T.R. Barton, E.J. Frazer and A.M. Vecchio-Sadus

CSIRO Minerals

Box 312, Clayton South VIC, 3169, Australia


Four samples of agglomerated alumina superfines, to be used as smelting feedstock, were produced using spray drying. The agglomerates were formed using either as-received or washed superfines (to reduce soda content) with the addition of either 5 or 10 wt% aluminium hydroxychloride binder. The feedstock formed from unwashed superfines and 10 wt% binder had an attrition resistance comparable with that of Smelter Grade Alumina (SGA). A segregation test on a mixture of 95 wt% SGA:5 wt% agglomerated alumina superfines showed that only a minimal degree of segregation of agglomerated superfines from SGA can be expected. The calcined agglomerated superfines feed contained residual chloride and this resulted in elevated levels of chloride in the cell fume compared with the 100% SGA feed.

A laboratory-scale aluminium smelting cell was used to establish whether the presence of the binder and other inorganic impurities in the agglomerated superfines had any detrimental effect on metal quality and current efficiency (CE). Smelting was conducted with agglomerated alumina superfines levels at 5 and 100 wt% of the alumina feed. Aluminium was deposited at a current density of ~0.75 Acm-2 for 1-6 h at 965°C from an electrolyte similar in composition to that used in industrial cells. CEs of >95% were obtained in high purity electrolyte using 100% SGA as feed, and there appeared to be no significant reduction in CE with an agglomerated superfines feed. Trace metal analyses revealed no significant difference in the purity of aluminium deposits produced from the various feedstock compositions. Overall, the inclusion of superfine agglomerates at the 5 wt% level in SGA did not appear to adversely affect the performance of a laboratory-scale aluminium smelting cell.


alumina, aluminium, agglomeration, smelting, superfines


T.R. Barton, E.J. Frazer and A.M. Vecchio-Sadus

1.0 Introduction

The fine aluminous material produced in alumina refineries (superfines dust) is currently redigested or blended back (Syltevik et al., 1996) into the Smelter Grade Alumina (SGA). It is vital to smelter productivity that SGA properties remain within well defined limits, in particular, with respect to the minimisation of fines content. High quality SGA contributes to smelter efficiency by reducing degradation, segregation and emissions, while maintaining dissolution rate. The study outlined in this paper considers superfines dust as a secondary resource and seeks to demonstrate that the utilisation of superfines dust, as a smelter feedstock component, does not adversely affect the aluminium smelting process.

The following sections describe the performance of various blends of SGA and agglomerated superfines as potential feedstocks for aluminium smelting. The studies were conducted in a laboratory-scale cell with an electrolyte having a composition similar to that employed industrially, and containing agglomerated superfines at 5 and 100 wt% of the feed. The principal objective was to establish whether the presence of the binder and other inorganic impurities in the agglomerated superfines had any detrimental effect on metal quality and current efficiency (CE).


2.1 Agglomerate Preparation

Four samples of agglomerated alumina superfines (see Fig. 1), to be used as a smelting feedstock component, were produced using spray drying (Barton et al., 1993; Hall et al., 1996) with the addition of aluminium hydroxychloride (Reheis, 1991). Two of the feedstocks used alumina superfines that had been washed using a method similar to that described in the patent literature (Grocott, 1995) to ensure that the soda (Na2O) content was well within the range specified for SGA (Allais et al., 1996).The feedstocks produced are shown in Table 1.

Table 1

Samples of agglomerated alumina superfines produced for use as smelting test feedstocks


Nominal Binder Level
















2.2 Physical Characterisation of Agglomerates

The agglomerates were subjected to attrition tests as a guide to the effects of the pretreatment of the alumina superfines (to reduce soda content) and of changes of binder level on agglomerate strength. Other characterisation tests included a brief assessment of the segregation of agglomerated alumina superfines from SGA, and measurement of residual chloride levels in the agglomerated material after calcination.

Fig. 1

SEM photomicrograph of agglomerated alumina superfines.

