D.s. Rossitera, D. Ilievskib and G.M. Parkinsona

A.J. Parker Cooperative Research Centre for Hydrometallurgy

a) School of Applied Chemistry, Curtin University of Technology,

GPO Box U1987, Perth WA 6845

b) CSIRO Division of Minerals, PO Box 90, Bentley WA 6102


Nucleation of gibbsite from synthetic Bayer liquor has been studied in the presence and absence of sodium gluconate as a model precipitation poison. Both primary and secondary nucleation mechanisms were examined. The mechanisms of primary nucleation were inferred from induction period measurements obtained using multi-angle laser light scattering (MALLS). The advantage of MALLS is its ability to detect changes in clear, unseeded, supersaturated sodium aluminate solutions earlier than conventional induction period measurement techniques. The secondary nucleation experiments examined the relationship of seed surface area and sodium gluconate dose to the induction period and precipitation rate.

From the unseeded experiments and classical theory, the interfacial tension and critical nucleus size were determined, and shown to be independent of the presence of the poison. These experiments showed that both the homogeneous and heterogenous mechanisms operate within the supersaturation range studied. Addition of poison increased the supersaturation range within which homogenous nucleation was the dominant mechanism. Samples were removed from unseeded experiments and examined using in-lens field emission scanning electron microscopy (IFESEM).

The seeded experiments suggest that the growth rate suppression from the addition of sodium gluconate poison is a surface process, and that it results from a reduction in the available surface area and not from a change in overall surface activity. The mechanism of poisoning was inferred from micrographs collected by scanning electron microscopy (SEM), and by using measurements of the growth rate and induction period for both poisoned and unpoisoned gibbsite precipitations.

The results from both the seeded and unseeded experiments are used to demonstrate that the secondary nucleation of gibbsite is from the surface of the seed and not the adjacent solution.

Key Words:

crystallisation; precipitation; nucleation; induction; poisoning; gibbsite



D.s. Rossiter, D. Ilievski and G.M. Parkinson


Nucleation is said to be secondary, if it requires the presence of seed and supersaturation to proceed, or primary, if it occurs in an unseeded supersaturated solution. Primary nucleation is further sub-divided into homogeneous nucleation, which occurs in truly clean solution, and heterogeneous nucleation, which occurs on foreign particles (eg dust) in the "clean" solution. This paper investigates the effect of sodium gluconate, a known gibbsite precipitation inhibitor, on the mechanisms of primary and secondary nucleation through the study of induction periods and with the aid of electron microscopy.

The induction period is the latent period before the onset of measurable precipitation, and its length is partially a function of the sensitivity of the technique used to measure precipitation. Smith et al (1995) proposed that for gibbsite precipitation, this is actually a period of growth dominated by the seed surface, which is deactivated. Once sufficient new material has been formed, it is actually growth on the new surface that dominates the measured precipitation rate. This is consistent with the observations of Smith and Woods (1993) who were able to measure precipitation during the induction period. SEM studies of gibbsite which reported crystal growth through the development of small nuclei over the seed surface (Misra and White, 1971; Brown, 1972; Gnyra et al., 1974; Smith and Woods, 1993; Cornell et al, 1996) would seem to support the proposed mechanism.

Industrially, the Bayer process is a seeded continuous process; hence, as shown by the above references, the tendency has been for induction period work on gibbsite to focus on seeded precipitation. However, the work of, for example, Mullin and Ang (1976) and Söhnel and Mullin (1978), on nickel ammonium sulphate and calcium carbonate, respectively, has shown that, through the application of classical theory, unseeded or primary nucleation mechanisms can also be investigated through induction period measurement. The observed latent period in this case is due to the time taken for sufficient clustering of molecules to exceed the stable critical nucleus size and survive in solution (below this critical size the cluster is thermodynamically unstable, and without further addition of molecules should "dissolve"). The difficulty faced by these earlier researchers is the ability to measure the true nucleation induction period. In the current work, advantage is taken of MALLS ability to detect changes early in unseeded, nucleating solutions.


