water research 43 (2009) 762–772

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Removal of sulfate from high-strength wastewater

by crystallisation

Stephan Taita, William P. Clarkeb, Jurg Kellera, Damien J. Batstonea,*

a
Advanced Water Management Centre, University of Queensland, St Lucia, Queensland 4072, Australia
b
Division of Environmental Engineering, School of Engineering, University of Queensland, St Lucia, Queensland 4072, Australia

article info abstract

Article history: Sulfate causes considerable problems in anaerobic digesters, related to generation of
Received 13 June 2008 sulfides, loss of electrons (and hence methane), and contamination of gas streams.
Received in revised form Removal of sulfides is generally expensive, and still results in methane losses. In this
19 October 2008 paper, we evaluate the use of precipitation for low-cost sulfate removal, in highly
Accepted 7 November 2008 contaminated streams (>1 gS L1). The main precipitate assessed is calcium sulfate
Published online 21 November 2008 (gypsum), though the formation of complex precipitates such as jarosite and ettringite to
remove residual sulfate is also evaluated. The four main concerns in contaminated
Keywords: wastewater are:- high solubility, caused by high ion activity and ion pairing; slow kinetics;
Crystallisation inhibition of nucleation; and poisoning of crystals by impurities, rendering product
Sulfate unsuitable for reuse as seed. These concerns were addressed through batch experiments
Sulfide on a landfill wastewater with a similar composition to other sulfate rich industrial
Wastewater wastewaters (high levels of organic and inorganic contaminants). Crystallisation rates were
Gypsum rapid and comparable to what is observed by others for pure solutions (2–5 h). The kinetics
Ettringite of crystallisation showed a 2nd order dependence on supersaturation, which have impli-
Jarosite cations for crystalliser design, as discussed in the paper. No spontaneous nucleation was
observed (seed was required). Seed poisoning did not occur, and product crystals were as
effective as pure seed. Solubility was increased by an order of magnitude compared to
a pure solution (2.6  103 M2 vs. 0.22  103 M2). As evaluated using equilibrium modelling,
this was caused equally by non-specific ion activity, and specific ion pairing. Jarosite and
ettringite could not be formed at reasonable pH and temperature levels. Given the lack of
complex precipitates, and relatively high solubility, gypsum crystallisation cannot practi-
cally be used to remove sulfate to very low levels, and gas-sulfide treatment will likely still
be required. It can however, be used for low-cost bulk removal of sulfate.
ª 2008 Elsevier Ltd. All rights reserved.

1. Introduction sewers during wastewater discharge, and leads to serious

difficulties during on-site anaerobic wastewater treatment,
Sulfate is a common contaminant in industrial wastewaters, due to sulfate reduction to sulfide, as described below:
including effluents from paper mills, fertilizer production,
textile industries, pesticide production, mining of metallif- 1. The production of sulfide is an occupational health and
erous sulfide deposits, root crop processing such as ginger and safety issue and a maintenance problem as hydrogen
beet and aluminium anodising. Sulfate causes corrosion in sulfide is a poisonous gas and, when being oxidised to SO2

* Corresponding author. Tel.: þ61 7 3346 9051; fax: þ61 7 3365 4726.
E-mail address: [email protected] (D.J. Batstone).
0043-1354/$ – see front matter ª 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2008.11.008
 water research 43 (2009) 762–772 763