2.2.1 Attrition

A measure of the level of attrition was obtained by subjecting samples of agglomerated alumina superfines (-212 m m/+53 m m) to a series of simple, timed sieving tests using a set of Ö 2 stackable sieves and a mechanical sieve shaking device ("Rotap"). Duplicate samples of green, dried (150° C overnight) or calcined (900° C for 2, 4 or 6 h) agglomerates were screened for either 5 or 30 minutes and a particle size distribution obtained. The attrition value (AV) was determined as follows:

where x = % -53 m (30 min) and y = % - 53 m (5 min)


The degree of agglomerate segregation that may occur in a mixture of SGA and agglomerated alumina superfines was assessed using a variation of a published method (Enstad, 1997). The agglomerated materials were differentiated from the white SGA by adding a small amount of red dye during the agglomeration process, thereby colouring the product. The coloured agglomerates were then mixed with the SGA in the ratio of 5:95 on a mass basis.

The homogeneous mixture of agglomerated alumina superfines and SGA was poured into the two-dimensional test unit (see Fig. 2). The tester was constructed with a clear perspex front wall and a stainless steel back wall and was equipped with nine tube-type samplers situated along a straight line parallel to the angle of repose of the SGA/agglomerated alumina superfines mixture. Samples of the mixture were obtained by pushing the samplers through guides in the back wall of the tester.

The resultant nine samples were split into quadruplicate sub-samples that were analysed for colour using image analysis. Clear plastic slides were covered with a monolayer of alumina particles obtained directly from a riffler. Forty images were collected from each slide; each image contained approximately 200 particles. The slides were viewed under white light from both transmitted and reflected sources on a Zeiss Axiophot microscope. A Sony 930P colour camera was adjusted to enhance the colour difference between SGA and agglomerated alumina superfines, and the colour image recorded. Important structural information was contained in the red and green channels which were processed using a mathematical transformation of the local shape characteristics. Results were collected as an area percentage. The densities of the SGA and agglomerated alumina superfines constituents were used to convert the data to a weight percentage, and differences in the ratio of the weight percentages between the nine sets of samples were used to evaluate the degree of segregation that had occurred along the slope.

Fig. 2

Schematic diagram of two-dimensional segregation tester

2.2.3 Chloride removal

Sub-samples of agglomerated alumina superfines were statically calcined by either: (i) heating to 900° C in air, using an electric muffle furnace, for periods of up to six hours, or (ii) heating to 600° C in humidified air, using an electric tube furnace, for periods of up to seven hours. The level of residual chloride remaining in the agglomerates after calcination was determined quantitatively by XRF and these values were compared with both the theoretical and actual chloride levels of the agglomerated alumina superfines prior to calcination. The theoretical chloride levels were calculated based on the nominal binder content of the superfines/binder mixture and the specified chloride content (16.3 wt%) of the binder [Al2(OH)5Cl.2H2O] (Reheis, 1991), and are reported as a percentage of the a-alumina content.

2.3 Smelting

2.3.1 Electrochemical Cell and Furnace Design

Anodes were fabricated from unsheathed graphite rod (Morganite EY941; diam. 25 mm) with a 5° taper at the tip. The molybdenum cathode (Metallwerk Plansee GmbH, 99.95%; diam. 25 mm) had a complementary conical cross-section with a 170° included angle. Electrodes were positioned at a nominal anode-cathode distance of 3 cm. The multi-element graphite electrochemical cell used here (see Fig. 3) was designed for previous laboratory studies of aluminium smelting (Dorin and Frazer, 1993).

The graphite cell was contained in an Inconel 600 furnace tube which was purged with argon to minimize oxidation. An electrode rotator (Pine Instrument Company, Model AFASRE) was mounted above the furnace and fitted with an extended Inconel arbour to which the anode was attached; the assembly entered the cell through a water cooled seal and bearing in the lid. The unit was heated by an 8 kW Globar (The Carborundum Co.) furnace with appropriate power and temperature control units. The cell was equilibrated for at least one hour after the furnace had attained its working temperature. The electrolyte temperature was measured with a Type-K chromel-alumel thermocouple.

