Jonas Addai-Mensah, Igor Ametov, John Ralston and Clive A. Prestidge

Ian Wark Research Institute,

University of South Australia,

Mawson Lakes, S.A. 5095, Australia.


AFM colloid probe and rheological studies were performed under conditions which simulate seeded Bayer Process crystallisation. The role of liquor cation, gibbsite crystal face and hydrodynamic conditions on the time-dependent surface forces and particle interactions were determined. Repulsive forces are observed between gibbsite particle surfaces in fresh liquors; these are considered to be both steric and hydrodynamic in origin. The latter is particle approach rate dependent, and may prevail under processing related conditions. Upon aging, the steric repulsion attenuates and a structured gibbsite interface develops with an associated adhesive force. Adhesive forces were greater at gibbsite non-(001) than (001) crystal faces; this reflects face-specific surface chemistry and different propensities for agglomeration. In seeded, synthetic Bayer liquors the repulsive-to-adhesive transition in particle interaction leads to flocculation and formation of a particle network, with the development of associated non-Newtonian flow, yield stresses and viscoelasticity. The onset of adhesive forces, yield stresses and elastic moduli occurs faster in sodium than potassium based liquors, which concurs with faster agglomeration kinetics in sodium based liquors. These findings together show that gibbsite particle interactions are controlled by the structure of the interfacial layer and influenced by whether sodium or potassium ions are present. Flocculation is a precursor to cementation and may be rate determining in gibbsite agglomeration. A more complete understanding of the gibbsite crystal growth and agglomeration mechanisms has resulted.


gibbsite, interaction forces, agglomeration, colloid probe AFM, rheology.




Jonas Addai-Mensah, Igor Ametov, John Ralston and Clive A. Prestidge


The crystallisation step of the Bayer Process for alumina production from bauxite involves gibbsite seed addition to supersaturated caustic aluminate solutions at elevated temperature and proceeds via nucleation, growth and agglomeration. Crystal growth is notoriously slow and it is generally the agglomeration process that controls the size of the final industrial product. However, the precise mechanism of agglomeration and its dependence on hydrodynamics, aluminate supersaturation, temperature and the liquor cation are poorly understood, hence are rarely used as a basis for improved crystal size control, crystalliser design and performance.

Several studies (eg. Scott, 1963; Misra and White, 1971; Low, 1975; Halfon and Kaliaguine, 1975; Ilievski, 1991) have indicated the reluctance of colloidal size gibbsite crystals to rapidly agglomerate during Bayer precipitation. Speculative mechanisms for this behaviour have emerged, but confirmatory experimental evidence on a microscopic scale is rarely reported. Clearly, interactions between slow-growing gibbsite crystals control the agglomeration process and are themselves critically dependent on the interaction forces. Until recently, however, particle interaction forces have been experimentally inaccessible. Colloid probe atomic force microscopy (AFM), coagulation rate and rheological studies of gibbsite-aqueous and gibbsite-Bayer liquor systems are now starting to address this unusual colloid stability (Addai-Mensah et al., 1998a and b, 1999a and b; Prestidge, 1998; Prestidge et al, 1999a). Repulsive forces, which are not electrostatic in origin have been directly measured between gibbsite surfaces under Bayer-related conditions (Addai-Mensah et al., 1998a) and result in particle interaction behaviour which does not follow the conventional theories of colloid science (Prestidge, 1998). Schematic particle interaction mechanisms have started to emerge which support the experimental observations (Addai-Mensah et al., 1998b, 1999a and b), but these are far from unequivocal and as yet do not address the specific interfacial chemistry and how this is linked to the surface speciation and gibbsite growth pathways. Further investigation in this area is clearly warranted.

In this work we concentrate on AFM colloid probe measurements of the interparticle forces between gibbsite particle surfaces under synthetic Bayer conditions. Specifically, the role of liquor cation (Na+ vs. K+), crystal faces [(001) vs. non-(001)], aging time and hydrodynamics are addressed. Complementary rheological measurements are made on seeded liquors to correlate interparticle forces, particle interaction and agglomeration. The general aim is to further our understanding of gibbsite agglomeration during Bayer crystallisation and to explore the interplay between interfacial chemistry, cation effects and the growth/agglomeration mechanisms.


