Michael Thornber1 and David Binet2

1. CSIRO Minerals, PO Box 90, Bentley, Western Australia 6982

2. Alcoa of Australia Ltd., PO Box 161 Kwinana, Western Australia 6167


The adsorption data for sodium, caustic, total alkalinity and aluminate were measured for various Bayer residues by a precise washing procedure. The results were expressed as functions of the of the associated liquor TA and fitted by both logarithmic and Langmuir functions within the experimental error. For a residue rich in goethite the Na adsorption was consistent with known adsorption on goethite at lower pH.

The changes in adsorption on residue solids with added DSP gave data on the contribution from the DSP and similarly adsorption data on the effect of lime additions was obtained. The contributions to the adsorption from the various components of the residue solids were identified as follows:

The results support the hypothesis that mineral surfaces, in a concentrated Bayer liquor have Na+ adsorbed at an inner layer and the negative counter ions such as OH-, CO32-, Al(OH)4- and organic and inorganic anions that are regarded as contaminants to the Bayer process are adsorbed as an outer or double layer. This could also apply to gibbsite surfaces involved in crystallisation and dissolution processes.


Bayer residues, alkalinity, adsorption, sodium, dsp, liquor TA, caustic, lime, goethite, adsorbed soda



Michael Thornber and David Binet


Caustic soda consumption is an important factor contributing to the operating cost of Bayer plants. Most of the plant soda loss occurs via residue disposal, either as chemically combined soda in desilication product (DSP), or as soluble soda in the associated liquor. The soluble soda loss depends on the amount of liquid lost with the residue and its soda concentration which in turn is governed by performance of the residue washing circuits.

The washing process involves chemical reactions between the surfaces of a complex mixture of mainly Fe, Si, Al, Ti, Ca, oxide and hydroxide mineral solids and the equally complex Bayer liquors that vary in temperature and composition from near boiling at the highest concentrations down to 40 oC or below at the low concentrations required for residue disposal. Any soda exchange that occurs between solids and liquor by adsorption/desorption, dissolution/precipitation, exchange reactions etc, should be taken into account if accurate mass balance calculations are required. Such processes will be referred to collectively in a general way as removal of "adsorbed" soda.

The objective of the work described here was to develop a laboratory test procedure and analytical methods that could be used to quantify this adsorbed soda as a function of liquor TA and residue mineralogy.


2.1 Solution Assay

Total alkalinity (TA), total caustic (TC) and alumina (A) were measured by an adaptation of the Watts and Utley, 1953, 1956, methods using a Mettler DL70 automatic titrator. Analysis was on a weight basis to simplify calculations and improve overall precision. The measurement at the second end point was very precise; it corresponds to total caustic + 1/2 carbonate and was recorded as TAC. A sample of known TA, TC and A was run as a control in each batch of samples.

Na in solution was measured by atomic absorption spectroscopy. Solutions were accurately diluted by weight in 0.2M HCl to give approximately 20 ppm Na2O and four completely independent assays were carried out on each solution to improve precision. In some tests, Al, Ca and Si were assayed by ICP-EES techniques.

2.2 Residue Preparation and Sampling

Where a bauxite was used as starting material it was rod milled in hot plant liquor, held at 95 C for 16 hours, digested in a Parr 20 L stainless steel reactor with the remaining liquor charge for 20 minutes at 143 C and then cooled rapidly to 100 C before discharge through a 150 m m sieve to remove sand. After addition of freshly prepared flocculant the fine mud slurry was allowed to settle at 90 C for up to one hour. Most of the clear liquor was then pumped off leaving approximately 1 litre of slurry containing about 500 g of fine mud solids to be transferred to a one litre vessel and placed in a water bath at 55 C to cool to the washing test temperature.

Other mud samples were taken directly from a refinery first washer underflow They were settled at 90 oC and treated as above to give 1 litre samples at 55 C.