and then sulfuric acid, corrodes concrete and metal struc- as well as re-adjustment prior to subsequent anaerobic
tures and gas engines. treatment.
2. Sulfide is a wastewater contaminant, contributes to BOD The application of gypsum and ettringite crystallisation to
and COD, and produces a high-odour wastewater product. treat sulfate rich acid mine waters is well documented. Gel-
3. For every mole of sulfate reduced, one mole of potential denhuys et al. (2003) developed a gypsum crystallisation
methane is lost, so the methane yield may be reduced by process for treatment of acid mine waters where limestone is
a considerable amount. used to neutralize acidic feed water (pH 2–3). Lime is used to
raise the pH to 12 in order that most of the metals in the water,
Methods to manage sulfide include precipitation by iron (or including magnesium, precipitate out as metal hydroxides.
other metals) addition, and removal in the gas stream with Limestone is recovered post-treatment as calcium carbonate
either sulfide removal in a caustic scrubber, reaction with precipitates by sparging with carbon dioxide (from the
packed beds of iron or zinc, or biological sulfide oxidation. neutralization reaction). This also neutralizes the pH. With
Both these options have relatively high operating and/or the use of flocculants for solids recovery this process removes
capital costs, and do not alleviate the potential loss in most metals, and sulfate to a level of 400 mgS L1 with
methane production. a practical amount of added calcium. Smit and Sibilski (2003)
An alternative is to remove sulfate from the wastewater describe a pilot scale application of a process similar to that of
prior to biological reduction to sulfide. Available methods Geldenhuys et al. (2003), but with ettringite crystallisation to
include ion exchange, ion selective membranes and crystal- reduce sulfate in the treated mine water to less than
lisation of sulfate minerals. Of these methods, crystallisation 50 mgS L1. While these methods have been used for heavily
is likely the most economical and much simpler to operate contaminated physicochemical streams, scientific literature
with high-strength wastewater (organics and inorganics). As on crystallisation for sulfate removal from organically
an added incentive, wastewaters rich in sulfate are often rich contaminated wastewaters is minimal. Some of the major
in metals which can support the crystallisation of sulfate concerns with sulfate removal from wastewater with both
minerals without further additives (for example calcium in inorganic and organic contaminants are as follows.
paper mill wastewater for calcium sulfate crystallisation).
Metal cations which precipitate with sulfate, and are also
1.1. Solubility
available in the form of low-cost additives, are limited. The
choice is likely restricted to calcium, to form calcium sulfate
Impurities in industrial wastewater may strongly influence
dihydrate (gypsum), a marginally soluble mineral. As stated
the solubility of the mineral being precipitated, due to, for
above, calcium is often present in the wastewater, and if
instance, non-ideal solution behaviour such as ion pairing and
added, lime is one of the lowest cost commodity chemicals
ion activity. To give an example, the solubility of calcium
available. In addition, co-precipitation to form minerals which
sulfate (gypsum) in pure water is approximately 2 g L1 over
are less soluble than gypsum may be available to provide
a range of temperatures (Seidell, 1940), similar to the solubility
enhanced sulfate removal. Examples include:
reported by Geldenhuys et al. (2003). In comparison, in the
experiments of Kabdasli et al. (1995) with textile wastewater,
1. The calcium-aluminium-sulfate mineral, ettringite,
the solubility of calcium sulfate is approximately 8 g L1. To
precipitated by the presence or the addition of aluminium
calculate this solubility value, it is assumed that the final
(for example as a calcium aluminate) and calcium at high
sulfate concentrations of Kabdasli et al. (1995) were at equi-
pH (>12). With this mineral the sulfate levels can drop to
librium. The main concern is that a higher solubility requires
less than 200 mgS L1 with reasonable amounts of
a larger concentration of calcium to reduce the sulfate
aluminium and calcium (Geldenhuys et al., 2003; Smit and
concentration to desired levels.
Sibilski, 2003).
2. The mineral jarosite, which forms by the combination of
iron, sulfate and another cation which may be ammonium, 1.2. Crystal growth rates
potassium or sodium. This mineral is commonly precipi-
tated at acidic pH (<2.8) in the metallurgical recovery of Substantial differences in the rates of gypsum crystallisation
zinc, where iron(III) must first be removed from solution. in wastewater are reported in the literature, ranging from an
Although conceptually discussed in the paper of Wang et al. impractical 1 week crystallisation time to reach near equilib-
(2006), no references were found in the literature to large rium conditions with textile wastewater (Kabdasli et al., 1995),
scale application of jarosite precipitation for sulfate to a feasible 3 h to reach equilibrium in acid mine drainage
removal from industrial wastewaters. (Geldenhuys et al., 2003). In synthetic solutions, gypsum
crystallisation attains equilibrium within 1–2 h. These differ-
Two major possible drawbacks with precipitation of jar- ences are a major concern due to the importance of crystal-
osite and ettringite are that significant crystallisation rates lisation rates in determining the crystalliser vessel size (and
may only be realised at elevated temperatures (60–90  C; see associated capital and operating costs) for a particular feed
Elgersma et al., 1993; Alvarez-Ayuso and Nugteren, 2005; water flow and influent sulfate concentration. These differ-
Christoe, 1976) and the minerals are usually only stable at ences could have resulted from either:
extreme pH conditions (as indicated above). These crystal-
lisation conditions are unfavorable due to substantial energy 1. inhibition of crystal growth by impurities in the textile
costs for heating, and chemical costs for the adjustment of pH, industry wastewater used by Kabdasli et al. (1995), which
764 water research 43 (2009) 762–772

are not experienced with acid mine water used by Gel- (B) The same sample as (A), but treated to remove organic
denhuys et al. (2003), or acids. The acids were evaporated off at pH 4.0 and 80–
2. inhibition of crystal nucleation by impurities in waste- 100  C, over a 24 h period, while keeping the concentra-
water. The experiments of Kabdasli et al. (1995) and Gel- tions of inorganics similar by re-dilution with purified
denhuys et al. (2003) implicate this as a critical factor, water (resulted in 87% reduction in organic acids).
because while crystallisation rates were impractically slow
in the textile wastewater experiments by Kabdasli et al., the (C) A long term aggregate sample consisting of leachate orig-
crystallisation rate was fast and practical in Geldenhuys inating from the leachate collection system. This sample
et al.’s experiments on acid mine drainage where seed was collected over a single 4 h period in April 2007, from
crystals were added on which crystal mass could readily a storage lagoon onsite.
deposit.
All these samples were centrifuged (30 min at 9000 g) and
If these differences are caused by crystal growth inhibi- filtered (0.22 mm polyethersulfone (PES) membrane, Millipore,
tion (see Hamdona and Al Hadad, 2007) crystallisation will Ireland) to limit complications in the experiments caused by
not be effective at removing sulfate. If inhibition of nucle- insoluble impurities in the wastewater. The wastewater was
ation is responsible, the crystallisation rate can readily be not pre-treated in any other way prior to use.
increased by the addition of sufficient quantities of seed
crystals. 2.2. Analysis