2.1 Materials

High purity, synthetic Bayer liquors were prepared by dissolving aluminium metal (99.95%, Merck) in either NaOH (99.0%, 1% Na2CO3) or KOH (85% 13.5% H2O, 0.5% Na) solutions. Upon complete Al dissolution, the solutions were filtered through a 0.2 mm caustic resistant membrane filter (Gelman Sciences HT-200 Tuffryn) to give optically-clear solutions. For both AFM and rheological studies the solutions used were at an initial A/C of 0.7 and C= 212 or 318 g dm-3Na2CO3, where A and C refer to Al and caustic concentrations expressed respectively as Al2O3 and Na2CO3 in g dm-3. The supersaturation levels, calculated based on equilibrium solubility data of Misra (1971), were representative of highly pregnant Bayer liquors.

Coarse monocrystalline gibbsite crystals (~150 x 150 x 300 mm), see Figure 1A, were supplied by Pechiney/Queensland Alumina and used as planar surfaces for colloid probe AFM. Gibbsitic spheres (see Figure 1B) were prepared by hydrating 10-30 mm diameter g-alumina spheres (Union Carbide, Linde Div., USA) after the methods of Ginsberg et al. (1962). The hydrated surfaces were shown by XPS valence band analysis to be gibbsite-like in structure. The sphere and plate were respectively glued to the tip of AFM cantilever and a nickel wafer using caustic resistant Shell Epikote 1004 resin. AFM imaging indicated the surfaces of both the plates and spheres were acceptably smooth (root mean square roughness of ~4 nm) for satisfactory interparticle force measurements.

  • Figure 1

    AFM colloid probe substrates: A mono-crystalline gibbsite flat and B hydrated alumina sphere on AFM gold-coated Si3N4 cantilever. Colloidal sized gibbsite particles (Hydral 710) were supplied by Alcoa of Australia and confirmed by XRD to be > 95% gibbsite, with < 5% bayerite and a trace of boehmite. SiO2 (0.04%) and Na2O (0.45%) were the major impurities. The particles were single crystals, prismatic in shape, with an average aspect ratio of 1.2; their size was in the range 0.4 to 2.0 mm, with a mean of 1.6 mm and BET surface area 7.2 m2 g-1. Washed samples were used for rheological and agglomeration studies of seeded Bayer liquors.

    2.2 Methods

    2.2.1 AFM colloid probe

    Colloid probe force versus distance measurements were conducted using a Nanoscope (III) AFM, equipped with an E-head and a fluid cell (Digital Instruments, CA., USA). Microfabricated Si3N4 cantilevers (Digital Instruments, CA., USA) with spring constants in the range of 0.05 to 0.08 Nm-1 were used subsequent to their calibration (Cleaveland et al., 1993). A "gibbsite" sphere was glued to the tip of an AFM cantilever and the gibbsite crystals glued to nickel wafers and mounted on the AFM piezoelectric transducer head. Shell Epikote 1004 resin was used as glue in both cases and found to be caustic resistant over the time scale of the experiments. Initially the cantilever deflection was monitored as a function of the piezoelectric transducer movement and converted into the force versus distance using proven methodology (e.g. Addai-Mensah et al. 1998a). Measurements were made on both the (001) and non-(001) crystal faces at 30 C. We note that equivalent measurements under Bayer crystallisation temperatures are in progress and that the trends observed are identical to those reported here, though kinetic differences due to temperature variation have been observed.

    2.2.2 Rheology

    Freshly prepared synthetic Bayer liquors were seeded with gibbsite particles at solid contents in the range 3 to 30% w/w. Seeded liquors were placed in a well-sealed polymethyl-pentene vessel and immersed in a thermostatically-controlled oil bath at 65 C; this acted as an inert, isothermal crystalliser. No external mixing was employed. At 30 min time intervals, aliquots of suspension were removed from the crystalliser and transferred for flow and yield stress analysis. Inductively coupled plasma atomic emission spectroscopy (ICP-AE) was used to determine changes in supersaturation as a function of time during aging. Laser diffraction (Malvern Mastersizer X) was used to size the filtered solid product as a function of aging.