2.3 Adsorption Measurement Method

This procedure was developed to make a precise balance for the mass of liquor and its dissolved components over each of four successive batch wash stages. An essential feature is that mud solids remain in the same 70 mL centrifuge tube throughout the whole procedure, thereby eliminating any solids loss. The centrifuge tubes are equipped with a leak proof lid and mixing rod so that slurries can be completely dispersed by shaking. A single wash cycle involves weighing the slurry contained in the tube, centrifuging to spin down solids, removal of clear supernatant, addition of water to dilute the slurry, dispersion and equilibration of the diluted slurry for 30 minutes, centrifuging and removal of the dilute supernatant. The centrifuge and shaking water bath were both thermostatically controlled at 55 C. Each removal of clarified liquor or addition of water was monitored by weighing. The quantity of solids in the tube is determined by drying the slurry from the final stage wash and allowing for dissolved solids present in the final dilute liquor.

For each slurry tested, eight independent washes were carried out in duplicate pairs at four different mud weights. Two blanks, consisting of liquor alone, were carried through the same procedure. The dilutions used were calculated so that the TA of the final wash liquor for all tubes was less than 2gL-1.

All liquors were assayed for TA, TC, A and Na and once the weight of residue solids in each tube had been determined the amounts washed off the solids at each stage could be calculated by mass balance. This procedure gives only relative changes in adsorption which occur as the liquor is progressively diluted. By arbitrarily assuming that there is zero adsorption at TA=2gL-1 it is possible to combine the data from all 8 tubes, even if the final TAs are different, and to compare adsorption data from different runs. This is a convenient reference point since it coincides with a practical definition of fixed soda The eight tubes each with 4 washes give a total of 32 data points for each data set.

It should be noted that all adsorption calculations were made in terms of the measured mass based concentrations. For display purposes the TA concentrations have been converted to volume based units (gL-1) using a density correlation.

2.4 Characterisation of the Solids

Care was taken to sample the solids by riffling or quartering techniques so that the samples were as representative as possible. All of the data from the measurements shown below were combined to calculate the mineral content of the solids.

Bauxites and washed residue solids were assayed for major and minor elements by X-ray fluorescence techniques and powder X-ray diffraction patterns were used to give a semi quantitative measurement of the mineral proportions based on the peak intensities of the diffraction lines that were not suffering overlap interferences. For the residue solids semi quantitative estimates of gibbsite and goethite content was made using DTA/TG data from a Rigaku Thermoflex instrument. Measurements of the surface area and particle size distribution of the residues were made by multi point BET and a Malvern SB instrument respectively.

DSP content of the residues was determined by selective leaching with dilute sulphuric acid. The solutions were assayed for Si and Al by ICP-AES and Na Mg Ca Fe and Cr by atomic absorption. Calculations were based on the noselite formula, Na6(AlSiO4)6.Na2SO4.


Results calculated from the Na data obtained for the –150m m residue from a goethite rich Darling Range bauxite are shown in Figure 1. The errors shown were due mainly to the adsorption being calculated from the difference between successive solution assays. Precision for the sodium assay was 0.3 %, which together with weighing errors leads to overall errors of about 25 % for the first wash, and 10, 4, 2 and 1% for successive washes.

In Figure 1a the data has been fitted by a logarithmic curve of the form:

Adsorbed Na2O = P1(logTA-log2), (2)

using a weighting scheme proportional to the inverse of the error squared to emphasise the more precise data at low TA. P1 is the slope of the log plot and is numerically equal to the adsorbed soda at 2 gL-1 TA.

Figure 1

Sodium adsorption data for a residue from a goethite rich bauxite, (a) logarithmic plot (weighting proportional to the inverse of the error squared) and (b) Langmuir plot (unit weighting)

Figure 1b shows the same data fitted (using unit weighting) by a Langmuir equation that forces the curve through the reference point for zero adsorption at 2 gL-1:

Adsorbed Na2O = P1*P2(TA/(1+P2*TA)-2/(1+2*P2)) (3)

