Guy Forté, Chemist

Process Consultant

Alcan International Limited

Arvida Research and Development Centre

1955, Mellon Blvd.

P.O. Box 1250

Jonquière, Quebec

Canada, G7S 4K8



This paper describes a process to desilicate either Bayer pregnant liquor or Bayer spent liquor supersaturated in silica, it applies also to other sodium aluminate solutions, silica supersaturated.

Silica in Bayer liquor is an undesirable impurity for two main reasons: first, it co-precipitates with alumina trihydrate and contaminates the product; second, it precipitates as sodium aluminum silicate (sodalite/cancrinite, desilication products, DSP) on surface of heat exchangers, thus reducing the heat transfer coefficient.

This process removes dissolved silica from a Bayer solution by contacting the solution in a moving packed bed of coarse quartz particles in a reactor, whereby the dissolved silica precipitates as desilication products. This process is carried out at temperatures between 60 and 140oC.

The solution can be filtered in situ and the filtrate, less supersaturated in silica, is immediately available either as pregnant liquor for alumina trihydrate precipitation or as spent liquor to be preheated in heat exchangers before being fed to the digestion. The coarse solids, which contain the desilication products, are withdrawn from the reactor. The process is fully autogenous.

The recovered desilication product can then be processed to recover valuable soda and alumina or be available for sale.


silica alumina, dsp, precipitation, desilication, quality


Guy Forté


In the Bayer process for producing alumina from bauxite, the bauxite containing aluminium hydroxides is contacted with solutions containing caustic soda to dissolve the aluminium hydroxides and some of the silica minerals, such as kaolinite, while leaving most of the remaining constituents of the bauxite essentially unattached in the solid form. The silica, which is dissolved by the caustic soda solution, forms a soluble sodium silicate. This reacts relatively slowly with the sodium aluminate in solution to form complex hydrated sodium aluminium silicates, known collectively as "desilication products". These desilication products, which include Bayer sodalite, are of low solubility in the resulting sodium aluminate-caustic solutions and eventually precipitate out of the solution.

Silica is a highly undesirable impurity in the Bayer liquor for two main reasons: (1) when present in a super-saturated aluminate liquor, it will co-precipitate with the alumina trihydrate downstream and will contaminate the product, (2) when present in the Bayer spent liquor at a concentration above its solubility, it tends to precipitate mainly on the surface of heat exchangers, thus greatly reducing the heat transfer coefficient.

In order to keep the silica concentration in the liquor at an acceptable level, most current Bayer plants maintain the digestion conditions for the extraction of alumina from the bauxite for a time sufficient so that the extracted aluminium hydroxides and the dissolved silica have time to form the desilication products prior to the separation of the unattached solids. This process may take up to one hour. In certain cases, the digestion step may also be preceded by a pre-desilication step performed at lower temperature, which is aimed at dissolving some silica in order to favour the subsequent silica precipitation reaction. Because of the longer residence time required at digestion conditions (high temperature and pressure) to allow the silica precipitation reaction to occur, premature alumina crystallization will occur and hence reduce productivity. In addition, a longer residence time means that larger size pressure vessels are required and energy requirements are higher.

This paper will present an alternative process to desilicate Bayer liquors. This process removes dissolved silica from silica supersaturated sodium aluminate solutions, either Bayer pregnant or Bayer spent liquor, by contacting the liquor with a slowly stirred bed of coarse silica sand (quartz) particles or a mixture of coarse quartz and sodalite particles. The pilot plant set-up and the experimental procedure will be described. The test results will be presented and the process industrial potential will be discussed.


The process set-up used at the Alcan International Ltd., Arvida Research and Development Centre, pilot plant and in the Alcan Smelters and Chemicals Ltd., Vaudreuil Plant in Jonquière, Québec is presented in Figure 1.



  • Figure 1

    Bayer Liquor Desilication Set-up.

    The Bayer plant liquors, either the spent liquor or the pregnant liquor, were stored in a feed tank, from where it was pumped to the 7.6 L Parr reactor through a positive displacement pump and a steam heater. This heater increased the liquor temperature to about 95° C. The reactor is a standard equipment from Parr Instrument Company, Moline, Illinois, USA. The model number is: 4552-HDM-MO-230-VS-2000-BDV-CC-4842-TDM-HTM-SOLCHG-GEARDR. The reactor has been previously filled up with either pure quartz or a mix of quartz and sodalite. The reactor was heated externally and electrically with the result that the wall temperature was around 200° C for tests at 140° C. The pressure in the reactor was maintained at 825 kPa with a control valve.

