PRECIPITATION OF GIBBSITE: DEVELOPMENT OF A NEW RATE EQUATION

R.M. Cornell, D.S. Pannett, N.S. Sullivan, P.C. Clarke and C.M. Bailey

Alcoa of Australia Limited. Cockburn Road, Kwinana 6167. Western Australia

ABSTRACT

The equation used by Alcoa (derived from the King equation) to describe gibbsite precipitation in Bayer liquors is applicable only over a narrow range of conditions. Outside of this range the rate "constant" is strongly dependent on Total Caustic (TC) and temperature.

A new rate equation that applies over a far wider range of TC (100-300 g/L) and temperature has now been developed using synthetic liquor. This equation is based on supersaturation ratio instead of supersaturation difference. The kinetics experiments were carried out with a carefully prepared, non-agglomerating seed with smooth, well grown, low index faces. Use of this material ensured that the surface area of the seed remained constant during the experiment.

Before the main kinetics experiments could proceed, considerable background work into the various factors influencing the precipitation induction time had to be carried out. Both the results of this work and the new rate equation itself, are presented in this paper.

KEY WORDS:

rate equation, induction period, kinetics, precipitation, gibbsite

PRECIPITATION OF GIBBSITE: DEVELOPMENT OF A NEW RATE EQUATION

R.M. Cornell, D.S. Pannett, N.S. Sullivan, P.C. Clarke and C.M. Bailey

1.0 INTRODUCTION

Precipitation of gibbsite is the central chemical reaction in the production of alumina. Various models to describe the precipitation of gibbsite under Bayer conditions are available (cf. Veesler and Boistelle, 1994). Alcoa utilises the following relationship to describe the rate of precipitation:

(1)

where:

C = [Al(OH)4-] at time t and

C* =[Al(OH)4-] at equilibrium.

This equation (which is derived from the King equation), works satisfactorily under current conditions, but it clearly does not adequately describe what is taking place because the rate constant (ko), which should be constant, decreases with rising caustic concentration (Kirke, 1981; Veesler and Boistelle, 1994). As increasing the caustic concentration can increase plant yield, a higher caustic level is a goal to which the refineries are moving. When this is achieved however, the current rate equation will no longer be applicable. Hence, a new rate equation that applies over a wide range of conditions is required.

The aims of this investigation were twofold:

1. To develop in synthetic liquor a precipitation equation that would be applicable over a wide range of caustic levels and temperatures.

2. To identify the causes of the induction period and ways of eliminating it.

The precipitation of gibbsite consists of two stages. There is an initial slow stage, termed the induction period, during which very little precipitation takes place (Misra and White, 1971; Brown, 1972; Haflon and Kaliguine, 1976; Smith and Woods, 1993) followed by the main precipitation reaction that is very much faster. The rate equation applies to this second stage. Earlier tests in synthetic liquor carried out at relatively high growth rates suggested that the induction time is eliminated if there is no calcium in the liquor (Cornell et al., 1996). The current experiments in "calcium free" liquor, however, covered a range of experimental conditions and in some preliminary tests, an induction period as long as 21 hours was noted. It was felt that before the main rate equation experiments could proceed, it was necessary to get further information about the factors influencing the induction time. To this end, experiments to examine the effect on induction period of seed surface roughness, solids loading, crystal morphology and cations in solution were carried out.

 

1.1 The Rate Equation

The thermodynamic driving force for crystallization is the difference in chemical potential (D m ) between the solution and the solid phases. This chemical potential difference is related to ln(a/a*): a/a* is the supersaturation ratio where a is the activity of the reacting species in solution and a* the activity at equilibrium solubility.

The traditional form of the rate equation (1) which expresses supersaturation as the difference, (C-C*), between the level of reactive aluminium species in solution and at equilibrium is not defensible from a theoretical point of view. The supersaturation difference depends upon the units in which it is expressed and reflects the strong influence of caustic concentration upon solubility of gibbsite. These problems are eliminated by replacement of supersaturation difference by the supersaturation ratio (C/C*), which is dimensionless and includes the effect of NaOH concentration on solubility. The first step in the development of a new equation, therefore, is to change the form of the driving force for precipitation.

