. Figure 6. Sodium in Al versus current efficiency, yearly cell data. 392
Current Efficiency (%) 98 97 96 95 94 93 92 91 90 89 0 500 1000 1500 2000 2500 3000 3500 Pot Age (days) Figure 9. Current Efficiency (one year average) and cell age. 2. Phosphorous in Al A detailed study for phosphorous, partly displayed in Figure 10, gave the following regression correlation: P (ppm) = 19.4 0.0052 (A - 1130) + 1.6 (CE 93.8) (25) A is cell age in days. This equation indicates that the phosphorous level in Al will increase with increasing current efficiency, but equation (25) could only explain 40% of the scatter. indicates that well-balanced magnetic fields (low k) as well as high current density (low A) should both increase current efficiency. As is well known, this is not always happening, and some of the complicating factors will be discussed in the following. Reduced turbulence will reduce k, but it will also increase c, since metal solubility will increase when the bath ratio increases, see Peterson and Wang (8). We have tried to quantify the relation between k and c, and our preliminary conclusion is that changes in k will be modified by changes in c, but the turbulence factor will always remain dominant. Electronic conduction is known (9) to increase in importance as bath ratio increases, but it is not clear if the interface ratio or the bulk bath ratio will determine the effect of electronic conduction on current efficiency Higher bath ratio at the interface will increase the possibility for cryolite to precipitate, thereby reducing the metal area and reducing back reaction rate. Obviously there will be a limit above which excessive cryolite formation can lead to erratic current distribution and even to bottom sludge. Cryolite can form directly at the metal interface, or during alumina dissolution. In both cases bath superheat will be an important parameter. Dewing (10) published the following empiric regression equation: Log (%CE) = 0.0095 dt 0.019%AlF3 (27) 0.06%LiF + const. Lower superheat (dt) will according to this equation lead to higher current efficiency, and cryolite formation may be a possible part of the explanation. High current density will increase heat production in the cell. In order to conserve energy, every pcell designer and cell operator will strive to reduce the anode-cathode distance (ACD). This will eventually increase turbulence, with deteriorating current distribution and increased gas release resistance. The reduced ACD may also result in more anode problems like anode spikes, and reduce the current efficiency. To some extent the low ACD is today compensated by multiple and deep slots in the anodes. Fig. 10. Phosphorous in Al versus cell age. A1 and A2 represent two different anode qualities with slightly different phosphrous content. Discussion The standard equation for back reaction rate: r Al = k Al A Al c Al (26) The high bath ratio at the metal interface will also tend to increase the Marangoni turbulence, which has been described by Utigard (11). Interfacial tension between bath and metal varies with the bath ratio, and this creates horizontal velocities in the bath boundary layer due to bath ratio gradients. The discussion of these phenomena also involves the slow transport of alumina (and fluoride) from the cell bottom to the electrode space, through the bath film assumed to surround the entire metal pool in the cell. The idea that cryolite is an active ingredient in the cell processes is not a new one. Recently, Solheim (12) modelled the conditions for cryolite formation, and Haupin (13, 14) has mentioned cryolite precipitation as a possible explanation for the so-called liqidus enigma found in many cells. The measured resistance hysteresis reported by Kvande and co-workers (15) can perhaps also be explained by means of cryolite precipitation. 393
In order to maintain excellent current efficiency and energy consumption at very high electrode current density in large cells, it may be necessary to optimize both alumina feeding (16) and the ACD in a more localized way than before, in order to control cryolite formation. Conclusion Alumina dissolution in cells with point feeders proceeds with intermediate cryolite formation. To melt, dissolve and distribute the alumina-cryolite mixture, sufficient turbulence (locally) is needed. High current density, low superheat and good magnetic stability may lead to precipitation of cryolite on the metal surface. 5. J.L.Holm, Thermodynamic Properties of Molten Cryolite and other Fluoride Mixtures, Inst. of Inorganic Chemistry, NTNU, Trondheim, Norway, 1971 6. J.Thonstad and S.Rolseth, Electrochimica Acta, 1978, Vol. 23, pp. 233-241 7. A.T.Tabereaux, Light Metals 1996, pp. 319-326 8. R.D.Peterson and X.Wang, Light Metals 1991, pp. 331-337 9. G.M.Haarberg et al., Light Metals 2002, pp. 1083-1089 10. E.W.Dewing, Met. Trans. B, Vol. 22, 1991, pp. 177-182 11. T.Utigard et al., Light Metals 1991, pp. 273-281 12. A.Solheim, Light Metals 2002, pp. 225-230 13. W.Haupin, Light Metals 1992, pp. 477-480 14. W.Haupin, Light Metals 1997, pp. 319-324 15. H.Kvande et al., Light Metals 1997, pp. 403-410 16. B.P.Moxnes, Light Metals 2009, in print. Some amount of solid cryolite formation can increase the current efficiency by slowing down the back reaction rate, but excessive amounts of cryolite could disturb current distribution, form bottom sludge and reduce current efficiency. Individual control of point feeders and anodes may benefit the development of large cells with high anodic current density, as discussed by Moxnes (16). Picture 1. HAL275 in Sunndal, Norway. References 1. O.A.Asbjørnsen and J.A.Andersen, Light Metals 1977, pp. 137-143 2. J.Thonstad et al., Met. Trans. 3, p. 403, (1972) 3. O.Kobbeltvedt et al., Light Metals 1996, pp. 421-428 4. D.Bratland and K.Grjotheim, Light Metals 1976, Vol. 1, pp. 3-21 394