1.3. Re-use of crystals The major elemental species for solubles in the wastewater
were quantified with Inductively Coupled Plasma Optical
The work of Hamdona and Al Hadad (2007) showed clear Emission Spectroscopy (ICP-OES) using a Varian Vista Pro ICP-
inhibition on crystal growth from seed crystals by various OES instrument operating at a forward power of 1200 W uti-
heavy metal cations in solution and adsorption of these lising a Sturman – Masters spraychamber (Varian, Palo Alto,
metal ions onto growth sites on the surface of these crys- CA). Sulphur as sulfate was shown by comparison with ion
tals. However, no studies could be found that investigated chromatography data (method details follow) to typically
whether such growth inhibition is amplified with re-use of represent greater than 90% of the total sulphur content of the
the crystals in subsequent crystallisations. This matter is of wastewater. Accordingly, it was assumed that the sulphur
concern in the practical application of crystallisation for content determined with ICP-OES was a direct measure of the
removal of sulfate, because re-use of crystal product as seed sulfate content of the wastewater (in the absence of biological
is far more economical than continual use of manufactured activity). Ammonia nitrogen, nitrate/nitrite and phosphate
seed. were measured on a Lachat QuickChem8000 flow injection
The present investigation addresses the listed concerns analyser as per the Lachat QuickChem Method 31-107-06-1-A.
through a series of crystallisation experiments, using landfill Chloride, sulfide, thiosulfate, sulfite and sulfate concentra-
wastewater with a similar composition to other sulfate rich tions were measured by ion chromatography (IC) using a Dio-
industrial wastewaters. To further interpret issues related to nex ICS-2000 IC System fitted with an AS50 Autosampler and
solubility, the results are also assessed with an aqueous AD25 Absorbance Detector (Dionex, Sunnyvale, CA). A RFIC
equilibrium model. In addition to this, this paper evaluates IonPac AS18 column was used with the instrument. The
whether co-precipitation with other trace minerals can sample was eluted with a KOH gradient (12–52 mM) and
assist in sulfate removal, by formation of ettringite and absorbance was measured at a wavelength of 230 nm. The IC
jarosite. measurements were performed at 35  C. Volatile fatty acids
(VFAs) were determined with a Perkin-Elmer GC. The column
was DB-FFAP 15 m  0.53 mm  1.0 mm (length  ID  film)
2. Materials and methods kept at 140  C during the analysis, while the injector and flame
ionisation detector (FID) were operated at 220  C and 250  C,
2.1. Sample sources respectively. High purity helium was used as a carrier gas at
a flow rate of 17 mL min1. The pH of the wastewater in all
Wastewater samples were collected from a landfill in a void cases were measured using a calibrated TPS pH probe (TPS,
produced by mining activity in a large sulfide deposit in Tar- Springwood, Australia).
ago, Canberra, Australia. The landfill received predominantly The mineral phases present in the crystals used and
municipal solid waste from Sydney, Australia. Three key generated in the experiments were identified with powder
samples of wastewater were used in the investigation; X-ray diffraction analysis and scanning electron microscopy
(SEM) in the back scattering mode. Particle settling velocity
size distributions were measured with an Andreasen Pipet
(A) An aggregate of samples collected at locations with method described by Farmer and Beckman (1984). Volume
a representative spread over the landfill void (from equivalent size distributions were measured by laser
leachate collection system at different depths, from diffraction with a Malvern Mastersizer 2000 (Malvern
leachate storage dams and from groundwater weeping Instruments, Worcestershire, UK) after ultrasound treat-
points on the wall of the mine void). All these samples ment to break up any loose agglomerates. Measured size did
were collected over a single 4 h period in February 2007. not change substantially by sonication. Specific surface area
 water research 43 (2009) 762–772 765

of the crystals was estimated from geometric analysis of 5. Re-use of seed crystals. 400 mg of the (dried) crystal product
SEM images. from experiment 1 was used to seed an additional experi-
ment with 100 mL of fresh sample (A).
2.3. Crystallisation experiments 6. Control crystallisation experiments in which gypsum seed
crystals were not added, were performed in parallel with
2.3.1. Seed crystals the other experiments.
Synthetic gypsum seed crystals were prepared by mixing
equal volumes of 0.6 M aqueous sodium sulfate and 0.6 M
aqueous calcium chloride. These solutions were prepared 2.3.4. Experimental sampling
with analytical reagent grade chemicals. While crystallising, In each experiment, a 10 mL sample of the crystalliser
the mixture was stirred (800 rpm) with a 6-bladed Rushton contents was collected immediately after the addition of the
turbine in a 1 L cylindrical glass vessel with four baffles gypsum seeds and typically again at 30 min, 1 h, 1.5 h, 2 h,
(Tait, 2007). The sodium sulfate was added drop-wise 2.5 h, 3 h and 24 h (considered to be at equilibrium) after the
(approximately 3 mL min1) up to a final mixture volume of addition of the gypsum seed. These samples were filtered
800 mL. To complete the crystallisation the mixture was left through 0.2 mm syringe filters (PES membrane) and kept
to stir for an additional 2 h. The crystals were filtered and refrigerated for subsequent measurements of the concentra-
rinsed with distilled water to remove any adhering sodium tions of sulfate, calcium and other metals in solution. At the
chloride. The water was displaced with absolute ethanol completion of each experiment, the crystal product was
and the ethanol was evaporated off in an incubator at 45  C. centrifuged (30 min at 9000 G), filtered (0.45 mm nitrocellulose)
Crystal seed prepared in this manner were used in all the and rinsed with 99% ethanol. Any residual ethanol was
experiments. evaporated off in an incubator at 45  C. The dry product was
retained for investigation of particle size, particle morphology
2.3.2. Crystallisers and the mineral phase composition.
Crystallisers used in the experiments were (a) 250 mL Schott
bottles, with working volumes of 100 mL, stirred with 50 mm 2.3.5. Modelling of crystallisation kinetics
magnetic bar stirrers (300 rpm) or (b) a 1 L glass vessel, also The apparent crystallisation rate (RCaSO4 , with units of M h1)
used in the preparation of crystal seed as detailed above. All was fitted with the expression (Hina and Nancollas, 2000):
experiments were performed at room temperature (19  C– !12 !n
 23
22  C). m ½Ca½SO4 
RCaSO4 ¼ kg S0 1 (1)
m0 ½Caeq ½SO4 eq
2.3.3. Experiments
The following experiment types were performed: where the product of the equilibrium concentrations of
calcium and sulfate (total, [Ca]eq[SO4]eq) was measured at the
1. Crystallisation without added counter-ions. 400 mg of synthetic completion of a number of the crystallisation tests. The
gypsum seed crystals were added to 100 mL of sample (A), surface area of the added seed crystals on which crystal mass
(B) or (C), and to deionized water as a control experiment. could deposit (S0, with units m2 L1) and the initial mass of
2. Crystallisation with excess calcium. Synthetic gypsum seed crystals in the crystallising solution (m0) were known. The
crystals and 1 g of excess calcium (3.68 g as analytical grade total molar concentrations of calcium and sulfate in solution
calcium chloride dihydrate) were added to 100 mL of ([Ca] and [SO4]) were measured continuously during an
sample (A). This level of excess calcium was selected based experiment and the mass of crystals in the crystallising
on experimental trials, to give a final sulfate level of solution at any time (m) was determined from a mass balance
approximately 1 g L1. over the crystallisation. kg (with units mol h1 m2) and n
3. Inducing co-precipitation. 6.04 g of analytical grade (dimensionless) were kinetic fitting parameters. Best-fit
aluminium (III) chloride hexahydrate or 6.76 g of analytical values for these kinetic parameters were estimated by a non-
grade iron (III) chloride hexahydrate or 5.59 g of analytical linear parameter estimation technique according to Batstone
grade potassium chloride anhydrous was added as powder et al. (2003), implemented within the software package
salts to 100 mL of sample (C). The total added concentra- Aquasim 2.1d (Reichert, 1994). The main limitation of Equa-
tions of the cations in the wastewater were 0.25 M for tion (1) is the assumption of proportionality in volume and
aluminium and iron and 0.75 M for potassium and the area – S0(m/m0)2/3 – which is a scaling rule of the surface area of
concentration of chloride was near identical in each case. the crystals in suspension with respect to their mass. This
The added reagents significantly acidified the wastewater assumption holds for systems with size independent growth
from its native value of 5.0, so the solution pH was adjusted and negligible nucleation.
to 7.0 by adding analytical reagent grade Ca(OH)2 (typically
3.2 g of anhydrous calcium hydroxide in 100 mL of waste-
water). After pH adjustment, 400 mg synthetic gypsum seed 2.4. Modelling of aqueous equilibria
crystals were added.
4. Crystallisation at increased pH. The pH of 100 mL of sample (C) Aqueous equilibria simulations were performed with the
was adjusted to 7.0 or 9.0 with analytical reagent grade software package PhreeqC (Parkhurst and Appelo, 1999). The
Ca(OH)2 salt and 400 mg of synthetic gypsum seed crystals solubility product constants for gypsum and potassium jar-
was added. osite were in the PhreeqC software database. The solubility
766 water research 43 (2009) 762–772