Figure 3

Schematic of multi-element grahite cell: (a) outer graphite sheath; (b) inner graphite sheath; (c) molybdenum cathode; (d) retaining screw; (e) outer graphite pot; (f) rotator mount.

2.3.2 Electrolyte composition

The electrolyte (total 300 g) was composed of vacuum-dried (120° C, 4 h) synthetic cryolite (Japanese; ex. Alcoa, Pt Henry), smelter grade alumina (SGA; ex. Alcoa), agglomerated alumina superfines, aluminium fluoride (CERAC Inc., 99.9%), and calcium fluoride (Aldrich, 99.9%). The starting electrolyte composition was fixed at 79 wt% Na3AlF6, 10 wt% AlF3, 5 wt% CaF2 and 6 wt% Al2O3 (SGA and/or agglomerated alumina superfines). The electrolyte had a composition similar to that used in industrial cells (12.5% AlF3 in excess of the cryolite composition; bath ratio (defined as wt NaF/ wt AlF3) of 1.14) (Grjotheim and Kvande, 1993a). Experiments were run in duplicate at 965° C.

High purity electrolyte (alumina free) of the above composition was synthesized by melting measured quantities of sodium fluoride (BDH, AnalaR), aluminium fluoride (CERAC Inc., 99.9%) and calcium fluoride (Aldrich, 99.9%) under argon at 1000° C over 1 h. XRF analyses of various electrolyte components are shown in Table 2. Note the very low level of SiO2 in the high purity electrolyte compared with synthetic cryolite.


Table 2

XRF analysis (wt%) of electrolyte components. (Note: elements reported as oxide equivalents.)




Synthetic Cryolite

High Purity Electrolyte



SGA (as received)

Calcined Agg. Superfines (average)



<0.01 <0.02 <0.0165 <0.005 <0.005



0.32 <0.01 <0.013 <0.006 <0.006


  <0.01 <0.01   0.001 <0.001


  0.02 0.02   <0.008 <0.008



<0.01 3.7 0.02-0.04 <0.006 <0.006


  41.4 35.5 0.3-0.4 0.40 0.36***


  <0.01 <0.01 <0.02 <0.001 <0.001



<0.01 <0.01 <0.005 0.005 0.013



0.01 <0.01 <0.0012 <0.002 <0.002



0.06 <0.01   <0.002 <0.002


  0.02 <0.01   <0.001 0.29


  <0.01 <0.01   <0.002 <0.002

* (Grjotheim et al., 1982); ** (Allais et al., 1996); *** Na2O equivalent in washed fines was 0.18 wt%.

Note: Na content in synthetic cryolite and high purity electrolyte reported as oxide.

2.3.3 Smelting Procedure

The graphite anode was gradually lowered into the electrolyte over a period of 5 min until it had reached the operating inter-electrode distance. Smelting was conducted at a nominal current density of ~0.75 Acm-2 for 1-6 h using a scanning potentiostat (EG&G PAR, Model 362) coupled to a current booster (EG&G PAR, Model 365). The cell voltage and current were monitored continuously on a Yokogawa HR1300 hybrid recorder.

At the termination of the electrolysis, the anode was removed to the cool zone of the furnace and the electrolyte allowed to solidify. The aluminium metal was collected for analysis. Current efficiencies were determined from the weight of the aluminium product using Faraday's law. The weight of aluminium was corrected for the molybdenum dissolved from the cathode substrate (up to ~5 wt% at 965°C); no correction for the co-deposition of sodium was made.

2.3.4 Analytical methods

The aluminium product was reduced to small fragments and the entire sample taken for analysis, as local molybdenum enrichment at the cathode interface may bias the results. Impurity metal concentrations (Mo, Si, Na, Ti, Cu, Fe, Mn, V, P, Zn) in the recovered aluminium product were determined by ICP or AAS, as appropriate (relative error <5% for Mo, Na; relative error 10-20% for Ti, Cu, Fe, Mn, Si, V, P, Zn). Chloride in the cell fume was determined by potentiometric titration (relative error of <5%).