    Shear stress versus shear rate curves were determined using a Haake CV20 couette type rheometer (shear rate controlled), fitted with a caustic resistant concentric cylinder sensor and a circulating oil bath for temperature control. The shear rate was increased linearly from 0 to 300 s-1 and back to 0 over a 2 min period, with the shear stress measured continuously. Yield stresses were determining using a Haake VT550 rheometer fitted with a caustic resistant Vane, using the method of Nguyen and Boger (1985). Dynamic measurements were made using a Rheometrics SR5000 dynamic stress rheometer, fitted with a caustic resistant parallel plate sensor and the temperature was controlled using a solid-state heat pump (Peltier element). The viscoelastic structure development was determined by applying a non-destructive oscillating deformation to a sample of seeded liquor (at 65 C) at a frequency of 1 Hz. The storage (or elastic) modulus (G’) and loss (or viscous) modulus (G") were and recorded continuously.


    3.1 AFM colloid probe

    3.1.1 Interparticle forces

    Typical AFM traces of the cantilever deflection versus piezoelectric movement acquired at different aging times are shown Figure 2. The data were recorded continuously at a scan rate of 0.49 Hz (approach velocity of 0.59 m s-1), where hydrodynamic effects were found to be insignificant. Initially (Figure 2A) the surfaces are repulsive, with no hysteresis on withdraw (retraction). At faster approach rates a longer ranged repulsion (up to several hundred nanometres) was observed and found to be linearly related to approach rate and solution viscosity. This corresponds to a hydrodynamic force rather than a surface force and is not considered to prevail at the lower particle approach rates experienced during processing; it may however be important in the high shear rate regions of a crystalliser. Upon aging for times greater than 2h (Figure 2B) a characteristic change in compliance is observed on particle approach with significant measurement hysteresis on retraction.


    Figure 2

    AFM cantilever tip deflection as a function of piezoelectric crystal position for gibbsite flat (non-(001) crystal face) and sphere aging in a potassium based supersaturated liquor (A/C = 0.7, C = 212, 30C): (A) within first 1 h and (B) aged for more than 2 h. C marks the change in compliance and X marks the maximum pull-off force required to separate the particles.

    These features are observed more clearly when we consider the corresponding force (normalised by radius, ie. F/R) versus separation curves given in Figure 3. Initially a strong, long-range (up to 40 nm), monotonic repulsion is observed with no jump into contact nor adhesion. During in situ aging a systematic attenuation in the repulsion occurs until at 2 hours there is a clear transformation in the behaviour. The approach curves now show a non-monotonic character indicative of a change in nanomechanical structure as the applied force is increased. A structured layer is considered to be being probed on approach of the particle surfaces and an adhesive force (separation or pull-off force) observed on retraction. With time, both the thickness and density of the structured layer and the magnitude of the adhesive force increase.

    (A)                                                                  (B)

    3.1.2 Crystal face specific interaction forces and the influence of liquor cation

    The repulsive to adhesive behaviour was exhibited at both crystal faces, in both sodium and potassium based liquors and over a wide range of temperatures, caustic concentrations and supersaturations. We now quantitatively explore the influence of cation and crystal face on the magnitude of the interaction forces and kinetics of their transition. For all combinations of crystal face and liquor cation type the maximum repulsive and adhesive forces are plotted against time in Figure 4 A and B, respectively. The magnitude of the initial repulsive force was ~25 % greater for the (001) compared to the non-(001) and ~20 % greater in sodium than potassium based liquors. The subsequent decrease of repulsion was also faster for the non-(001) face.

    (A)                                                                           (B)


    Figure 4

    The influence of liquor cation and crystal face on the maximum forces between gibbsite surfaces in synthetic Bayer liquors: (A) repulsive forces and (B) adhesive forces.

