DESALINATION

ELSEVIER Desalination 143 (2002) 255-267

www.elsevier.com/locate/desal

Silica and metals removal by pretreatment to prevent

fouling of reverse osmosis membranes

R. Sheikholeslami*, J. Bright
School of Chemical Engineering and Industrial Chemistry The University of New South Wales,
Sydney 2052, Australia
Tel: +61-2-9385-4343, Fax: +61-2-9385-5966, email: [email protected]

Received 28 January 2002; accepted 5 March 2002

Abstract

Reverse osmosis is being increasingly used for desalination of seawater, brackish water and as well the
marginal and polluted wastewaters. The operation of waters bearing substantial amount of silica is only feasible
if there is a chemical pretreatment in process to prevent excessive and unbearable fouling by this refractory
deposit. In this study, model solutions representing tailings wastewater which usually has high silica content was
used in standard jar tests using the Boltac Coagulation and Flocculation simulator to determine the effectiveness
of chemical pretreatment (by precipitation and coagulation) for removal of silica and other species such as
magnesium, calcium, iron and manganese which affect silica fouling. Two precipitants were tested: lime and
soda ash, and caustic soda. Precipitant aids (alum and ferric chloride) were also examined in combination with
optimum precipitant doses to observe their differential effects. The optimum precipitant dose of caustic soda was
200 mg/L. Ferric chloride and alum were tested with this optimum caustic soda dose, but did not improve
removal significantly, if at all. The optimum precipitant dose of lime and soda ash was 150 mg/L lime and 450
mg/L soda ash. Ferric chloride and alum were tested with this optimum lime and soda dose, but again did not
improve removal enough to warrant their use. Neither of the methods were as effective in removal of magnesium
and calcium as in removal of other components; lime and soda ash treatment as expected actually increased the
calcium content of solutions.

Keywords: Silica fouling; RO; RO fouling; RO pretreatment; Silica removal; Metal removal; RO desalination

1. Aims and significance increasingly important aspect of process

operation, particularly with the advent of more
Wastewater treatment is becoming an stringent environmental regulations.
Environmental pressures, along with other
factors such as fresh water cost and availability,
*Corresponding author

001 l-9164/02/$- See front matter 0 2002 Elsevier Science B.V. All rights reserved
PII: SO0 I I-9 164(02)00264-3
256 R. Sheikholeslami and J. Bright/Desalination 143 (2002) 255-267

wastewater disposal costs and climate applications, with desalination being the largest
influences have caused many companies to market. Despite all the advances in design of
consider recycling their wastewater. Selection RO processes, one major problem persists and
of the most appropriate water recycling process that is membrane fouling which is defined as a
can be quite difficult, with a wide variety of condition under which membrane undergoes
traditional processes available and new, less coating by some element or elements in the
established technology evolving. Reverse stream in such a way that its output or flux is
osmosis (RO) is becoming an increasingly reduced with time. Therefore the micro-
popular choice for the separation of water from environment near the membrane surface is not
various types of solutions, and is often applied at the steady state condition. Membrane fouling
as a tertiary step in treatment of saline waters reduces the quality of product water as well as
such as mineral tailings. However, like all its quantity in addition to often damaging the
membrane processes, RO membranes membrane itself. There are several mechanisms
experience substantial problems with fouling of fouling such as particulate deposition,
especially with marginal waters. Significant precipitation, accumulation and growth of
fouling acts to reduce the effectiveness of the biological matter, deposition of the reaction
membranes and makes the process impractical products from the stream constituents, chemical
and uneconomical. For this reason, pretreatment or physical reaction of stream constituents with
is often essential when wastewater contains a the membrane surface, and any combination of
large quantity of foulants. Of particular interest the above mechanisms.
is the possibility of silica fouling of RO units, Pretreatment of feedwater is an effective and
which is a major problem in the desalination of necessary means of fouling mitigation
brackish waters. Silica fouling is also affected specifically with marginal waters such as
by the presence of other species such as industrial wastewaters. Obviously, the most
calcium, magnesium, iron, and manganese. A effective pretreatment scheme must be designed
common pretreatment method used for the based on chemical and physical analyses of the
removal of metals and silica is precipitation, feed as well as the desired feed water quality.
often using a precipitation aid (or coagulant) Generally, in practice Silt Density Index (SDI)
and as such several chemicals in various or a variation of it is used to assess the fouling
combinations were tested in this study to propensity of water. However, though high SD1
determine their effectiveness and optimum values insure membrane fouling, low SD1
dosage on model solutions. The effect of values should never be considered a measure of
precipitant aids (coagulants) was assessed. fouling-free operation as SD1 is a measure of
colloidal (particulate, organic matter, and
bacteria) fouling but as mentioned above, there
2. Background are other types of fouling such as precipitation
fouling that its measure cannot be assessed by
Wastewater treatment technologies can be SD1 values. In cases where feedwater is
divided into three categories: physical, chemical relatively clean (e.g. well water), a simple
and biological. Often a combination of these addition of antiscalant or slight modification of
processes is used to achieve the desired water stream pH may be enough to mitigate fouling.
quality in the most economical way [l]. However, for streams with high fouling
Reverse osmosis is a membrane filtration potential, possibly from several feed
physical separation technology that is becoming constituents, more extensive pretreatment is
an increasingly popular for a number of necessary.
 R. Sheikholeslami and J. Bright/Desalination 143 (2002) 255-267 257