The new rate equation has the form:

Rate = k`.A.mass.(supersaturation)n (2)

Here, rate is the change in total aluminium molality over time. A is the specific surface area (BET) and in the current experiments, is constant over the course of the reaction. Mass is the instantaneous total mass of gibbsite/kg water and supersaturation is defined as ln(a/a*) where a is the molality of the aluminate species in solution and a* is the molality at equilibrium (at the instantaneous NaOH concentration). The effect of using molalities rather than activities in this term is small because the ratio of the two activity coefficients (they are at the same NaOH concentration) is close to unity. The value of a* was calculated using the thermodynamic solubility equation developed by Alcoa R&D (Vernon, 1998). The reaction is assumed to involve monomeric, monovalent aluminium species, i.e. Al(OH)4-.

It is probable that at high caustic and/or aluminate levels, ion pairing of caustic and aluminate may need to be considered. To date however, pairing constants for NaOH and for sodium aluminate have only been estimated up to 0.6M sodium (i.e. 64 g/L TC).

2.0 EXPERIMENTAL METHODS

Data for development of a new rate equation was obtained from the desupersaturation curves for seeded systems. To ensure that only growth took place during these experiments (i.e. no agglomeration, attrition, or secondary nucleation), the supersaturation was kept to moderate levels and the seed carefully prepared to provide particles with very smooth, well-defined surfaces and sharp edges.

A stock batch (30 kg) of seed was prepared from C31 hydrate. The starting material was wet sieved to give the 45-75 mm fraction. This material was still contaminated by very fine particles (<10 mm) clinging to the surfaces. These were removed by two successive treatments with ultrasonics followed by decantation of the fines. Following this, the high index faces (which are sites of secondary nucleation and attrition) were grown out under conditions of moderate TC, temperature and supersaturation. Two cycles of growth were applied. The seed was examined by SEM and Malvern light scattering and also tested in a preliminary precipitation run to ensure that no fines were generated and that the surface area did not change significantly during the experiment. The BET surface area of this material was 0.06 m2/g.

Synthetic Bayer liquors were prepared by digesting AR grade NaOH or KOH with very pure gibbsite (prepared from aluminium flakes) and deionised water in a PFA beaker with a graphite base: this procedure prevented metal ion contamination. Concentrations in this work are based on the North American convention of Total Caustic (TC) being expressed as grams/litre Na2CO3 and Alumina (A) being expressed as grams/litre Al2O3.

All experiments were carried out under isothermal conditions in 1 litre teflon reactors with inverted cone bases. A stirring rate of 200 rpm was normally used. These teflon reactors replaced the stainless steel reactors used initially, as the latter corroded with use and the iron released was believed to interfere with the kinetics of precipitation (cf. Pearson, 1955, Veesler and Boistelle, 1993). The use of teflon reactors ensured that contaminants such as iron could be avoided.

The TC values for these experiments ranged from 100-300 g/L and the temperature from 60-80oC. The same initial supersaturation ratio (1.75) was used for all kinetics experiments and the experiments were run long enough to reduce the relative supersaturation to about 1.1 (i.e. for 3-4 days at 70oC). At high TC values, the viscosity of the liquor is very high and it has been suggested (Kuznetsov and Abanin, 1975) that under these conditions, crystallization may be diffusion controlled. In the present experiments, the rate of precipitation was independent of stirring speed (100-300 rpm) over the entire TC range considered indicating that even at high TC, the rate determining step is the chemical reaction.

The teflon reactor inserts were held in water jackets equipped with heating coils to enable temperature to be controlled to +/-0.2oC with auto-tuned Shimaden temperature controllers. The insulating properties of the teflon inserts meant that heating to the required temperature took longer than for the steel reactors: the initial heating procedure depended upon the temperature selected for the reaction.

Surface areas were determined by a BET method (MicroMetrics Gemini III surface area analyser) using N2 as the adsorbing gas. Scanning electron microscopy was carried out using a JEOL JSM.6400 instrument.

2.1 Data measurement and Treatment

A conductivity probe (Smith and Woods, 1993) was used to follow the changes in Al203 concentration (A) with time and the conductivity readings were converted and stored as voltage values using a DT50 Datataker. Noise in the voltage data was smoothed using an Excel macro (developed in-house) to produce a smooth kinetic curve (Figure1).

Figure 1

Graph of Raw and Smoothed Voltage Data

The smoothed voltage data was calibrated against Al2O3 concentration and TC data obtained from subsamples taken throughout the experiment. Levels of alumina and TC in the liquor were measured by an automatic liquor titration method using the common gluconate-fluoride chemistry. As the initial liquor concentrations were known, the A and TC values were then calculated at each point from the smoothed conductivity measurements. Following this, the instantaneous solubilities were calculated at every point on the curve using the thermodynamic solubility equation and all concentrations converted to molalities. The mass of hydrate at each point was calculated from the initial seed loading and the mass of hydrate deposited. Following this, the supersaturation term (ln a/a*) was calculated for each point and the rate calculated using simple first order numerical differentiation. From these two terms the rate equation was fitted. The specific surface area changed less than 1% during the experiment and the constants n and k were fitted for each experiment. The activation energy was obtained using the Arrhenius equation and used in the calculation of k`o.