3.1 Attrition

Agglomerates produced from all four feedstocks were dried (150° C overnight) and calcined (900° C for 2 h). In addition, samples of AGG3 agglomerates were calcined for periods of 4 and 6 h at 900° C. The attrition values obtained for green, dried and calcined agglomerates are given in Table 3; these values have a relative standard deviation of ± 10%.

Table 3

Attrition values (%) and soda levels (wt%) for smelting test feedstocks




AV (%) Following Thermal Treatment




150° C (O/N)

900° C

(2 h)

900° C

(4 h)

900° C

(6 h)


SGA 0.40




5 wt% binder 0.36 7.0 10.2 13.2




5 wt% binder, washed 0.18 15.1 13.4 18.1




10 wt% binder 0.36 1.2 1.3 3.7




10 wt% binder, washed 0.18 27.9 24.7 21.8



* As received, ex-calciner.

Table 3 shows that the attrition value obtained for SGA (benchmark) was 1.3%. Only the attrition values obtained for AGG3 green and dried (150° C) equalled the benchmark. A comparison of the attrition values obtained for agglomerates prepared using as-received alumina superfines (AGG1 & AGG3) demonstrates that, for each given thermal treatment, the attrition value increases with decreasing binder content. The higher attrition values for AGG4 may have been caused by a pH control failure during the spray drying/slurry preparation.

It appears that prolonged time (>2 h) at temperature (900° C) has little effect on the attrition values of agglomerates produced from the AGG3 feedstock. This observation may be significant when considering longer-term thermal treatment to reduce residual chloride content. The attrition values (see Table 3) increased after calcination at 900° C for all feedstocks except AGG4, with only those obtained for agglomerates formed from AGG3 being comparable with those of SGA. Based upon the attrition values alone, feedstocks AGG1, AGG2 and AGG4 would be less suitable for blending with SGA.

3.2 Segregation

Data obtained via image analysis of samples taken from the segregation tester are shown in Table 4.

Table 4

Mass (wt%) of agglomerated alumina superfines contained in sub-samples taken from segregation tester as determined by image analysis

Sample ID










Fraction Agg. Alumina Superfines (wt%)










Average 5.15; Std Dev 0.25

In an ideal case where the mixture of SGA/agglomerated alumina superfines remains homogeneous, the fraction of agglomerates contained in any sub-sample should equal 5.0 wt%. A comparison between the data in Table 4 and this figure shows little variation, with the test data generally lying within two standard deviations of the mean. Overall, these results indicate that, where measures such as static mixers are in place to maintain homogeneity prior to bulk storage, segregation of agglomerated alumina superfines from the SGA would not be expected.

It should be noted that samples S2 and S3 (4.68 and 5.58 wt%, respectively) were more than one standard deviation either side of the mean suggesting that a segregation mechanism may be at work in this region of the tester. During the filling of the tester, it was observed that layers of material flowed down the side of the heap in the tester and this suggests that a segregation mechanism known as "sieving" (Enstad, 1997) may be operating. However, the observed flowing layers may also be caused by the difference of particle shape between that of the round agglomerates (see Fig.1) and the hexagonal SGA, with the agglomerates having a greater tendency to "roll" down the side of the heap and hence segregate from the SGA. At this level of investigation, either segregation mechanism seems possible.

3.3 Chloride Removal

Ideally, any chloride contained in the agglomerates would be removed during calcination prior to smelting. The calcination step is necessary to reduce the loss on ignition (LOI) of the green agglomerates, from ~2.5% to the level set in the specifications for SGA (~0.7%) (Syltevik, 1996). A specification for chloride level in SGA was not found during a brief literature search. Nevertheless, the behaviour of chloride during calcination and smelting is of interest from process engineering and occupational health and safety viewpoints.

3.3.1 Results of calcination experiments

Sub-samples of the four agglomerated feedstocks were calcined at 900° C for 2 h, and additional sub-samples of AGG3 were calcined at 900° C for 4-6 h. Sub-samples of green agglomerates, together with corresponding samples of calcined agglomerates, were analysed for chloride using XRF. The results of these analyses are shown in Fig. 4.