    Interparticle adhesion was stronger and observed earlier in sodium than potassium based liquors; this is in agreement with rheology and discussed further in the following sections. The repulsion to adhesive transition occurs faster at non-(001) faces than the (001) face, but the magnitude of the adhesion is crystal face independent. Recent studies (Addai-Mensah et al., 1999a) have shown that the non-(001) faces of colloidal gibbsite nuclei show a propensity towards initial agglomeration in comparison with the (001) faces and is in agreement with direct force measurements.

    3.2 Rheological and Agglomeration Behaviour

    Flow curves and yield stresses/viscoelasticity were measured at 20% and 25% (w/w) gibbsite seeding, respectively. Under batch conditions, at these relatively high levels of seed surface area, no significant particle size enlargement was determined by laser diffraction. That is, supersaturation is consumed prior to significant growth or agglomeration. The observed rheological changes therefore reflect particle network formation (ie. flocculation), which occurs as a precursor to cementation. At 3% (w/w) seed loading, size enlargement was significant and the higher supersaturation-to-surface area ratio enabled crystal growth, cementation and agglomeration to prevail; these findings are included below for comparison.

    3.2.1 Flow Curves

    Shear stress versus shear rate curves for supersaturated, synthetic Bayer liquors seeded with colloidal gibbsite (see Figure 5) are extensively time-dependent. Increased shear thinning and pronounced thixotropy (flow curve hysteresis) is evident upon aging. It should be noted that the flow behaviour of equivalent unseeded liquors is Newtonian and time-independent, furthermore, no such time-dependent rheology has been observed for under-supersaturated seeded liquors. Given that growth and agglomeration are insignificant, it is the change in the gibbsite particle interaction potential that results in the formation of a gibbsite particle network with the observed thixotropy.

    Figure 5

    Flow curves for a sodium based liquor (A/C = 0.7, C = 318 g dm-3, at 65 C) seeded with 20% (w/w) colloidal gibbsite.

    Figure 6

    Thixotropic structure in aging, synthetic liquors (A/C = 0.7, C = 318 g dm-3, at 65 C) seeded with 20 % (w/w) colloidal gibbsite.

    The area within a flow curve hysteresis loop equates to the energy per unit volume applied to the suspension per unit time to breakdown microstructure. Figure 6 shows that the rate and extent of thixotropic structure formation is greater in sodium than the potassium based liquors and may reflect cation effects in controlling gibbsite particle interactions. This is not, however, unequivocal since the measured thixotropic structure is influenced by both particle interactions and liquor viscosities. Yield stresses and elastic moduli, as reported below, are more directly related to particle interaction forces. With this in mind, studies presented elsewhere (Prestidge et al., 1999a), have shown that the rate of thixotropic structure formation is strongly dependent on the initial supersaturation but only weakly dependent on temperature. Therefore, for any liquor strength there may an optimum temperature for particle network formation. Further studies in this area are clearly warranted.

    3.2.2 Yield stresses

    Yield stresses, measured by the Vane method, have been determined over a wide range of liquor strengths, seed loading levels and liquor cation types. For supersaturated liquors with seed loading greater than ~15% (w/w) a significant increase in the yield stress is observed upon aging, as demonstrated in Figure 7. The increasing yield stress is directly related to the increasing adhesive particle interactions and confirm the formation of a particle network structure through flocculation. In sodium based liquors, the faster rate of yield stress development is in agreement with the less pronounced repulsive forces and greater adhesion as determined by AFM. Particle interaction energies may be estimated from yield stresses and then quantitatively compared with AFM colloid probe measurements; this is however beyond the scope of the present paper.

    At seeding levels less than 5% (w/w) the liquors are effectively Newtonian and no yield stresses or viscoelastic structure are determined; agglomeration is however prevalent as shown in Figure 8. At these lower seeding levels the supersaturation is sufficient to enable the flocs to cement into agglomerates, which have been confirmed by scanning electron microscopy. Agglomeration occurs faster and more extensively in sodium compared with potassium based liquors, which is in agreement with rheology and confirms that flocculation and cementation are directly linked.

    In these experiments the aging was undertaken in the absence of external shear, where the flocculation and cementation processes are therefore unperturbed. Under the hydrodynamic conditions of a plant crystalliser, however, the flocculation step is significantly perturbed and it is envisaged that the shear forces may cause floc rupture. Shear dependent agglomeration is in fact well established. In these cases the rate of flocculation may be the rate-determining step of agglomeration.