Table I avenues for pretreatment of these streams. The

Typical pretreatment methods and chemicals main methods are chemical treatment
(precipitation, coagulation, flocculation)
Method Chemicals followed by deep bed media filtration or
membrane filtration. Typical chemicals used for
Precipitation Lime (Ca(OH)J and soda ash
these purposes are listed in Table 1.
(NaXO3)
Sodium hydroxide (NaOH) The ultimate aim of pretreatment is to
Coagulation Alum (AI,(SO,)3), pH-5.5 to completely remove feedwater constituents that
8 (6 to 7 is preferable) cause fouling; however, this is rarely
Iron salts: ferric chloride and achievable. So, generally pretreatment schemes
ferric sulphate, pH-4 to 1 I are designed to produce a maximum-tolerable
(6 to 7 is preferable)
RO feed concentration of each particular
Flocculation Polyelectrolytes (cationic,
anionic, and non-ionic)
foulant.
Silica has been described as the most
problematic foulant in RO desalination units [2]
and is a predominant cause of RO fouling in
Table 2
desalination of industrial wastewaters such as
Silica concentrations in various RO feedwaters [4,5]
tailings, pulp and paper, and brackish water and
Water source Si02 (mg/l) its reduction is considered an important
pretreatment step to prevent silica scale
Natural water l-30 formation within the RO modules. Typical
Well water 20-100 silica concentration in various RO feed waters
Seawater Negligible is given in Table 2.
Brackish water up to 1000 Silica (SiOz) exists in many crystalline and
Tailings 20-60
amorphous forms. Crystalline silica has very
low solubility in water, of the order of 6 ppm
SiOz. Amorphous silica has a much higher
Table 3 solubility of 100-140 ppm SiOz [2]. When
Size range of particles usually found in water supplies silica is dissolved in water, it forms monosilicic
acid (Si(OH)d) which will remain in the
Type Particle size Stability
monomeric state as long as its concentration
(nm)
remains less than about 2 mM [2].
True solution <I Completely However, at higher concentrations
stable monosilicic acid dimerizes and polymerizes to
Colloidal dispersion l-500 Can remain form (larger) polysilicic acids, the larger of
dispersed for which are of a colloidal size. Colloidal silica is
years also sometimes referred to as silica sol. Table 3
Suspension >500 Settles rapidly shows the size range definitions of particles in
water [3].
Silica fouling of membranes can occur by
one of three mechanisms: monomeric
Examples of very “dirty” streams with high deposition on the membrane surface, colloidal
fouling potential include brackish water, landfill deposition whereby colloids form in the
leachate water, dairy waste, pulp and paper solution and then accumulate at the membrane
waste, and tailings water. There are several surface, and biogenic amorphous silica by
258 R. Sheikholeslami and J. Bright/Desalination 143 (2002) 255-267