In this equation P2 is the Langmuir constant that gives an indication of the surface affinity and the constant P1 is the saturation adsorption limit when TA is very large (Veith and Sposito, 1977). The value of P1 found for this residue, 0.812 0.013 g Na2O per 100 g of mud solid, and its measured surface area (25.56 m2g-1 ) lead to an estimate of 10.2 micromoles of Na+ per m2, in excess of the measured surface sites for a synthetic goethite, 4.25 (Balistrieri and Murray, 1981) and 3.82 micromoles per m2 (Davis and Kent, 1990). This calculates to 6.2 Na+ ions per nm2 and could indicate that the surface is near to saturated with Na+ in the Bayer liquor or that the 1% DSP in the solids is absorbing the excess Na+. The adsorption at TA = 0 can be calculated to be, -2P1*P2/(1+2P2) = -0.1055 g of Na2O per 100 g of mud solid. The asymptote for limiting adsorption where TA is very large relative to zero adsorption at TA = 2 is P1-2P1*P2/(1+2P2) and calculates to be 0.7064, as indicated by the arrow in Figure 1b. The equation can be extrapolated to calculate adsorptions of 0.46, 0.11 and 0.072 micromoles Na+ per m2 at pH 12, 11 and 10 respectively and these fall within the ranges 0.37 to 0.61 at pH 12, 0.11 and 0.35 at pH 11 and 0.02 and 0.21 at pH 10 measured for goethite using tracer techniques (Rundberg et al, 1994).

Figure 2, a b and c give adsorption curves calculated from the TA, TC and TAC data for the same goethite rich bauxite residue. In each case the adsorption is expressed in terms of Na2O and fitted with logarithmic curves so that they can be compared to the Na data. As expected the slope for the TA data, 0.360, (Figure 2a) is greatest and the same as that for the sodium data which means that the total alkalinity washed off which includes, hydroxide, aluminate and carbonate has sodium as the counter ion washed off to maintain charge balance. The slope for the TC data, 0.314, (Figure 2b) is significantly lower as it does not include the carbonate washed off. The slope from the TAC data, 0.328 (Figure 2c) is between the other two because it includes half the carbonate washed off and its value lies in the smaller errors at higher TA.

Figure 2

Adsorption data for the residue from a goethite rich bauxite fitted to logarithmic plots using the shown weighting schemes (a) TA, (b) TC and (c) TAC; (d) Aluminium adsorption data only

The Aluminium data for the same goethite rich bauxite residue, Figure 2d, indicates that alumina precipitation dominates at high TA values but is washed into solution at low TAs. The precipitation was confirmed by data from the blank tubes with liquor only which also showed alumina loss at high TA values. The data for TA<25 gL-1 was reasonably well fitted by the logarithmic model and corresponding values for P1 are given in Table 1.

3.1 Effect of DSP

To assess the contribution of DSP to adsorbed soda, washing tests were carried out on the residues prepared from the goethite rich bauxite after spiking with 60 g of kaolin per kg, and from a high quartz bauxite after spiking with 0, 30 and 60 g per kg. A washing test was also conducted on a synthetic DSP prepared from a mixture of Bayer liquor and sodium silicate. The P1 values obtained for the log plots are given in Table 1.

The increase in DSP content of the muds from the spiked bauxites, and the corresponding increases in P1 values can be used to estimate values for the DSP component. For the Na data the estimates are 0.83, 0.97, and 0.60; the average value of 0.80 is consistent with the value of 0.88 measured for DSP alone. This is more than twice that of the goethite rich bauxite residue. The high adsorption associated with the DSP must be due to exchange reactions with the zeolite type of cage sites that are part of the DSP sodalite and cancrinite structures (Leiteizen et al, 1975) and close examination of the data indicates that this is mainly washed off at TA values less than 30. The saturation calculation based on the Langmuir fit of the Na data for the synthetic DSP gives a surface Na+ density of 140 atoms per nm2 and this is consistent with an exchange reaction.

Table 1

Slopes for the adsorption data fitted to log curves for the residues from a goethite rich bauxite and a quartz rich bauxite both with increasing amounts of DSP due to kaolin addition. Adsorption fits for a synthetic DSP are included. (Unit weighting was used.)

Table 2

Comparison of surface area and particle size measurements for residues where kaolin or lime was added to the bauxite


The additional DSP did not significantly alter the surface area of the residue but it caused a significant increase of particles in the size range 0.3 to 6 m m as shown in Table 2.