Fig. 4

Chloride levels (wt%) in green and calcined agglomerates

A comparison of the theoretical levels of chloride with those found in the green agglomerates for each feedstock suggests that a small amount of chloride is lost by hydrolysis during the agglomeration process. After calcination of all four feedstock samples at 900° C for 2 h, the residual chloride levels were remarkably constant at 0.29± 0.05 wt%, irrespective of the initial chloride level in the green agglomerates. Chloride lost at this stage would be via hydrolysis with water provided by the dehydroxylation of gibbsite contained in the superfines. Further calcination of the AGG3 feedstock at 900° C for 4-6 h reduced the residual chloride level to a minimum of 0.17 wt%. This result indicates that prolonged calcination in air at temperature is not a viable option to reduce residual chloride levels; a certain amount of chloride apparently remains in the agglomerate structure after calcination and must diffuse to the surface for removal by hydrolysis.

The hydrolysis mechanism for chloride removal was generally confirmed by calcination of a further sample of AGG3 agglomerates in a stream of humidified air at 600° C for ~7 h. A reduction of residual chloride level to 0.19 wt% was achieved at this lower temperature because of the more favourable hydrolysis conditions. Further reduction in residual chloride should be possible by optimization of calcination/hydrolysis conditions.

3.3.2 Chloride in cell fume

The calcined agglomerated alumina superfines contained residual chloride and this resulted in the appearance of elevated levels of chloride in the cell fume (~0.5 wt%) compared with the 100% SGA feed (~0.1 wt%). Interestingly, much lower chloride levels were obtained in all cases when using high purity electrolyte (<0.1 wt%). This may partly reflect the lower chloride content in the high purity electrolyte compared with synthetic cryolite (see Table 2). In addition, the high purity electrolyte was pre-fused at 1000° C, as opposed to vacuum drying at 120° C for the synthetic cryolite, reducing the water available for subsequent formation of HCl.

3.4 Current Efficiency

The aluminium metal products were collected to determine CE and metal quality. The CE values obtained with synthetic cryolite varied over a wide range (68-89%). The CEs obtained when using high purity electrolyte (see Table 5) were similar to those in industrial cells (³ 92 wt%) (Grjotheim and Kvande, 1993b). Unfortunately, the variability in CE (± 2.5%) experienced from run to run was much higher than normal. Thus, the better of the duplicate results is quoted here and no significance can be ascribed to the apparent variation in CE between feedstocks. Overall, the present results do not suggest any variation in CE with feedstock, but this would need to be verified by longer term, larger scale tests.

Table 5

Comparison of Ces (%) after 6 h smelting for various feedstocks in high purity electrolyte


100% SGA

95% SGA:5% AGG2

100% AGG2

CE (%)





3.5 Aluminium Product Quality

Trace metal analyses revealed no significant difference in the purity of aluminium deposits produced from the various feedstock compositions (see Table 6). In general, when detection limits were sufficiently low, the impurity levels were observed to decrease with smelting time in the 1-4 h range. For example, for an electrolyte containing 100% SGA feed, the copper content decreased from 0.07 wt% to ~0.01 wt%. The aluminium product obtained from high purity electrolyte had much lower levels of silicon. This is to be expected since the high purity electrolyte contained significantly less silicon than synthetic cryolite (see Table 2).

Table 6

Trace metal concentrations (wt%) in aluminium product after 6 h smelting from various feedstocks (synthetic cryolite and high purity electrolyte)


Using synthetic cryolite

Using high purity electrolyte *

(typical range)**

100% SGA

95% SGA:

5% AGG2

100% AGG2 *

100% SGA

95% SGA:

5% AGG2

100% AGG2







0.03 0.04







0.006 0.006







0.004 0.002







0.01 0.02







0.004 0.004







0.12 0.06







<0.001 <0.001







0.007 0.006







0.002 0.002

* Change in detection limits; ** (Kirk-Othmer, 1992).