product constant for ammonium jarosite was assumed to be

Table 1 – Wastewater matrix characteristics.
approximately equal to the measured ionic product reported
by Elgersma et al. (1993) for pH 1.5 and 95  C (log Ksp ¼ 91). Parametera Units Sample Ab Sample Bb Sample Cb
The solubility product constant for ettringite was assumed to Ammonia g NH4–N L1 2.2 2.2 1.1
be approximately equal to the value estimated by Warren and Calcium g L1 4.1 2.9 1.1
Reardon (1994) (log Ksp ¼ 44.91 over the pH range 10–13, at Chloride g L1 3.3 16.3 –
25  C, for Ca6Al2(SO4)3(OH)12.26H2O). Because (a) sulfate Iron g L1 1.4 1.4 0.48
Magnesium g L1 1.0 1.0 1.7
reduction is relatively slow compared to crystallisation, and
Phosphorus mg L1 20 20 7
(b) actual samples were far from equilibrium with respect to
Potassium g L1 2.1 2.1 1.0
organics, sulfate, and sulfide (indicating minimal sulfate Sodium g L1 3.1 3.1 1.5
reduction), the sulfate reduction to equilibrium was excluded Sulfate gS L1 1.8 1.1 3.8
from the model. To represent the organic acid content of the Total g L1 20 15 17
wastewater, acetic acid was defined as a new species in the organic
model database. The solution temperature for each simula- carbon
Total mg L1 200 300 –
tion was set at 25  C and the pH was set at a number of values
inorganic
within the range investigated in the experiments (4.0–9.0). carbon
Chloride ion concentration was allowed to vary to balance the pH standard 5.5 3.8 5.0
charge of the solution. PhreeqC permitted simulation of units
equilibrium conditions for a number of complexes simulta- Acetic acid mg L1 10800 3000 5100
neously (for example gypsum, lime and ettringite). Simula- Hexanoic mg L1 1500 34 600
acid
tions were performed for hypothetical batch crystallisations
Iso-butyric mg L1 10400 270 3500
to equilibrium (or dissolutions if the wastewater was under-
and
saturated with respect to mineral phases). The relative influ- butyric
ence of ion activity and specific soluble ion pairs on the acid
solution equilibria was investigated by sequentially elimi- Propionic mg L1 3000 260 1200
nating these influences in virtual experiments. Ion activity acid
effects were removed by nominating parameters for each ion Valeric mg L1 1400 0 600
and
species which would give them activity coefficients of 1. Ion
Iso-valeric
association effects were eliminated by deleting the ion asso-
acid
ciation relationships from the model database. Total VFAs g L1 27.0 3.5 11.0

a Other elements, including Ba, Cr, Ca, Cu, Mg, Li, Pb, and Sr were
all below the 1 mg L1 level. Zn, Si and Al were approx. 100 mg L1.
3. Results b Sample A – From wastewater collection system; Sample B –
Sample A which was treated in the laboratory to reduce levels of
3.1. Wastewater matrix VFAs; C – From wastewater storage dam.

The measured compositions of the 3 wastewater samples

used in this study are listed in Table 1. As is typical of young
landfills (Cecen and Gursoy, 2000), the wastewater contained behaviour. A larger product corresponds to a higher solubility.
a high level of organics, mainly as organic acids (or volatile Accordingly, it could be stated that gypsum was approxi-
fatty acids, VFAs), and was heavily contaminated with nutri- mately 3 times more soluble (10 times higher [Ca]eq[SO4]eq) in
ents and metal salts. The landfill water was heavily buffered the landfill wastewater than in a pure aqueous solution
due to the presence of these impurities. Low pH values, typical saturated with respect to gypsum. This added non-ideal
of waters affected by acid mine drainage, was apparent. The behaviour (pure saturated solution also non-ideal) was not
wastewater contained a considerable background of calcium influenced by: (i) reduced volatile organic acids content (as
and consequently appeared to have good potential to precip- analysed by sample B) or (ii) increased inorganics as assessed
itate gypsum without any added calcium. This potential was by increasing pH and adding excess calcium, iron, aluminium
confirmed by the desktop simulations and in the actual or potassium. Thus, the equilibrium of gypsum crystallisation
experiments by an equi-molar decrease in the soluble from the wastewater was much less favourable for the
concentrations of calcium and sulfate over time, suggesting removal of sulfate than the equilibrium of a pure aqueous
the precipitation of calcium sulfate. gypsum solution.