    To maintain the solid packed bed of coarse quartz particles slowly in motion, a home-made ribbon type agitator was used. The liquor leaving the reactor with some entrained fine solids fed a FundabacR candle filter with a surface area of 0.012 m2. To minimize heat losses, the filter was heated with steam. The filtrate leaving the filter was cooled through a water cooler but the pressure was maintained at 410 kPa with a control valve. The purpose of it was to maintain a pressure differential of about 410 kPa across the filter. Finally the liquor was released to atmosphere. Most of the desilication product formed accumulated in the reactor, which has to be manually emptied from time to time.

    Liquor samples were taken before (stream 1) and immediately after (stream 2) the reactor and after the filter (stream 3). They were analyzed for caustic, alumina and carbonate by potentiometric titration and for 14 cations, including silica by ICP (Induced Coupled Plasma). The desilication product (stream 4) and the solid accumulation in the reactor was analyzed for particle size distribution by sieving and laser diffraction, for composition by XRF (X-ray Fluorescence) and for crystallography by SEM (Scanning Electronic Microscopy), EDX (Energy Dispersion of X-ray), XRD (X-ray Diffraction) and IR (Infra-red).


    The operation was continuous, 24 hours per day, 7 days per week. In total, 25700 liters of plant spent liquor and 5400 liters of decanter overflow liquor (pregnant liquor) were processed over the equivalent of 117 days. The spent liquor was filtered to avoid alumina trihydrate precipitation during prolonged storage but the pregnant liquor came directly from the plant liquor decanter overflow at near the boiling point temperature. The test’s temperature range has been 60 to 140° C and the agitation speed between 7 and 68 rpm. The residence time in the moving packed bed varied between 2 and 44 minutes while the total residence time in the reactor (i.e. packed bed and free space) went from 5 to 54 minutes. The flow rates were between 4.4 and 51.4 L/hour and the solids concentration in the reactor between 650 and 1250 g/L. Overall, about 30 different conditions were tested.


    The test results will be presented in three sections: the desilication efficiency for spent liquor, the desilication efficiency for pregnant liquor and the seed behavior in terms of particle size distribution and chemical composition.

    4.1 Spent Liquor Desilication Efficiency

    The spent liquor desilication efficiency, defined as the reduction of the silica supersaturation in the solution, is presented in Figure 2 as a function of the temperature at a constant residence time and a fixed agitation speed.

    As expected, the desilication efficiency is sensitive to the reaction temperature, all other things being equal (residence time, agitation and liquor silica supersaturation which was around 0.5 g/L SiO2). It went from 55 % to more than 90 % as the test’s temperature increases from 100 to 140° C.

    The desilication efficiency is also a function of the residence time as it is shown in Figure 3.

    For a given spent liquor silica supersaturated (about 0.5 g/L SiO2) at 140oC and agitation at 17 rpm, the desilication efficiency is less than 50 %, with 5 minutes residence time in the reactor, while it is above 90 % at 40 minutes and more.

    Figure 2

    Spent Liquor Desilication Efficiency as a Function of Temperature

    Figure 3

    Spent Liquor Desilication Efficiency as a Function of Residence Time

    4.2 Pregnant Liquor Desilication Efficiency

    Typical test results are presented in Table 1. For the first series of test, the filtration set-up was as described in section 2.0, but for the second series, the filter has been in the reactor itself, so all the fines were kept in the vessel.

    Table 1

    Typical Results – Pregnant Liquor Desilication

    Test Conditions

    Filtration Outside Reactor

    Filtration in Reactor

    Temperature oC



    Residence in Reactor min.




    Feed Liquor


    NaOH as Na2CO3 g/L



    Al2O3 g/L



    Na2CO3 g/L



    SiO2 g/L




    Treated Liquor


    NaOH as Na2CO3 g/L



    Al2O3 g/L



    Na2CO3 g/L



    SiO2 g/L




    SiO2 Solubility* g/L




    Desilication Efficiency


    Reduction) %



    * based on Oku-Yamada solubility equation

    Again the desilication efficiency is defined as the reduction of the silica supersaturation in the solution. For pregnant liquor, the silica solubility in g/L is defined by the Oku-Yamada equation (Oku, 1971).

    [SiO2] eq. = 1.58 x 10-5 x A x C (1)

    where A is the alumina concentration in g/L and C is the caustic concentration in g/L expressed as sodium carbonate.

    It can be seen, from Table 1 above, that up to 60 % desilication efficiency can be achieved with this process at 140oC and about 30 minutes of residence time with feed pregnant liquor of relatively low silica supersaturation at about 0.25 g/L SiO2. This pregnant liquor contained up to 1.2 g/L of red mud solids.

    4.3 Particles Size Distribution

    The original coarse quartz used to initiate the desilication reaction in the reactor, analysed >97 % SiO2 with a median particle size of 1.11 mm and about 10 % <0.9 mm and a maximum of  % >1.7 mm.