living organisms; these are discussed by silicate {e.g., MgzSiOd (foserite)) may be
Sheikholeslami and Tan elsewhere [2]. formed, but the most likely mechanism is
The problem of silica fouling is mostly simple adsorption onto the surface of Mg(OH)z.
exacerbated in presence of even minute The A13’ ion precipitates colloidal silica, but
amounts of other cations, anions, and colloidal only in the pH range 4 to 5. It requires only one
matter. As such it is important to select a model part A13’ to precipitate 40 parts SiOz. The A13”
system that consists other troublesome ion precipitates soluble silica in the pH range
constituents and assess silica removal in their 5-11, but at least 4 parts Al are required for 1
presence. Tailings water is a good part of SiO2.
representative of such streams and as such it Lime and magnesia (which can be made
was chosen as the model system for the from magnesite) are commonly used for
purposes of this study. removing silica from water. They flocculate
Silica, both soluble and colloidal, can be colloidal silica along with other precipitated or
removed from water to varying levels in neutral suspended matter and combine with and adsorb
or slightly alkaline solution by coprecipitation soluble silica. Lime is often used as the first
with soluble metals which form insoluble step to remove silica. Dolomitic lime is also
hydroxides in situ, or by adsorption upon effective, as it contains 32% MgO. It may be
freshly formed insoluble hydroxides added to used in a cold process (70°F) however under hot
water. Although alumina should be the best conditions of over 120°F only half as much
adsorbent, magnesium hydroxide is the dolomite is required. Lime or magnesia
commonest reagent, requiring a relatively high treatment can reduce silica to as low as 3 ppm
pH, recirculation of sludge and a hot process. In under rather simple process conditions.
laboratory studies, silica was adsorbed and Coagulation aids silica removal. Silica is
precipitated by hydroxides of aluminium, iron, “hydrophilic”, although under some conditions
manganese and magnesium [6]. With a 15 ppm where salts cause coagulation it has become
AI solution, the silica concentration classed as somewhat hydrophobic. When
dropped from an initial amount of 3 ppm to a organic flocculating agents are adsorbed onto
residual value of 0.8 ppm; whereas with 30 ppm the surface, the surface definitely becomes
aluminium hydroxide, no measurable silica hydrophobic. Coagulation has been considered
remained. Dissolved silica may also be to be the result of van der Waals attraction
adsorbed onto a surface such as ferric which draws two particles together at the
hydroxide, rather than precipitation as a discrete moment of collision, unless opposed by a
compound. Ferric hydroxide is precipitated hydration barrier or by the electrostatic
from ferric sulfate by addition of sodium or repulsion forces between the similarly charged
calcium hydroxide for silica removal. In this particles, or both. There are therefore two
case, a pH of 9.0 or greater is required [7]. factors that retard coagulation of silica; one
Silica is adsorbed onto the surface of being the “hydration” of the surface of the
insoluble magnesium compounds, such as particles by a layer of water molecules
Mg(OH)I or MgCOx. If a natural water contains hydrogen-bonded to the SiOH groups and the
a considerable portion of magnesium hardness, second being the negative ionic charge on the
then silica will be removed to a certain extent particles above about pH 3.5 and the
with concurrent precipitation with Mg(OH)? surrounding cloud of positive counter ions such
[71- as Na+, forming the “double layer”.
However, the chemistry of silica removal by The mechanism of the coagulation of
magnesium hydroxide is unclear. Magnesium colloidal silica by electrolytes is still not well
 R. Sheikholeslami and J. Bright/Desalination 143 (2002) 2.55-267 259