3.2. Equilibrium of crystallisation 3.2.2. Desktop simulations

Experimental results were in qualitative agreement with
3.2.1. Experiments results from the desktop simulations of gypsum crystal-
Table 2 lists the apparent solubility for gypsum as products of lisation. Typically, [Ca]eq[SO4]eq values from the desktop
the total concentrations of calcium and sulfate in solution at simulations and the [Ca]eq[SO4]eq values measured in the
equilibrium [Ca]eq[SO4]eq (correspond to total elemental experiments differed by less than 50% (Table 2). This obser-
concentrations of Ca and S at equilibrium). The [Ca]eq[SO4]eq vation indicated that the model used could represent the real
values listed in Table 2 included all non-ideal solution system. It also showed that ionic effects in solution largely
 water research 43 (2009) 762–772 767

Table 2 – Multiples of the equilibrium concentrations of total dissolved calcium and sulfate.
Condition [Ca]eq[SO4]eq (103 M2)a

Literature Experiment Simulation

b
Sample A Sample C
c
Pure water þ gypsum Infinitely dilute solution 0.026
[Ca2þ][SO2
4 ]
Saturated solution 0.22d 0.24
Wastewater þ gypsum As is 2.6 (27%)e 1.8 (27%) 1.7f
crystallised Reduced VFAsg 2.5 (15%)
No ion pairing effects 0.63f
No ion activity effects 0.83f
With added excess 3.1 (30%)
calcium
With added excess Al, 2.4 (4%)
Fe or K at pH 7.0
pH adjusted 7.0–9.0 1.3 (10%)

a Multiples of total equilibrium concentrations of dissolved calcium and sulfate at 25  C. A higher [Ca]eq[SO4]eq value allows less sulfate removal
by crystallisation.
b Sample A – From wastewater collection system; Sample C – From wastewater storage dam.
c Solubility product (Ksp) which is the product of calcium and sulfate ions in solution for an infinitely dilute solution at 25  C, also with units of
103 M2. The value was taken from PhreeqC database.
d Seidell (1940).
e 95% confidence intervals as relative percentages.
f Simulation with wastewater with composition of Sample A.
g Sample B, which was sample A treated in the laboratory to reduce VFA content.

contributed to the increased apparent solubility, rather than Predictions of the model with Equation (1) using the optimal
interactions of calcium and sulfate with non-ionic impurities [Ca]eq[SO4]eq, kg and n values (above) were valid for signifi-
in the wastewater. A comparison of model simulation results cantly different crystallisation conditions (Fig. 1(C), higher
with and without the effects of ion activity and ion association initial calcium). Both kinetic and equilibrium experimental
in turn (Table 2), showed that these two influences were results were substantially in agreement with the model
equally important in determining the non-ideal equilibrium predictions.
behaviour observed with the wastewater. At native pH (5.5) When no crystal seeds were added to the wastewater,
the anions with the greatest level of ion association with the rate of crystallisation was not significant, as shown by
calcium in the wastewater (non-precipitating) were sulfate the results in Fig. 1(A). This means the wastewater is
and bicarbonate, whereas at high pH (12.0) it was hydroxide metastable with respect to gypsum (no nucleation, but
and sulfate. At native pH (5.5) and high pH (12.0) the cations crystal growth). As shown in Fig. 1(D) (unseeded run in
with the strongest ion association with sulfate in the waste- 250 mL bottles) for a different set of experiments without
water (non-precipitating) were calcium and magnesium. any added crystal seed, the system did not remain meta-
stable indefinitely, and did eventually spontaneously
nucleate.
3.3. Kinetics of crystallisation The rates of crystallisation in the 250 mL crystalliser and
the 1 L crystalliser (very different degrees of mixing) did not
An average apparent solubility constant ([Ca]eq[SO4]eq) of 2.6 differ significantly (Fig. 1 (D)) and neither did the rates of
(27%) 103 M2 (Table 2) was used as a fixed parameter crystallisation when crystal product rather than fresh
(constant) in order to model the observed kinetics of gypsum synthetic seed crystals was used to seed the crystallisation
crystallisation according to Equation (1). (Fig. 1 (D)). The implications of these findings are discussed
When crystal seeds were added the crystallisation was below.
rapid, as shown by representative kinetic data from the
experiments given in Fig. 1. When crystal seeds were added,
equilibrium concentrations of calcium and sulfate were typi- 3.4. Seed crystals and product crystals
cally attained within 2–5 h (see Fig. 1(A)). This indicated
favourable kinetics of crystallisation for sulfate removal. Synthetic gypsum crystals used as seed in the experiments
Kinetic parameters for seeded crystallisation were estimated and product crystals generated by the experiments are shown
with the data set shown in Fig. 1(B). The optimal parameter respectively in Fig. 3(A) and (B). The settling size distributions
values were kg ¼ 0.028 mol m2 h1 and n ¼ 2.1. Fig. 2 displays and volume equivalent size distributions are displayed
the 95% confidence region for these parameters. This region respectively in Fig. 3(C) and (D). Volume weighted mean
indicates that both parameters were significant at the 95% diameters were 37 (1) mm for the synthetic seed, and 35
confidence limit, i.e. no intersection with the zero axes. (1) mm for the product crystals. The specific surface area of
768 water research 43 (2009) 762–772

Fig. 1 – Representative data from the crystallisation experiments for (note change of scale and scale breaks); (A) an
experiment where seed crystals were added and not added for comparison; (B) three seeded experiments for which the
combined data were used to estimate kinetic parameters; (C) an experiment at higher initial calcium than the experiment
for which the kinetic parameters were estimated; (D) experiments at various conditions as listed in the legend.