                     Figure 7

    Yield stresses in synthetic Bayer liquors (A/C = 0.7, C = 212, 65 C) seeded with 25% (w/w) colloidal gibbsite.


    Figure 8              

    Particle size in synthetic Bayer liquors (A/C = 0.7, C = 212, 65 C) seeded with 3% (w/w) colloidal gibbsite.


    3.2.3 Viscoelastic structure

    In fresh, synthetic Bayer liquors seeded with colloidal gibbsite the particle network structure, as determined by dynamic rheological methods, is initially weak and dominated by viscous (steric) forces, ie. G" >> G’. This concurs with colloid probe AFM, which shows that repulsive steric forces initially dominate. Upon aging there is a dramatic increase in the elastic structure, as exemplified in Figure 9. With time, as the adhesive forces develops, G’ becomes greater than G" and the particle network strengthens. The precise form of the viscoelastic structure development is directly related to the interfacial chemistry as particles flocculate.

    The rate of elastic structure development is considerably faster in sodium compared with potassium based liquors, again confirming the influential role of the cation in controlling the interfacial chemistry and flocculation process.

    Figure 9

    Viscoelastic structure development in synthetic Bayer liquors (A/C = 0.7, C = 212, 65 C ) seeded with 25% (w/w) colloidal gibbsite

    3.3 Interfacial Structure and Particle Interaction

    On immersion of gibbsite seeds into fresh synthetic Bayer liquors the interfacial layer is initially diffuse in nature and thought to be composed of polymeric aluminate species which lead to repulsive interparticle surface forces. Particle flocculation is weak, agglomeration is unlikely to prevail at this time and the seeds may be considered "dead". Upon aging the thickness of this diffuse layer decreases and the repulsive interparticle force attenuates. A clear transition in the interfacial properties then occurs with the formation of a structured interface which is adhesive. On further aging the thickness of the structured interface and its adhesive nature increases. The seeds are now considered "active", flocculation proceeds and the likelihood of agglomeration improves. The structured interface is believed to be composed of a polycondensed, Al(III)-containing structures formed by the densification of diffuse, Al-containing steric layers which formerly led to repulsion (Addai-Mensah et al., 1998a). The adhesion is presumed to be due to the intergrowth of the densifying Al-containing interfacial layers. The probability that a floc cements into an agglomerate is controlled by the solution and hydrodynamic conditions.

    Though the precise chemistry of the interfacial structuring has not been fully exposed it is clear that both surface and solution properties play important roles. The presence of sodium, in comparison with potassium ions, enhances the rate of particle interaction and flocculation; this may be linked to their stronger aluminate ion pairing in solution and greater surface adsorption affinity (Prestidge et al., 1999b). The stronger activity of the non-(001) than the (001) face towards particle interaction is also informative and may be related to the chemical difference in their surface hydroxyl groups (Hiemstra et al., 1989). The challenge for the future is to fully deconvolute the interfacial chemistry which leads to the complex particle interaction behaviour and use this knowledge for improved agglomeration control during Bayer crystallisation.


    Gibbsite particle interaction forces under Bayer processing related conditions are directly measurable by colloid probe AFM and indirectly determinable by rheological methods. Repulsive surface forces dominate in fresh liquors and are considered to be steric in origin. The repulsion attenuates upon aging, a structured gibbsite interface develops and an adhesive force is apparent. The repulsive to adhesive transition results in the formation of a shear-sensitive particle network in seeded, supersaturated, synthetic Bayer liquors. The flocculation process responsible for particle network formation is a precursor to the cementation step and may be the rate determining step in agglomeration. The kinetics and extent of network formation is reflected in the development of non-Newtonian flow, yield stresses and viscoelasticity with time. Adhesive forces, yield stresses and elastic structure all develop faster in sodium than potassium based liquors and concur with the agglomeration kinetics. Adhesive forces are greater at gibbsite non-(001) than (001) faces and has as bearing on agglomerate structure. Knowledge concerning gibbsite particle interactions, which are critically controlled by the time-dependent interfacial chemistry and structure, enhance our understanding of the agglomeration mechanism and may ultimately lead to improved process control.