3.2 Effect of Lime Additions

Two sets of tests were carried out where lime was added to selected bauxites to determine its effect on the adsorption properties of the residues. One involved the goethite rich bauxite where 0.5 and then 1.0 % CaO slurried in liquor was added to the bauxite just prior to digestion. The other involved adding 1% CaO before predesilication in one case and after predesilication and prior to digestion of a Jamaican bauxite that gave fine hematite rich residue of large surface area. The changes to surface area particle size distribution, shown in Table 2, indicate that the surface area has decreased with lime additions for both bauxite residues and yet the adsorption has increased substantially as indicated by the slopes to the logarithmic adsorption curves shown in Table 3. The particle size distribution appears to show a general decrease where the lime was added just prior to digestion especially for the goethite rich bauxite (Table 2) but there has been an increase in particle size where the lime was added at the start of predesilication.

Comparing these washed residues by SEM examination indicated that where there was lime addition the presence of Ca was ubiquitous and associated with all areas examined and in far greater amounts than the ~3.3 wt % CaO (measured by XRF). this was not so for the residues with out lime additions or if the residues were well ground before examination. These results suggest that a Ca oxide or hydroxide has coated the surfaces of the solids and this surface has an increased capacity to adsorb Na+, OH-, CO32- and Al(OH)4- compared to the surfaces without the coating. Langmuir fit of the data also showed the same increased capacity but that the affinity constant remained the same indicating that the alkalinity could be washed off just as easily. In fact XRF analysis of the washed residue solids showed less Na2O remaining with those solids where lime had been added compared to those where there were no lime additions.

Table 3

Slopes of logarithmic fits to adsorption data for bauxites reacted with lime

The decrease in surface area is probably due to a calcium oxide or hydroxide coating filling micropores and preventing formation smaller particles during digestion. About half of the Ca that remained with the washed solids was able to be leached with cold weak acid during the DSP assay leach and there was a decrease in the leachable Al as the Ca additions increased. If it is assumed that the coating is acid leached so that it was unlikely that the coating was a leachable calcium aluminate and more likely to be a form of calcium hydroxide, oxide or carbonate. In all cases the amount of DSP in the residue solids decreased when lime was added, more so when the lime was added before predesilication and this same process could have caused the apparent increase in particle size in this case.

3.3 General Correlations

Adsorption data was produced for 17 differing bauxite residues and correlations were found between the adsorption slopes for logarithmic fits of the data and various characteristics of the residue solids. The main factors that correlated with the Na adsorption slopes also applied to the other adsorption data.

Most important was the amount of DSP and as already indicated this was due to the ability of DSP to absorb Na and caustic into the cage sites in the minerals. Next was amount of Ca in the solid in excess of the Mg indicating it was not tied up in dolomite related carbonate minerals and thus available to activate mineral surfaces for caustic adsorption as described above. The amount of goethite, boehmite and hematite minerals, both of which can have aluminium, manganese and titanium as well as iron incorporated was next in importance and thus there was a strong correlation with the combination of the of the Fe, Al, Mn and Ti content of the solids. Because of this and probably because of some absorbed water in the DSP and Ca coating there was a strong correlation between the loss on ignition (LOI) of the solids and the adsorption. Quartz gave a strong negative correlation indicating that it was a minor contributor to adsorption.

The correlation of adsorption with surface area was not as good as expected. This was probably due to two effects in the data set used:

    1. DSP absorbed into the cage sites within the solids as well as on the surfaces.
    2. The effect of increasing lime caused a decrease in surface area associated with and increase in adsorption. Thus a negative correlation with surface area.

3.4 Solid/Liquor Interface

The adsorptions were measured using concentrated Bayer liquors and diluting down to TA values that were as low as 0.5 which is near to a pH of 12. The high concentrated liquor can be regarded as sodium hydroxide, aluminate, carbonate, etc/ water mixture where the water activity is low enough that there is insufficient H2O molecules to solvate all the ions that are in solution and thus the solid liquor interface is competing with the ions in solution for H2O to be adsorbed. As the solids are washed to equilibrate with liquors of lower concentration the water activity increases and the interface changes to be more near to that known for aqueous solutions. It was shown that the data presented here could be extrapolated to pH 12 and below and be consistent with the results of Rundberg et al, 1994, for the adsorption of Na+ onto goethite. Goethite and the other residue minerals have their interfaces with aqueous solutions varying in surface charge that becomes more negative as the pH increases (Bowden, et al, 1980). Rundberg et al, 1994, argue that as the pH increases more and more Na+ is bound as an inner layer to the surface but there is no theory to carry the understanding into less aqueous Bayer liquors. The adsorption data presented here indicates that as the pH or the Bayer concentration increases then the Na+ continues to be adsorbed at this inner layer displacing water so that charge balance is maintained by negative counter ions such as OH-, CO32-, AlO4- and organic and inorganic anions that are regarded as contaminants to the Bayer process being adsorbed as an outer or double layer. The data also indicates that a cation such as Ca2+ that is more strongly charged than Na+, will displace it from the interface and chemically bind with the charged surface as an oxide that is negatively charged and favourable to adsorbing Na+ (Bowden et al, 1980).