the synthetic seed was 1.2 (0.2) m2 g1 (used as an input used in other gypsum crystallisation studies with synthetic
parameter for modelling crystallisation kinetics). The seed solutions (see van Rosmalen et al., 1981).
crystals had a similar appearance to synthetic seed crystals Crystal products from all the experiments consisted of
predominantly gypsum. The crystal product from all the
experiments had a similar appearance to the seed, but was
less agglomerated and with some additional finer crystals.
These differences are clearly seen on the crystal size distri-
butions. The crystal product typically contained about 50% by
weight added seed crystals and the remainder (as determined
from a mass balance over the crystallisation) consisted of
mostly gypsum crystallised during the experiments. Where
solution pH was increased (4th group of experiments), small
quantities of metal hydroxides (for example iron hydroxide)
were detected in the product. Both the seed crystals and
product crystals exhibited favourable settling rates for
hydrocyclone and gravity separation (Fig. 3), i.e. 95% of crystal
mass settled faster than 1.6 m h1 in the wastewater matrix
used here.
Whilst simulations predicted that jarosite of ammonia
Fig. 2 – Confidence region at the 95% confidence limit for and ettringite would precipitate at pH 4.0–9.0 and 25  C
kinetic parameters determined from the crystallisation (even in the presence of excess alumina or goethite), no
experiments. jarosite or ettringite could be detected in any of the crystal
 water research 43 (2009) 762–772 769

Fig. 3 – Scanning electron micrographs of (A) synthetic gypsum seed crystals used in the crystallisation experiments and (B)
a typical crystal product from a crystallisation experiment, and the corresponding (C) settling size distributions (with
estimated 95% confidence intervals as dashed lines) and (D) volume equivalent size distributions.

products from the experiments, even when encouraged to be precipitated at mild pH conditions (5–9), even with the
precipitate with a large excess of aluminium, iron and addition of excess iron or aluminium (see Section 3.4). Simi-
calcium (Experiment 3). larly, previous studies report the precipitation of iron
hydroxide rather than jarosite at pH > 3.8 (Wang et al., 2006)
and aluminium hydroxide rather than ettringite at pH < 10.7
4. Discussion (Gabrisova et al., 1991). Consequently, the precipitation of
ettringite and jarosite can only assist sulfate removal from
4.1. Gypsum crystallisation and co-precipitation industrial wastewaters if pH was drastically altered to
extremes not favourable for direct feeding to subsequent
4.1.1. Gypsum crystallisation in metals-contaminated waters anaerobic treatment processes. This is important, since
The wastewater used in this study contained considerable without these complexes, sulfate cannot be effectively
background calcium (see Table 1), such that a significant removed to <100 mgS L1, the minimum concentration
amount of gypsum was precipitated simply by adding gypsum required to truly achieve a low level of sulfide in the gas phase
seed crystals. This is an incentive for using crystallisation to (given substantial methane production). Benefits are therefore
remove sulfate from other calcium rich industrial wastewa- limited to enhancement of methane production, decreased
ters, such as that from a paper mill. The cost savings are not post-sulfide treatment (in liquid via iron, or gas via biological
simply associated with the calcium costs (approximately $US sulfide oxidation), and value-adding to lime by generating the
140 per tonne as CaO in Australia), but also with lime blending, more valuable product, gypsum.
dosing, and storage facilities, the cost of which can be more
than a simple crystallisation and separation system. 4.2. Equilibria

4.1.2. Co-precipitation Similar to the report of Kabdasli et al. (1995), impurities in the
In the present work, the marginally soluble co-precipitants, wastewater of this study made the equilibrium of gypsum
jarosite (ammonium-iron-sulfate-hydroxide complex) and crystallisation unfavourable for sulfate removal (see Section 3.2).
ettringite (calcium-aluminium-hydroxide complex), could not Consequently, a large excess of calcium could be required to
770 water research 43 (2009) 762–772