    Financial support for the project is acknowledged from the Australian Research Council’s SPIRT grants program and the alumina industry (Alcoa of Australia, Billiton, Comalco, Nabalco, Queensland Alumina and Worsley Alumina) through the Australian Minerals Industries Research Association (project P380B). Discussions with Dr. Andrea Gerson and Dr. Rob Hayes are warmly acknowledged.


    Addai-Mensah, J., Dawe, J., Hayes, R., Prestidge, C.A., and Ralston, J. (1998a). The Unusual Colloid Stability of Gibbsite at High pH, J. Colloid and Interface Science, 203, 115-121.

    Addai-Mensah, J., Gerson, A., Prestidge C.A., Ametov, I., and Ralston, J. (1998b). Interactions Between Gibbsite Crystals in Supersaturated Caustic Aluminate Solutions. Light Metals, 159-166.

    Addai-Mensah, J., Prestidge, C.A., and Ralston, J. (1999a). Gibbsite Particle Interactions and Agglomeration During Bayer Crystallisation. Minerals Eng., Accepted for publication.

    Addai-Mensah, J. and Ralston, J. (1999b). The Influence of Interfacial Structuring on Gibbsite Interactions in Synthetic Bayer Liquors, J. Colloid and Interface Science, Accepted for publication.

    Cleaveland, J.P., Manne, S., Bocek, D., and Hansma, P.K. (1993). A Non-Destructive Method for Determining the Spring Constant of Cantilevers for Scanning Force Microscope. Reviews in Science Instr. 64(2), 403.

    Ginsberg, V.H, Huttig, W., and Stiehl, H. (1962). Uber die Bildung von Kristallinen Al(OH)3 und die Umwandlung von Bayerit in Hydragillit. Zietscrift fur Anorgassche ubd Allemeine Chemie, 318, 239.

    Halfon, A., and Kaliaguine, S. (1976). Canadian J. of Chem. Eng. 54, 160-167.

    Hiemstra, J., De Witt, J.C.M., and Riemsdijk, W.H. (1989). Multisite Proton Adsorption Modelling at Solid/Solution Interface of Hydr(oxides): a new Approach. J. Colloid and Interface Science, 133, 105.

    Ilievski, D. (1991). Modelling Al(OH)3 Agglomeration During Batch and Continuous Precipitation in Supersaturated Caustic Aluminate Solutions. Ph.D Dissertation, University of Queensland, Australia.

    Low, G.C. (1975). Agglomeration Effects in Aluminium Trihydroxide Precipitation. Ph.D Dissertation, University of Queensland, Australia.

    Misra, C. (1971). The Precipitation of Bayer Aluminium Trihydroxide. Ph.D Dissertation, University of Queensland, Australia.

    Misra, C. and White, E. T. (1970). The Precipitation of Bayer Aluminium Trihydroxide. J. Crystal Growth, 8, 172-178.

    Nguyen, Q.D., and Boger, D.V. (1983). Yield Stress Measurement for Concentrated Suspensions. J. Rheology, 22, 335-349.

    Prestidge, C.A. (1998). Rheological Investigations of DLVO and non-DLVO Interaction in Concentrated Mineral Suspensions. Mineral Processing and Extractive Metallurgy Reviews, pp. 1-19.

    Prestidge, C.A. Ametov, I. and Addai-Mensah, J. (1999a). Rheological Investigations of Gibbsite Particles in Synthetic Bayer Liquors. Colloids and Surfaces A, in press.

    Prestidge, C.A., Rowlands, W.N., and Fornasiero, D., (1999b). Surface Electrical Properties of Gibbsite Particles in Highly Concentrated Salt Solutions. J. Colloid Interface Sci., submitted for publication.

    Scott, J. (1963). Effect of Seed and Temperature on the Particle Size of Bayer Hydrate. Extractive Metallurgy of Aluminium, 1, Alumina, Interscience, New York, pp. 203.