2.2 Induction Time Experiments

Induction time experiments involved addition of a known surface area of seed to preheated synthetic liquor held in 250 ml HDPE bottles. After seeding, the bottles were returned to a temperature controlled, rotating water bath. The bottles were sampled and the liquor analysed for Al2O3 concentration at 10-minute intervals over a 4-hour period. The data provided an accurate measure of the time at which precipitation commenced. Samples of seed were examined at the end of the experiment with the scanning electron microscope. These images provided useful information about mechanism of growth.

Two types of seed were used for the induction experiments, one having a smooth surface and the other a rough, poorly defined surface. Most experiments were carried out with the smoothed seed, which was identical to that used in the rate equation experiments. The ‘rough’ seed originated from a coarse plant seed where rubbing in the precipitation circuit had developed an irregular surface. This seed had been slurried and decanted to remove a considerable portion of the fine particles, so its surface area as determined by Malvern light scattering was very close to that of the smoothed seed.

In most cases, synthetic liquor prepared from C31 hydrate was used. A range of A/TC and TC values were investigated although most of the experiments were performed at TC 100, A/TC 0.50 and 65oC, because the different effects showed up most clearly under these conditions. The relative supersaturation for all these experiments was 1.31. Calcium levels in the liquor were measured using ICP.

3.0 RESULTS AND DISCUSSION

3.1 The Induction Time

Under any set of conditions, the length of the induction period was related to the level of calcium in the liquor (Table 1). The induction period decreased markedly with increased seed loading, TC or A/TC (Table 1). Increasing the temperature of the suspension (TC 200 g/L, A/TC 0.6) from 60oC to 80oC reduced the induction period from 21 hours to 10 minutes. These results are in accord with the findings of earlier investigators (Illievski et al., 1989 and references therein). The induction time could also be reduced if ‘rough’ seed at the same solids loading replaced smooth seed. The induction time was slightly longer in potassium-based synthetic liquors than in sodium based liquors at the same concentration (Table 1).

The liquor prepared from C31 hydrate contained approximately 5 ppm calcium (at TC 100 and A/TC 0.5) with the level increasing as the A/TC of the liquor rose. ‘Calcium free’ liquor prepared from pure hydrate actually contained around 1ppm calcium. This came from the NaOH or KOH used to prepare the liquor and was sufficient to give rise to a 20-minute induction time. When the level of calcium was reduced to <1 ppm by step precipitation, the induction period was eliminated under all conditions considered (Table 1).

Table 1

The effect of seed roughness, solids loading, A/TC & TC, cation in liquor and level of calcium in liquor on the length of the induction time. Relative supersaturation (C-C*/C*) = 1.31

 

TC

(g/L)

A/TC

Seed loading

(m2/L)

Temperature

oC

Gibbsite precipitated

g/L

Induction time

(minutes)

[Ca] in liquor

(ppm)

200

0.60

2

65

10.0

90*

6.4

200

0.60

2

65

0.9

>240

6.4

200

0.60

3

65

14.8

20-30

6.4

200

0.60

4

65

18.4

20-30

6.4

100

0.50

3

65

9.0

80

5.1

100

0.50

3

65

10.9

20

1.1

100

0.50

3

65

10.6

0

0.6

100

0.50

3

65

6.2

110#

5.1

300

0.72

3

65

23.7

0

7.8

* Rough seed used: in all other experiments smooth seed used. # Potassium aluminate liquor: otherwise sodium aluminate liquor.

The effect of calcium appears to arise from competition for growth sites between calcium and aluminium species. The strong effect of calcium at the beginning of the reaction suggests that its adsorption is faster and stronger than that of the aluminium species (which are present in excess). If sufficient calcium to form an equivalent monolayer is present, adsorption of the aluminium species and formation of gibbsite is usually hindered. Where the concentration of calcium is high enough to permit formation of several equivalent monolayers, the level of calcium in such a system decreases steadily over a period of hours. The time required to remove the calcium from the liquor depends on the rate of gibbsite precipitation. Each new layer of gibbsite that precipitates must compete with the calcium which is also adsorbing on the surface. The removal of calcium from the liquor can be greatly retarded when rate of precipitation of gibbsite is slow, i.e. at low temperature and/or low A/TC.