For the unseeded experiments, the liquor was prepared from sodium hydroxide pellets (AR grade), deionised water and aluminium wire (AR grade, BDH) degreased with acetone. A portion of the liquor was vacuum filtered twice and then syringe filtered three times with the last filtration being through a 0.1 m m membrane into a previously cleaned scintillation vial. The liquor was ultrasonicated (20 s) to remove bubbles from the solution and then rapidly brought to 60° C in the MALLS instrument, a Dawn-F or DAWN-DSP photometer (Wyatt Technology Corporation, Santa Barbara, CA, USA) fitted with a helium-neon laser (wavelength 632.8 nm). The time elapsed in reheating being minimal as the instrument was at temperature and the liquors retained most of their heat from preparation.

Liquor for the seeded experiments was prepared from high purity gibbsite (C31, produced by Alcoa Chemicals Division, Arkansas), sodium carbonate (AR grade), sodium hydroxide pellets (AR grade) and deionised water. The liquor was then pressure filtered through a 0.45 m m membrane filter and diluted to final mass with deionised water. This was then added to a 4.0 litre (working volume) baffled pitch base batch crystalliser, stirred at 400 rpm with a 4-bladed (45o) paddle stirrer, and the temperature maintained at 60.0oC. Both conductivity and turbidity probes were fitted to the crystalliser, enabling continuous in-situ measurement.

The composition of each liquor was expressed using North American alumina industry terminology where: A is the liquor alumina content expressed as equivalent g L-1 Al2O3; C is the total caustic concentration (both free and combined) expressed in equivalent g L-1 Na2CO3; and S is the total caustic and carbonate concentration, also expressed as g L-1 Na2CO3. For the unseeded experiments, C was kept constant at 200 ± 4 g L-1 and A varied between 100-160 gL-1. The composition of the liquors to be seeded was also kept constant at C=199± 3, S=242± 3 and A=138± 1 g L-1 of liquor.


3.1 Primary Nucleation

Using synthetic Bayer liquor with and without poison (Figure 1), we have measured the induction period for a series of liquors of varying supersaturation. The unpoisoned experiments demonstrate two distinct regions that from classical theory are ascribed to (A) homogeneous and (B) heterogeneous nucleation, the transition occurring at a supersaturation ratio of S = 2.51± 0.07 (equivalent to 0.68 A/C at C = 200). Addition of gluconate to the second series of liquors was found to extend the range of supersaturation ratios where homogeneous nucleation is the dominant mechanism; the transition now occurs at S = 2.21± 0.06 (equivalent to 0.60 A/C at C = 200). It appears that the poison is preventing precipitation on either the heterogeneous surfaces or the new material produced.

Figure 1

The precipitation induction period of gibbsite at varying supersaturation ratios with and without poison: A-homogeneous nucleation, B-heterogeneous nucleation

The slope of the line in the homogeneous region was used to calculate the interfacial tension, g = 45 ± 6 mJ m-2, and from this the critical nucleus size, rc = 1.2 ± 0.1 nm. The poisoned and unpoisoned sets of data collected for region A were equivalent within 95% confidence limits, and hence are shown as one line. From this it is apparent that, within experimental error, the interfacial tension and critical nucleus size are independent of the presence of poison. The slopes of both the lines determined for the heterogenous region were also shown to be parallel within 95% confidence limits. This is most probably due to the poison acting to shut down only a proportion of the heterogenous or new surface, hence the remainder is left to act independently or mirror the unpoisoned experiments.

A correlation published by Mersmann (1990) can be used to predict interfacial tension, and for the liquors used in these experiments predicts g = 46 mJ m-2. Van Straten and de Bruyn (1984) published values of g = 67 ± 20 mJ m-2 and ~25 mJ m-2 for bayerite and psuedoboehmite, respectively. These values were obtained from experiments carried out in 10-3 M NaOH solutions. The interfacial tension determined here is from experiments in ~4M NaOH solution and compares well with the available literature data.