understood. However, the effectiveness of Table 4

various types of electrolytes has been examined Solubility product values for some metal hydroxides at
by experiment. It has been found that below pH 25°C
1 1, and above critical concentrations of salts of
Metal hydroxide log Ks
all the alkali metals, coagulation occurs because
the particles become initially linked together by -5.19
CaWU2
acid-base bonds, rather than Si-0-Si bonds WOW2 -15.1
since the fresh coagulate is peptised by WOW3 -38.8
lowering the pH. However, above pH 11, where Mg(OWz -11.15
the surface is covered by basic groups, sodium MnW% -12.8
and lithium salts continue to bring about
coagulation but potassium, rubidium, and
cesium slats do not.
At pH 7.5, tetramethylammonium ion is a It is important to assess the effectiveness of
powerful coagulant with critical concentration metals removal by pre-treatment as well
of coagulant (c.c.c.) of 0.01 M, as compared because any amount of metals left in the stream
with ammonium and sodium, which have C.C.C. will exacerbate fouling. Heavy metals [S] are
of 1.0 M and 2.2 M, respectively. It has been precipitated as hydroxides through the addition
generally observed that in a mixture of of lime or caustic to a pH of minimum
polyvalent metal salt and polysilicic acid or solubility. However, several heavy metals are
colloidal silica at low pH, coprecipitation amphoteric and exhibit a point of minimum
occurs as the pH is raised to just below the pH solubility. The pH of minimum solubility varies
at which the metal hydroxide is precipitated with the metal in question. Often, pretreatment
from the metal salt solution when no silica is is necessary to remove substances that will
present. Examples of such salts include Fe3+, interfere with precipitation, such as cyanide and
Cr3+, and Ca2’. Silica particles may also be ammonia which form complexes with many
coagulated by hydrolyzed aluminium ions. metals. The logs of the solubility products (KS)
Flocculation of colloidal silica occurs with for the metal hydroxides of the metals to be
the addition of a polymeric or particulate removed in this study are given in Table 4 and
colloid; the micelle or particle of which is may be used to compute the optimum pH for
adsorbed simultaneously on the surfaces of two removal. Apart from lime precipitation, heavy
silica particles, thus linking them together. Such metals may also be precipitated as the sulfide
adsorption occurs either because the agent is and in some cases as the carbonate, such as
cationic and thus attracted to the negatively lead; however, this is not discussed here, as it is
charged silica surface, or because it contains not the subject of this study.
electron donor groups such as the oxygen of a Although the effect of cations on silica
polyether, and is held to the silica surface by solubility and polymerisation is not well
hydrogen bonding. The latter only occurs at low understood, both magnesium and calcium are
pH. The amount of flocculant added is critical, believed to adversely affect silica fouling on
since a certain amount is required to give RO membranes [4]. Calcium and magnesium
complete flocculation, however if excess salts increase fouling as they facilitate colloidal
flocculant is added, the flocculant may act as a silica formation and hence result in particulate
stabilising agent. To prevent redispersion in the deposition in addition to direct precipitation of
presence of excess flocculant, the colloid monomeric silica on the membrane; that is the
should first be coagulated. reason they have been selected as part of the
 R. Sheikholeslami and J. Bright/Desalination 143 (2002) 2.5.5-267