reduce sulfate to desired levels. Kabdasli et al. (1995) sug- (optimum of n ¼ 2.1). Gypsum crystallisation from synthetic
gested that the sodium content of the wastewater largely solutions at low levels of supersaturation exhibits n values
contributed to the equilibrium of sulfate. From the desktop between 1.5 and 2.1 (Christoffersen et al., 1982). There are two
simulation results, general non-ideal behaviour (ion activity) implications of 2nd order crystallisation kinetics:
accounted for approximately 50% of the increased solubility,
and high sodium concentrations will contribute to this influ- 1. a 2nd or greater order dependence on supersaturation can
ence. However, contrary to what is suggested by Kabdasli et al. indicate that the rate of crystal growth is limited by incor-
(1995), specific ion associations of sulfate with magnesium, poration of crystal mass into the faces of the growing
calcium and ammonia in solution appeared to be more crystals and not simply by mass transfer (linear depen-
influential than specific interactions with sodium. Accord- dence for mass transfer) (Mullin, 2001). This interpretation
ingly, it is reasonable to suggest that Geldenhuys et al. (2003) was shown to be valid by similar rates of crystallisation
in their experiments greatly improved the equilibrium of measured in the 250 mL crystallisers and the 1 L crystalliser
gypsum crystallisation from acid mine water by raising the pH (very different degrees of mixing). Industrially, this obser-
to above 11.5 with lime. This treatment led to the precipitation vation implies that the rate of crystal growth would be
and removal of most of the divalent cations in solution, reasonably insensitive to differences in mixing with
including magnesium. Consequently, the solubility of gypsum process scale-up.
in their treated wastewater was much more favourable for 2. The concentration driven kinetics of 2nd order systems
sulfate removal than that of the present study and the study of lead to much less efficient crystallisation (much longer
Kabdasli et al. (1995). Under non-ideal equilibrium conditions residence times to produce the same amount of crystal
the cost of excess calcium as lime to drive the removal of with the same crystal surface area initially available) in
sulfate contributes considerably to the operating costs of continuous crystallisers than in batch crystallisers.
a crystallisation process, but so does lime and/or sodium
hydroxide addition to raise the pH of the wastewater above 12. For applications in cold climates, it is important to note
Carbon dioxide produced by any subsequent anaerobic that the crystal growth kinetics of gypsum significantly
process can be used to neutralize the high pH of the treated decreases with decreasing temperatures (Liu and Nancollas,
wastewater. 1970). In such cases the crystallisation process would either
require heating or would need to operate at higher solids
4.3. Crystallisation kinetics concentrations (to provide additional crystal surface for mass
deposition) or longer retention times.
4.3.1. Inhibition of nucleation
Although, the wastewater used in this study was supersatu- 4.3.3. Crystal re-use
rated with respect to gypsum (see Section 4 above), it did not Rates of crystallisation measured for experiments where the
readily self-nucleate gypsum crystals and seed crystals had to wastewater was seeded with crystal product, did not differ
be added to induce crystallisation. Pure aqueous solutions of significantly from that measured when fresh synthetic
gypsum at corresponding levels of supersaturation self- gypsum seed crystals were used (see Section 3.3). This
nucleate and crystallise within a few seconds (Klepetsanis observation indicated that any inhibition of crystal growth by
et al., 1999). This effect, commonly termed metastability impurities in the wastewater (Hamdona and Al Hadad, 2007)
(crystal growth, but no nucleation), was also noted by Smit was not amplified with re-use of the crystals in subsequent
and Sibilski (2003) for gypsum crystallisation from acidic mine crystallisations. It suggests that crystal product may be reused
waters. Nucleation is likely inhibited by impurities in the as seed for subsequent removal of sulfate without detrimental
wastewater. As noted in the introduction of this paper, Kab- effects on the process, a much more economical option than
dasli et al. (1995) recommended a 1 week crystallisation time continual use of manufactured seed.
to remove significant amounts of sulfate from textile industry
wastewater, whereas the crystallisation experiments of Gel- 4.4. Crystallisation process
denhuys et al. (2003) on acid mine waters typically attained
equilibrium within 2–5 h. It is thought that inhibition of The present findings indicated that gypsum crystallisation
nucleation by impurities in the wastewater of Kabdasli et al. occurs at practical rates in wastewater and consequently
(1995) was responsible for their long crystallisation time, would be a feasible method for the removal of sulfate. A basic
because unlike the experiments of Geldenhuys et al. (2003), design for the crystallisation process would consist of a crys-
Kabdasli et al. (1995) did not seed their solutions with gypsum talliser, an optional chemical dosing system, and a method of
crystals. Further, based on our results, this inhibition is likely crystal solids separation. Given the favourable properties of
due to inorganic contaminants, since a decreased level of crystals such as those evaluated in this study, gravity sepa-
organic acids did not enhance nucleation rates. ration by settling in the crystalliser vessel and intermittent
draw-off of crystal product may be appropriate and would
4.3.2. Crystal growth kinetics probably be most economical. To enhance the effectiveness of
In the experiments of this study, seeded wastewater samples crystal separation from the effluent the option of a hydro-
typically reached equilibrium concentrations of calcium and cyclone may be considered with little added cost.
sulfate within 2–5 h (see Section 3.3), which is favourable for The selection of either batch or continuous modes of
sulfate removal. The rate of crystal growth was observed to be crystalliser operation depends on the scale of treatment.
2nd order with respect to the supersaturation of the solution Batch crystallisers tend to have higher operating and capital
 water research 43 (2009) 762–772 771

costs and are typically only economical for small scale treat- crystallisation. This effect adds a degree of control over the
ment. If continuous gypsum crystallisation is selected, the size distribution of the crystal product from a batch crys-
2nd order dependence of crystal growth on supersaturation tallisation process.
must be considered, because as a consequence, continuous  Impurities in the wastewater do not substantially inhibit the
stirred tank crystallisers would be much less efficient than rates of crystal growth and seeded crystallisation of gypsum
a stirred batch system. Options include; reaches equilibrium within practical batch times of 2–5 h.
 Crystallisation in the wastewater which is seeded with
1. increasing the volume of a continuous crystalliser to crystal product from an earlier crystallising batch, exhibits
operate at longer hydraulic retention times (capital similar crystal growth rates to crystallisation in the waste-
intensive), water seeded with fresh synthetic gypsum seed crystals. At
2. operating a continuous crystalliser at high crystal concen- an industrial scale, it is more economical to operate with
trations to increase available crystal surface area for mass crystal recycle than with manufactured seed only.
deposition and increasing the overall rate of mass crystal-  Impurities in the wastewater substantially increase the
lisation, or solubility of gypsum (3-fold compared to pure solution) and
3. operating the continuous crystalliser in a plug-flow mode a substantial excess of calcium may be necessary to attain
for which the influence of higher-order kinetics is practi- the desired level of sulfate removal (for example 10 gCa L1
cally the same as in a batch crystalliser. to achieve 0.25 gS L1).
 Complex sulfate minerals with lower solubilities than
Recycling of crystal product may be employed with gypsum (such as jarosite and ettringite) do not crystallise at
continuous systems to increase available crystal surface area. mild pH conditions (5–9) favourable for feed to a subsequent
The loss of crystal surface area with discharge of coarse anaerobic process.
crystals from the system is replaced by seed crystal charge,
unless nucleation (shown to be inhibited in the present
wastewater) provides the additional new crystal surface area
for mass deposition.
Acknowledgements
The overall economics of a gypsum crystallisation process
would be heavily driven by requirements of lime to achieve
Veolia Environmental Services (Australia) and the Environ-
the desired reduction in sulfate (see above). The positive
mental Biotechnology CRC (EBCRC) Australia funded this
offsets for these costs may include;
study. Mr. Justin Houghton and Mr. Chris Alexander from
Veolia Environmental Services, and Dr. Brenton Hamdorf
1. methane gain from a subsequent anaerobic process,
from EBCRC are gratefully acknowledged for their input into
2. cost savings from reduced loading on existing sulfide
this project. Dr. Rachel Smith (Chemical Engineering, UQ)
treatment systems,
assisted in performing the crystallisation experiments and Mr.
3. lower overall sulfide risk and
Paritam Dutta (AWMC, UQ), Mr. David Page (Chemical Engi-
4. value-adding to lime, by conversion to gypsum, and so
neering, UQ), Mr. David Appleton (Land and Food Sciences,
resource recovery from the wastewater.
UQ), Dr. Beatrice Keller-Lehmann (AWMC, UQ), Ms. Kar Man
Leung (AWMC, UQ) and Ms. Jianguang Li (AWMC, UQ) per-
The extent to which these four factors affect the overall
formed some of the analyses.
economical viability of the process would be site and appli-
cation specific. For instance, the interest in methane may be
references
low in a particular application, while concern over sulfide risk
mitigation may be high, or vice versa.