Figure 2

The effect of a series of sodium gluconate concentrations on the induction period, for two fixed supersaturation ratios.

The different effect of the poison on homogeneous and heterogenous nucleation was examined further by using two different supersaturations and several different gluconate concentrations. The results are shown in Figure 2, where again for the supersaturation ratio expected to be in the homogeneous region (S=2.60), within error, there is no effect on the induction period. For the heterogeneous case (S=2.23) the poison is again seen to affect the measured induction period. This is consistent with the mechanism proposed, where the poisoning is a surface process.

Image150b.GIF (2753 bytes)

Figure 3

A - Schematic of the indexing of crystal faces for gibbsite. B - Mechanism for the formation of diamond shaped crystals

Material collected from the unpoisoned experiments was examined using IFESEM (Appendix 1). The first two micrographs are of samples extracted at 15 and 30 minutes into a precipitation experiment, respectively. Both show the hexagonal disc morphology consistent with literature (e.g. Greenwood and Earnshaw, 1984). The second two micrographs are of material collected at the end of a similar experiment (240 minutes). Again, the hexagonal morphology is apparent, and there is also evidence of twining or agglomeration. Of note, however, is the surface roughening, or what would seem to be layer growth. The last micrograph also shows a crystal with diamond shaped morphology, which lacks the surface features of the adjacent hexagonal material. Formation of the diamond plates is consistent with two of the hexagonal faces growing faster than expected, and subsequently being grown out, as shown in Figure 3.

3.2 Secondary Nucleation

Gibbsite precipitation is notoriously slow, with the resulting production of small material (as demonstrated by the micrographs of the unseeded gibbsite – Appendix 1). In batch precipitation, seed surface area is fundamental to determining the rate of desupersaturation and the secondary nucleation rate. The seed surface area is known to affect the measured induction period. This relationship has been quantified by Rossiter et al (1996), Ilievski et al (1989) and Brown (1975). In Rossiter et al (1996), it was shown that the increased induction times resulting from sodium gluconate addition could be satisfactorily explained by a reduction in the available surface area.

The effect of measurement technique is demonstrated in Figure 4, where it seems that the turbidity probe detects the release of the secondary nuclei, and their subsequent growth is measured by the conductivity probe (shown as alumina concentration). Also shown is the difference obtained between choosing the breakpoint and the intersection of the tangents as determining the end of the induction period.

Figure 4

Comparison of two in-situ techniques, conductivity (shown as alumina concentration) and turbidity, used to follow the seeded gibbsite precipitation


Cornell et al (1996) linked the induction period to the presence of calcium in the liquor, reporting that with a reduction in the calcium concentration they eliminated the previously observed induction period. They also linked the induction period to the formation of new material on the surface of the seed, and implied that this is not observed if the calcium concentration is reduced. Similar surface roughening is shown in Appendix 2 in the micrographs taken at 30 minutes, which are of material collected from both poisoned and unpoisoned seeded experiments. Also visible is the formation of diamond shaped material which is then either overgrown or removed. These micrographs are consistent with the observations of Cornell et al. However, the micrographs in Appendix 1 show surface roughening also in the absence of seed or calcium. Thus, it would seem that whilst impurities may encourage surface roughening, under the appropriate conditions they are not essential.

Figure 5

Gluconate dose addition at the end of the first induction period

The complementary nature of the measurement techniques was applied to the seeded gluconate study with the results shown in Figures 5 and 6. Each shows the independent behaviour of the measurement techniques, showing different responses depending upon the timing of the gluconate dose. Turbidity is more sensitive to changes in cross sectional area of suspended solids, while conductivity measures solution concentration. The cross sectional area will change rapidly with the release of nuclei, with negligible change in the solution concentration. The expected subsequent growth onto the large surface area of the released nuclei does change the cross sectional area but not as drastically, whereas it does significantly decrease the solution concentration. Hence, it seems that the turbidity is sensitive to the release of the nuclei and the conductivity to growth. From this, Figure 5 with the poison dose at the end of the "unpoisoned induction period", shows a longer period of limited growth, but the nuclei have still been released, and hence no change in the turbidity profile is observed.