An attempt was made to use these techniques to measure the Na+ adsorption on gibbsite but the interface was too dynamic to obtain useful data. However it is still likely that the gibbsite/liquor interface has Na+ adsorbed as an inner layer during crystallisation and dissolution and that Ca2+ is capable of forming chemical bonds that interfere with these processes (Cornell et al, 1996).


An empirical method has been developed to quantify the amounts of Na2O, total alkalinity, total caustic and alumina that are removed from Bayer residue solids as they are progressively washed from 250 to 2 gL-1 TA. Within the experimental errors, the release of sodium matched that of the alkaline anions. This release has been described in terms of removal of "adsorbed" species but this is unlikely to be the only mechanism involved.

The adsorption decreased linearly with the logarithm of liquor concentration and the slope of this line has been used to compare adsorption characteristics of different residues. The data was fitted equally well by Langmuir type equations. For a goethite rich residue the Langmuir equation could be extrapolated to fit with known data for the adsorption of Na onto goethite at pH 12 to 10 and indicated that the surfaces were near to saturation in concentrated liquors.

Correlation studies using adsorption data from 17 differing Bayer residues showed that Na+, OH-, CO32- and Al(OH)4- adsorption was influenced, in decreasing order, by DSP, some Ca species, and minerals such as goethite, hematite and boehmite that are known to have charged surfaces. Quartz was essentially inert. Adsorption correlated well with LOI but, surprisingly, only poorly with surface area.

The results support the hypothesis that mineral surfaces, in a concentrated Bayer liquor where water activity is low, have Na+ adsorbed at an inner layer and negative counter ions such as OH-, CO32- and Al(OH)4- are adsorbed as an outer or double layer. This could also apply to gibbsite surfaces involved in crystallisation and dissolution processes.


Significant contributions to this study came from the careful work of Caroline Hughes, Teresa Bocking, Geraldine Woods, Alan Fletcher and Phil Shawcross.


Balistrieri, L. S. and Murray, J. W. (1981). The surface chemistry of goethite (a FeOOH) in major ion seawater. Am. J. Science, vol 281, pp 788-806.

Bowden, J. W., Posner, A. M. and Quirk, J. P. (1980). Adsorption and charge phenomena in variable charge soils. from "Soils with Variable Charge". Theng, B. K. G. ed., N. Z. Soil Science Spec. Publication. Ch. 9, pp 147-165.

Cornell, R. M., Vernon, C. F. and Pannett, D. S. (1996). Growth of gibbsite in Bayer liquors. Fourth International Alumina Quality Workshop, Darwin. pp 98-103.

Leiteizen, M. G., Pashkevich, L. A.., Firfarova, I. B. and tsekhovolskaya, D. I. (1975). Characteristics of sodium hydroxoaluminosilicates forming during bauxite digestion. in Mineralogical and Technological Evaluation of Bauxites. FKI, Research Inst Non-ferrous Metals, Budapest, pp 167-179.

Rundberg, R. S., Albinsson, Y. and Vannerberg, K. (1994). Sodium adsorption onto goethite as a function of pH and ionic strength. Radiochimica Acta, 66/67, pp 333-339.

Veith, J. A. and Sposito, G. (1977). On the use of the Langmuir equation in the interpretation of "Adsorption" phenomena. Soil Soc. Am. J., vol 41, pp 697-702.

Watts, H. L. and Utley, D. W. (1953). Volumetric analysis of Sodium Aluminate in solution: determination of carbonate, hydroxide and alumina. Analytical Chemistry, vol 25, p864.

Watts, H. L. and Utley, D. W. (1956). Sodium gluconate as a complexing agent in the analysis of aluminium compounds. Analytical Chemistry, vol 28, pp 1731-1735.