Alvarez-Ayuso, E., Nugteren, H.W., 2005. Synthesis of ettringite:

a way to deal with the acid wastewaters of aluminium
5. Conclusions anodising industry. Water Res. 39, 65–72.
Batstone, D.J., Pind, P.F., Angelidaki, I., 2003. Kinetics of
Gypsum crystallisation is an effective means for bulk removal thermophilic, anaerobic oxidation of straight and branched
of sulfate from highly contaminated wastewaters. Co-precip- chain butyrate and valerate. Biotechnol. Bioeng. 84, 195–204.
itation to other compounds such as jarosite and ettringite does Cecen, F., Gursoy, G., 2000. Characterization of landfill leachate
and studies on heavy metal removal. J. Environ. Monit. 2,
not occur under reasonable process conditions. Precipitation
436–442.
processes therefore cannot economically reduce sulfate to very Christoe, J.R., 1976. Removal of sulfate from industrial
low levels, and post anaerobic sulfide removal is still required wastewaters. J. Water Pollut. Control Fed. 48, 2804–2808.
to produce high-quality biogas. However, sulfate precipitation Christoffersen, M.R., Christoffersen, J., Weijnen, M.P.C., van
can substantially lower the cost of sulfide removal, and will Rosmalen, G.M., 1982. Crystal growth of calcium sulfate
enhance methane production by reduced sulfate reduction. dihydrate at low supersaturation. J. Cryst. Growth 58, 585–595.
Elgersma, F., Witkamp, G.J., van Rosmalen, G.M., 1993.
This conclusion is based on the following experimental
Incorporation of zinc in continuous jarosite precipitation.
outcomes:
Hydrometallurgy 33, 313–339.
Farmer, R.W., Beckman, J.R., 1984. Particle size distribution
 Impurities in the wastewater inhibit nucleation, such that analysis of blended solids by a modified Andreasen pipet
the addition of seed crystals is required to induce method. Ind. Eng. Chem. Process Des. Dev. 23, 341–343.
772 water research 43 (2009) 762–772

Gabrisova, A., Havlica, J., Sahu, S., 1991. Stability of calcium Parkhurst, D.L., Appelo, C.A.J., 1999. User’s Guide to PHREEQC
sulphoaluminate hydrates in water solutions with various pH (Version 2) – a Computer Program for Speciation, Batch-
values. Cem. Concr. Res. 21, 1023–1027. reaction, One-dimensional Transport, and Inverse
Geldenhuys, A.J., Maree, J.P., de Beer, M., Hlabela, P., 2003. An Geochemical Calculations: U.S. Geological Survey Water-
integrated limestone/lime process for partial sulfate removal. Resources Investigations Report 99-4259.
J. S. Afr. Inst. Min. Metall. 103, 345–353. Reichert, P., 1994. AQUASIM – a tool for simulation and data
Hamdona, S.K., Al Hadad, U.A., 2007. Crystallization of calcium analysis of aquatic systems. Water Sci. Technol. 30 (2),
sulfate dihydrate in the presence of some metal ions. J. Cryst. 21–30.
Growth 299, 146–151. Seidell, A., 1940. Solubilities of Inorganic and Metal Organic
Hina, A., Nancollas, G.H., 2000. Precipitation and dissolution of Compounds. van Nostrand, NY.
alkaline earth sulfates: kinetics and surface energy. In: Smit, J., Sibilski, U.E., 2003. Pilot plant study to treat typical
Alpers, C.N., Jambor, J.L., Nordstrom, D.K. (Eds.), Reviews in gold mine minewater using the savmin process. Water in
Mineralogy & Geochemistry, 40, The Mineralogical Society of Mining, October 13–15, Brisbane, Australia.
America, Washington, DC, pp. 277–301. Tait, S., 2007. PhD thesis. The University of Queensland, Brisbane,
Kabdasli, I., Tunay, O., Orhon, D., 1995. Sulfate removal from indigo Australia.
dyeing textile wastewaters. Water Sci. Technol. 32, 21–27. van Rosmalen, G.M., Daudey, P.J., Marchee, W.G.J, 1981. An
Klepetsanis, P.G, Dalas, E., Koutsoukos, P.G., 1999. Role of analysis of growth experiments of gypsum crystals in
temperature in the spontaneous precipitation of calcium suspension. J. Cryst. Growth 52, 801–811.
sulfate dihydrate. Langmuir 15, 1534–1540. Warren, C.J., Reardon, E.J., 1994. The solubility of ettringite at
Liu, S., Nancollas, G.H., 1970. The kinetics of crystal growth of 25  C. Cem. Concr. Res. 24, 1515–1524.
calcium sulfate dihydrate. J. Cryst. Growth 6, 281–289. Wang, C., Ma, S., Lu, A., Zhou, J., 2006. Experimental study on
Mullin, J.W., 2001. Crystallization. Butterworth-Heinemann, formation conditions of ammoniojarosite and its
Oxford. environmental significance. Acta Geol. Sin. 80, 296–301.