Figure 6

Gluconate dose at 120 minutes, only affecting the turbidity

Figure 6, where the poison dose was in the period associated with growth, shows no measurable effect on the conductivity, but a slight decrease in the rate of change in turbidity. This would seem to fit the observations of Friej et al (1998), where it has been reported that the appearance of surface roughening followed by transformation to the smooth appearance of gibbsite is a cyclic process happening many times, i.e. nuclei are continuously being generated and released. In Figure 6, the late dosing of the poison was sufficient to inhibit some of the release of new material; however, there remained sufficient growth such that the conductivity probe did not detect a significant change in the rate of decrease of solution concentration.

The question arises whether these effects result from some growth sites being completely deactivated whilst the others remain unaffected, or whether all the sites are affected and exhibit reduced activity. We have shown from the unseeded work that the critical nucleus size is independent of poison concentration. The observations in Figures 5 and 6 are again consistent with the gibbsite precipitation being controlled by the amount of surface area available for precipitation. The initial deposition rates were determined for each of the seeded experiments, and, when divided by the initial seed surface area (assuming the change in surface area during the induction period is negligible), gave a constant mass flux. From the poisoned experiments, taking the observed induction period in the unpoisoned study, we are able to predict an equivalent surface area. Taking the mass flux and dividing it by the constant mass flux determined in the unseeded work, we can also predict an equivalent initial seed surface area. For a poisoned system we are thus effectively predicting equivalent seed surface areas from the induction time and from the observed deposition rates after the end of the induction period. Within 95% confidence limits these two predicted areas were found to be equivalent. This again suggests that the poison is acting to shut down or reduce the effective surface area, and that the seed subsequently behaves as if all that has been reduced is the surface area.

If the gluconate reduced the activity of all the active sites by some amount, since the critical nucleus size is constant, then it would take longer to reach the end of the induction period. However, the same amount of surface would be producing the same number of nuclei or amount of new material. At the end of the induction period, the deposition rates would be expected to be the same as measured without poison, assuming that once the critical nucleus size is reached, that the growth rate of the nucleus produced on a poisoned site is the same as for nuclei produced on unpoisoned sites. This is not observed. The observed lower deposition rates are consistent with the postulated mechanism that poisoning occurs by shutting down a portion of the active surface and not by a reduction in the activity of all the surface.


Homogeneous primary nucleation of gibbsite results in particles of classical hexagonal disc morphology, and is independent of gluconate concentration. Heterogenous nucleation in the same system is affected by gluconate concentration. This deactivation of the heterogenous material is consistent with the poison being absorbed onto either the heterogeneous surface or newly produced material. If the poisoning were a result of a solution effect, then this would be observed in the homogeneous case as well.

The measurement technique used to detect the onset of precipitation has again been demonstrated to affect the magnitude of the observed induction period. With this system, and the use of gluconate, it has also been shown that through the simultaneous in-situ application of conductivity and turbidity we can characterise the nucleation behaviour. Selective use of the timing of the poison addition can be used to illustrate features of the nucleation and growth mechanisms.

Electron micrographs of material produced from the seeded experiments show the nuclei forming on different crystal faces, with diamond like morphology. This morphology was observed in both the seeded and unseeded cases. If the new material were produced in the bulk solution then it would be expected to have an hexagonal nature. Consequently, it is concluded that the material is being produced on the seed surface, and the morphology of nuclei is affected by subsequent growth after release into suspension. Hence, it is considered that the secondary nucleation mechanism occurs on the seed surface and not in the adjacent solution, and is based on interaction with the surface.