  • At a laboratory scale, inclusion of agglomerated alumina superfines in SGA is not detrimental to cell current efficiency and product metal quality.
  • The cell fume from smelting tests (synthetic cryolite) which used agglomerated alumina superfines as a feed had elevated levels of chloride compared with tests using 100% SGA.
  • Alumina superfines agglomerated using 10 wt% binder (Al2(OH)5Cl) have an attrition resistance comparable with that of SGA, as indicated by the Rotap attrition test.
  • Reduction of soda content of the alumina superfines prior to agglomeration results in a lower resistance to attrition.
  • Longer agglomerate (AGG3) calcination times at 900° C appear to have minimal effect on attrition resistance.
  • Segregation of a smelting feedstock into its components (SGA and agglomerated alumina superfines) is unlikely to occur if segregation prevention mechanisms are employed during bulk handling of the mixture (e.g., static mixers, stopper plates).
  • Static calcination in air of agglomerates (AGG3) resulted in a residual chloride level of 0.17 wt%, even after 6 h at 900° C. This is equivalent to <0.01 wt% in a 95 wt% SGA:5 wt% agglomerated alumina superfines feed. Similar residual chloride levels were obtained by calcination of agglomerates (AGG3) for 7 h at 600° C in humidified air. The calcination/hydrolysis conditions could be further optimised to lower residual chloride levels in agglomerated alumina superfines.
  • Further work should be conducted on a larger scale to:
  • Confirm observed effects on CE and metal quality
  • Determine maximum allowable levels of alumina superfines agglomerates in feedstock
  • Check tendency to form sludge in the smelter bath
  • Determine deportment of residual chloride in smelting cell
  • Investigate longer term effects on electrolyte composition
  • Assess bulk handling characteristics of blended feedstocks


We wish to acknowledge Alcoa of Australia Ltd for support of the agglomeration process development. We thank J.F. Kubacki for cell maintenance, B. Jenkins for segregation tests, the Minerals Chemical Analysis Service, and Alcoa of Australia Ltd for supplying some electrolyte materials.


Allais, B., Lillebuen, B. and Larsen, J.A. (1996). Alumina quality: Smelters experience with fluctuating and out of bound alumina: point feeding and side break technologies. Technical, environmental and other costs to a smelter when using out of specifications alumina. Fourth International Alumina Quality Workshop, Darwin. pp 9-19.

Barton, T.R., Robson, B.J., and Hall, J.S. (1993). Agglomeration of alumina material. Australian Patent No. 664328 & US Patent Application No. 08/290950.

Dorin, R. and Frazer, E. J. (1993). Operational characteristics of laboratory scale alumina reduction cells with wettable cathodes. J. Appl. Electrochem., 23, 933-942.

Enstad, G.G. (1997). Segregation of powders and its minimisation. Second Israeli Conference for Conveying and Handling of Particulate Solids, Jerusalem.

Grocott, S.C. (1995). Process for the production of a low soda content smelting grade alumina. Australian Patent Application No. 33127/95.

Grjotheim, K. and Kvande, H. (1993a). Introduction to Aluminium Electrolysis: Understanding the Hall-Héroult Process, 2nd ed., Düsseldorf, Aluminium-Verlag. p. 228.

Grjotheim, K. and Kvande, H. (1993b). Introduction to Aluminium Electrolysis: Understanding the Hall-Héroult Process, 2nd ed., Düsseldorf, Aluminium-Verlag. p. 238.

Grjotheim, K., Krohn, C., Malinovskı, M., Matiašovskı, K., and Thonstad, J. (1982). Aluminium Electrolysis, 2nd ed., Düsseldorf, Aluminium-Verlag. p. 363.

Hall, J., Barton, T., and Robson, B. (1996). On agglomeration of alumina calcination fines. Fourth International Alumina Quality Workshop, Darwin. pp 567-572.

Kirk-Othmer (1992). Kirk-Othmer: Encyclopaedia of Chemical Technology, 4th ed., NY, John Wiley and Sons. Vol. 2, p. 212.

Reheis (1991). Product Data (Chlorohydrolâ Powder). p. 2.

Syltevik, D., Lillebuen, B. and Larsen, J.A. (1996). Alumina quality – user perspective. Fourth International Alumina Quality Workshop, Darwin. pp 21-31.