Subsamples taken during and after the induction time were examined by scanning electron microscopy. Earlier work had shown that on a rough surface, many small gibbsite nuclei formed during the induction time (Cornell et al., 1996). On the smooth seed used in the current experiments, growth followed a spreading mechanism and numerous nuclei were not observed (Figure 2a). During long (hours rather than minutes) induction periods, the basal plane of the smooth seed is covered with feathery, spreading growth (Figure 2b). Such growth patterns (which have also been reported by Brown, 1972) were not observed if there was no calcium in the system (ie no induction period). It seems probable that these features correspond to normal spreading growth of gibbsite that had been disrupted by blockage of some sites (or growth surfaces) by adsorbed calcium.

A.                                            B.

Figure 2

A. Spreading growth on the basal plane (smooth seed: 3 m2/L, Na-based liquor containing 5 ppm Ca, TC 100, A/TC 0.50, 65oC).
B. Growth pattern which develops on the basal plane during the (>4 hr) induction period (smooth seed: 2 m2/L, Na-based liquor, 6.4 ppm Ca, TC 200, A/TC 0.60, 65oC). Magnification 3000x

In potassium aluminate liquor as in sodium aluminate liquor, growth on the basal plane involves a spreading mechanism (Figure 3a). Even in the presence of calcium species, the growth layer on the basal plane appeared thicker after 4 hours in potassium aluminate liquor than that formed in the equivalent sodium aluminate system although the overall amount of gibbsite deposited was less (Table 1). This result is not unexpected because in potassium aluminate liquors, gibbsite grows preferentially along the c axis as needles. It is the first time, however, that the differences in growth rate appears to have been directly observed. In the presence of calcium, the rate of growth along the c axis in potassium aluminate solution appears to be reduced (Figure 3). In addition, there are holes in the growth layer (Figure 3a) which suggest that nucleation of gibbsite does not proceed readily and that subsequent spreading growth is disrupted to some extent. The growth layer is far smoother and thicker when almost all the calcium has been removed from the liquor (Figure 3b), in part owing to the elimination of the induction period which results in a higher level of precipitation (10.2g gibbsite) during the four hour period.

A.                                        B.

Figure 3

A. Disrupted spreading growth on the basal plane (smooth seed, K-based liquor containing 5 ppm Ca, TC 100, A/TC 0.50, 65oC).
B. Uniform spreading growth on the basal plane (smooth seed, K-based liquor containing <1 ppm Ca, TC 100, A/TC 0.50, 65oC). Magnification 3000x

Over the range of conditions considered in the present work, the induction period could be attributed entirely to the presence of calcium in the liquor. An induction period was not observed even at A/TC 0.5 and 60oC, if there were no calcium in the liquor. Although Veesler and Boistelle (1994) reported that iron in the system caused an induction period, no such effect was observed in these systems, provided that the calcium level was below one ppm.

3.2 The Rate Equation

Several crystal growth models were tested. The best fit of the kinetic data under the conditions considered was obtained using an equation based on ln(a/a*)n with n = 3. Previous investigators (Pearson, 1955; King, 1971; Overby and Scott, 1978; White and Bateman, 1988; Audet and Laroque, 1989; Veesler and Boistelle, 1994) found an exponent of 2, but their equations usually incorporated the supersaturation difference rather than the supersaturation ratio.

The form of the new Alcoa rate equation is:

da/dt = ko.e(-Eact/RT).A.mass.(ln a/a*)3 (3)

where ko = 1.10x1020 (RSE 1%) and Eact = 141kJ/mol.

Over the range TC 100-300 and 60-80oC, ko was constant (Table 2). For the same data fitted to King based equation (1), on the other hand, ko was strongly dependent upon TC (Table 2).

Table 2

Rate constants, ko for the new precipitation equation and for the King based equation

TC

(g/l)

Temperature

oC

kox1020

(new equation)

kox109*

(King based equation)

300

70

1.12

2.71

250

70

1.06

4.61

200

70

1.14

7.39

100

70

1.11

11.75

200

60

1.11

6.55

200

80

1.07

6.13

* Eact for King-based equation (this work) 14.9kcal/mol: BET surface area

With the new relationship, the rate of precipitation can be accurately predicted over a much wider range of conditions than was previously possible. This new equation is not intended to provide any information about the mechanism of precipitation. Few crystallization or dissolution equations are capable of this. Such information has to come from a combination of other techniques including electron microscopy. Our earlier work (Cornell et al., 1996) showed that on the basal plane, growth involves a birth and spread mechanism. In fact two successive steps appear to be combined – spreading of the initial nuclei followed by initiation of a new growth layer on the previous one. As the SEM images (Figure 3) indicate, the second step proceeds more readily in potassium than in sodium aluminate systems.