The authors gratefully acknowledge support from the Australian Government’s Cooperative Research Centres Program. Also acknowledged are the Centre for Microscopy and Microanalysis at the University of Western Australia and C. MacRae, P. Fawell and P. Smith at CSIRO Div. of Minerals.

Acknowledged also is that much of the experimental data used as the basis for this paper have already been published (Rossiter et al, 1996; Rossiter et al, 1998).


Brown, N. (1972) "Crystal growth and nucleation of aluminium trihydroxide from seeded caustic aluminate solutions" J. Crystal Growth, 12, pp 39-45.

Brown, N. (1975) "A quantitative study of new crystal formation in seeded caustic aluminate solutions", J. Crystal Growth, 29, pp 309-315.


Cornell, R., Vernon, C. & Pannett, D. (1996) "Growth of gibbsite in Bayer liquors", 4th Int. Alumina Quality Workshop, (Darwin, NT) proc, pp 97-103.

Friej, S., Lee, M-Y., Reyhani, M.M. & Parkinson, G.M. (1998) "Mechanism and kinetics of growth of the basal faces of gibbsite crystals in Bayer liquors", 5th Int. Alumina Quality Workshop, (Bunbury, WA) proc.

Greenwood, N.N. & Earnshaw, A. (1984) "Chemistry of the Elements", Pergamon Press, Sydney, ch. 7, pp 273-278.

Gynra, B., Jooste, R.F. & Brown, N. (1974) "Polycrystallization of aluminium trihydroxide", J. Crystal Growth, 21, pp 141-145.

Ilievski, D., Zheng, S.G., & White, E.T. (1989) "Induction times for growth in seeded supersaturated caustic aluminate solutions", CHEMECA 89, pp 1012-1019.

Mersmann, A. (1990) "Calculation of interfacial tensions", J. Crystal Growth, 102, pp 841-847.

Misra, C. & White, E.T. (1971) "Kinetics of crystallization of aluminum trihydroxide from seeded caustic aluminate solutions", Chem. Eng. Prog. Symp. Series, 67(110), pp 53-65.

Mullin, J.W. & Ang, H-M. (1976) "Nucleation characteristics of aqueous nickel ammonium sulfate solutions", Faraday Disc. Chem. Soc. Precipitation, 61, pp 141-148.

Rossiter, D.S., Fawell, P.D., Ilievski, D. and Parkinson, G.M. (1998) "Investigation of the unseeded nucleation of gibbsite, Al(OH)3, from synthetic Bayer liquors", J. Crystal Growth, 191, pp.525-536.

Rossiter, D.S., Ilievski, D, Smith, P.G. and Parkinson, G.M. (1996) "The mechanism of sodium gluconate poisoning of gibbsite precipitation", Chem. Eng. Res. Des., 74 A7, pp.828-834.

Smith, P. & Woods, G. (1993) "The measurement of very slow growth rates in the induction period in aluminium trihydroxide growth from Bayer liquors", Light Metals, pp 113-117.

Smith, P., Austin, P. & Ilievski, D. (1995) "Modelling the induction period in gibbsite precipitation", Light Metals, pp 45-49.

Söhnel, O. & Mullin, J.W. (1978) "A method for the determination of precipitation induction periods", J. Crystal Growth, 44, pp 377-382.

Van Straten H. A. & De Bruyn P. L. (1984) "Precipitation from supersaturated aluminate solutions - II. Role of temperature", J. Colloid Interface Sci., 102, pp 260-277.


IFESEM Micrographs - Gibbsite produced from unseeded experiments - material collected at (1) 15 minutes, (2) 30 minutes and (3 & 4) approx. 120 minutes.


SEM Micrograph Series 1 – Gibbsite grown in seeded synthetic liquor – (1) 0 minutes, (2) 30 minutes, (3) 60 minutes and (4) 90 minutes.





















SEM Micrograph Series 2 – Gibbsite grown in gluconate poisoned seeded synthetic liquor – (1) 30 minutes, (2) 60 minutes (2b adjacent and left of 2a) and (3) 90 minutes.