    After about 20 days of continuous operation, the solids in the reactor had a median particle size of 400 microns with 5 % >1.7 mm and 40 % <250 microns. After about 80 days, the median particle size seems to stabilize at 225 microns.

    The solids collected in the filter at the beginning of the tests were very fine at <5 microns, but the solids collected after 60 days of operation had a median size of 100 microns and, after 80 days, it was at equilibrium with the solids in the reactor at 225 microns of median size. The solids filtration rate has been good at 4.3 m3/(h-m2) and 4.2 kg/(h-m2).

    4.4 Solids Chemical Composition

    As mentioned above, the starting solid to initiate the desilication reaction has been pure (>97 % SiO2) coarse quartz. The surprise has been to realize that this quartz was slowly and completely transformed into Bayer sodalite as the desilication reaction progressed without destroying its particle size. Figure 4 shows a transversal cut of a polished section of a quartz particle after 4 days of operation. Bayer sodalite started to form in the upper right corner and an X-ray mapping of the same figure presented in Figures 5a and 5b, clearly indicates a silicon core on the left side and presence of sodium and aluminum in the right corner. More than that, one can see, in the middle of Figure 5b, that sodalite formation is progressing toward the centre of the particle.

    After about 20 days of operation, this phenomena is completed; XRD and XRF analysis confirmed this. The solids in the reactor analyzed <2 % quartz and >97 % DSP. A typical elemental analysis is 7 % loss of mass, 30.7 % Al2O3, 35.6 % SiO2, 24.3 % Na2O, 0.2 % Fe2O3 and 0.3 % CaO.


    The above results indicate not only very good desilication efficiency (~90 %) within a reasonable residence time (~40 minutes) for spent liquor, but also good results on pregnant liquor with 60 % desilication after about 30 minutes of residence time, both at 140° C.

    This process uses low cost, commercial quartz sand as starting material. This quartz is completely transformed into Bayer sodalite product after about 20 days. The relatively slow agitation maintains the solid activity in such a way that the process is autogenous. In other words, the desilication efficiency is self-maintained internally (in the reactor) without external seed recycling and/or activation. The pure Bayer sodalite product is easily filter- able. The presence of red mud, up to 1.2 g/L in pregnant liquor, does not reduce the process efficiency. Other than alumina consumption associated with the DSP formation, the alumina losses related to this process are very low.

    Figure 4

    Polished section, transversal cut, magnification 200X, solid from the reactor after 4 days of operation.

    Figure 5a

    X-ray mapping of Figure 4 showing silicon.

    Figure 5b

    X-Ray Mapping of Figure 4, showing sodium and aluminium.


    Although the reactor walls were at relatively high temperature (200oC) because externally and electrically heated, they were free of Bayer sodalite scale because of the rubbing action of the coarse particles entrained by the ribbon type mixer.



    Compared with other processes, this process has the advantage of being autogenous, it uses inexpensive commercial solids (quartz) to initiate the reaction and when linked with the Alcan pressure decantation technology, it has a good industrial potential application.

    It is possible to imagine a Bayer circuit where low silica bauxite could be processed without predesilication step, with short digestion residence time, say 10 minutes, followed by a solid-liquid separation in a pressure decanter at 140oC, followed by the desilication process described in this paper, also at 140oC. The filtration step will be in situ, in the desilication reactor, and will not only remove the entrained fines DSP but also the fines red mud, no other polishing filter will be required. Finally, the desilicated pregnant liquor will be flashed to the appropriate temperature to precipitate alumina trihydrate. The Bayer desilication products which are of good quality will be removed from the desilication reactor and soda and alumina could be recovered by known means (Cresswell, 1984). This process will have the following advantages: low caustic consumption because of low silica bauxite, no capital investment in predesilication, low capital investment in digestion because of short residence time, less energy consumption because of filtration in situ at around 140oC, maximum liquor productivity because of higher charging ratio in digestion (no limitation in filtration), better performance in precipitation and finally better alumina quality, less silica.

    A similar scenario is possible with spent liquor and this is mainly advantageous for the high temperature digestion plants. A spent liquor stream, preheated to around 140oC, can be desilicated with the process described in this paper, thus reducing the DSP scale formation in the high temperature heat exchangers. The heat transfer coefficient will increase, the steam consumption will decrease as well as the maintenance descaling costs.


    Cresswell, P.J. and Milne, D.J. " Hydrothermal Recovery of Soda and Alumina from Red Mud " Light Metals 1984, Comalco.

    Oku, T. and Yamada, K., " Dissolution of Quartz and the Rate of Dissolution in the Bayer Liquor ", Light Metals 1971, (ed. T.G. Edheworth), AIME, NY, 1971.