The activation energy for precipitation of 14 kJ/mol is comparable to that reported by Veesler and Boistelle (1994), but considerably higher than the 40-60 kJ/mol reported for crystal growth reactions in general (Mullin, 1993). This observation suggests that an apparent activation energy that covers at least two parallel reactions, has been measured.

4.0 CONCLUSIONS

The induction period observed during precipitation of gibbsite under Bayer conditions appears to be due solely to the presence of calcium in the liquor. It can be eliminated by removal of the calcium from the liquor before carrying out the precipitation experiment. The induction period can also be greatly reduced if gibbsite growth is fast enough.

When calcium is present in the liquor, an induction period can exist in both potassium and sodium aluminate liquors.

A new rate equation based on the supersaturation ratio instead of the supersaturation difference has been developed using synthetic liquors. This equation avoids the dependence of the precipitation rate constant on TC. It applies over a TC range of at least 100-300 g/L and a temperature range of 60-80oC. An apparent activation energy of 141 kJ/mol for the precipitation of gibbsite has been obtained.

REFERENCES

Audet, D.R. and Laroque, J.E. (1989). Development of a Model for Prediction of Productivity of Alumina Hydrate Precipitation. Light Metals, Ed. P.G. Campbell, pp 21-26.

Brown, N (1972). Kinetics of Crystallization of Aluminium Trihydroxide from Seeded Caustic Aluminate Solutions of Industrial Composition. Paper presented at June 22 meeting of the ICSOBA in Czechoslovakia.

Cornell, R. M., Vernon, C. F. and Pannett, D.S. (1996). Growth of Gibbsite in Bayer Liquors. Alumina Quality Workshop, Darwin.

Haflon, A. and Kaliguine, S. (1976) Alumina Trihydrate Crystallization Part I. Secondary Nucleation and Growth Rate Kinetics. The Canadian Journal of Chemical Engineering, 54, pp160-167

Illievski, D. Zheng, S.G. and White, E.T. (1989). Induction Times for Growth in Seeded Supersaturated Caustic Aluminate Solutions. Chemica 1989. Technology for out Third Century. Broadbeach, Qld. 23-25 August. pp 1013-1019.

King, W. R. (1971) Some Studies in Alumina Trihydrate Precipitation Kinetics. Light Metals pp551

Kirke, E. (1981) Effect of Caustic Concentration on Alumina Trihydrate Growth Rate. R&D Alcoa Internal Report.

Kuznetsov, S. I and Abanin, R. A. (1975) Kinetics of Decomposition of Aluminate Solutions in Increased Concentrations. Tseutnye Metally/Non Ferrous Metals 16, (2) pp 40.

Misra, C. and White, E.T. (1971). Kinetics of Crystallization of Aluminium Trihydroxide from Seeded Caustic Aluminate Solutions. Chemical Engineering Progress Symposium Series, no. 110, vol. 67, pp 53-65.

Mullin, J. W. (1993) Crystallization. 3rd Ed. Butterworth-Heinemann.

Overby, T.L. and Scott, L.E. (1978) Characterization of Bayer Plant Liquors and Seeds Utilizing a Mathematical Model for Precipitation. Light Metals pp163.

Pearson, T. G. (1955) Chemical Background to the Alumina Industry. Royal Institute of Chemistry Monograph No 3 (London).

Smith, P. and Woods, G. (1993) The Measurement of Very Slow Growth Rates During the Induction period in Aluminium Trihydroxide Growth from Bayer Liquors. Light Metals pp 113-117.

Veesler, S. and Boistelle, R. (1993). About Supersaturation and Growth Rates of Hydragillite Al(OH)3 in Alumina Caustic Solutions. Journal of Crystal Growth, (130), pp 411-415.

Vernon, C. F. (1998) Gibbsite Solubility in Bayer Liquors. R&D Alcoa Internal Report.

White, E.T. and Bateman, S.H. (1988). Effect of Caustic Concentration on the Growth Rate of A/(OH)3 Particles.