constituents of the model solution as their components whose concentrations were too low
removal is important for reducing the fouling to enable direct accurate measurement of the
potential. Manganese and iron have also been salt with the given analytical balance. Sodium
seen to adversely affect silica fouling even metasilicate solution was prepared (pH-1 l), its
when in minute quantities and as such their pH was reduced by hydrochloric acid to 7 while
removal was assessed in this study. the stock solutions of metals were added in
order to prevent the possibility of silicate
formation during solution preparation. The pH
3. Experimental materials and procedures was monitored to insure that it was between 5
to 7 during the solution preparation. Once the
A Boltac Coagulation and Flocculation constituents of the model solutions were mixed
Simulator (or “Jar Star”) was used to perform the pH was brought to 4.9 and the solution
the Jar Tests. This equipment simulates the type volume to 5L.
of mixing and settling conditions found in a The “Jar Star” was programmed with the
clarification plant, and allows the accurate desired experimental procedure. The
control of stirring speed. The simulated model precipitant/precipitant aid solutions were
solutions were prepared using distilled micro- prepared by dissolving required amount of
filtered water with analytical grade reagents. chemicals. A 500 mL measuring cylinder was
Sodium metasilicate was used for preparation of used to transfer 500 ml of model solution into
silica stock solutions. The salts used for each of the four 1L jars. Desired quantities of
preparation of the metals stock solutions were precipitant were pipetted into each jar. The jars
chosen based on their relatively high solubilities were then positioned in the apparatus and the
and their anion constituents. The iron and program started. Each jar was rapid mixed at
manganese salts were chosen such that the 200.0 rpm for 3 minutes. Precipitant aids (i.e.
metal ion is in the correct oxidation state to alum or ferric chloride) were added towards the
allow for precipitation to occur. If manganous end of this period for those runs in which their
and ferrous salts had been chosen, pretreatment effectiveness was being assessed. Stirring speed
by chemical oxidation or aeration would have was reduced to 40.0 rpm for 12 minutes, to
been required. allow flocculation to occur. Stirring was
It should be noted that some of the stopped and the sample was allowed to settle
components included in the model solution (i.e. for 30 minutes. During this time, the pH of each
cobalt, nickel and barium) were not analysed jar was measured. Samples of supernatant from
for their removal as they were present in small each beaker were taken. Varian Vista AX CCD
amounts and were not necessary to remove in Simultaneous inductively coupled plasma (ICP)
pretreatment. They were included as technique was used to determine the
background components so that the model concentration of iron, magnesium, manganese,
solution could more accurately represent and silica in supernatant samples. Millipore
tailings water. Corp. Millex-GN Nylon 0.2 _m microfilters
The model solution was prepared in five- with 20 mL syringes, and also Millipore
litre batches from their respective salts. The ultrafilters from regenerated cellulose were
components were measured either directly by used for those runs during which the
the analytical balance, or a more concentrated supernatant was filtered to determine the size
solution was prepared and the desired quantity range of colloidal particles formed and to
of it measured by a pipette. The latter devise a proper separation technique for their
preparation method was required for the removal prior to entering RO.
 R. Sheikholeslami and J. Bright/Desalination 143 (2002) 2.55-267 261

Table 5 dose; the removal at this dose was 90.6%,

The composition of the model solution and salts used which corresponds to a final silica
for its preparation concentration of 3.5 mg/L, down from the
initial concentration of 37.3 mg/L. The
Salt used for Desired Initial concentr.
preparation component of component mechanism of silica removal could be
(ppm)
coprecipitation with the insoluble metal
hydroxides that were formed when the caustic
Fe2(S04)3.9H20 Fe 0.57-0.71 soda was added; specifically iron, manganese
MnSO+H*O Mn 5.8-6.7 and magnesium hydroxides have been known to
MgS0+7H20 Mg 530-560 achieve silica removal in this way [6]. Silica
CaC12 (anhydrous) Ca 51-59 could also be removed by precipitation as
NiS04.6Hz0 Ni 0.3
magnesium silicate (MgzSiOJ) or calcium
CoCl*.6H*O co 0.1
BaC12.2H20 Ba 0.4 silicate (CazSiOa). Change in millimoles of
Na203Si.9H20 Si 17-18 silica removed with increasing caustic soda
SiOl 36-39 doses was plotted with change in magnesium
NaHC03 co3 15 and calcium removal in Fig. 2 to give an
indication of whether calcium and magnesium
removal changed proportionately with silica
removal, i.e. to indicate if removal was via
Table 5 lists the composition of the model formation of calcium and/or magnesium
solution. The dosages of precipitants were silicate. It is difficult to draw any definitive
50-400 ppm, 50-200 ppm, and 150-800 ppm, conclusions from Fig. 2, however it does appear
respectively for NaOH, lime, and soda ash. A that silica removal follows calcium removal
dosage range of lo-25 was used for precipitant quite closely for doses of 250 mg/L caustic and
aids (Alum and ferric chloride) with the above. This could be explained by the pH,
optimum dosage (NaOH = 200 ppm; lime = 50 which is 10.1 (measured) or higher for these
ppm, soda ash = 450 ppm) of the precipitants. doses and pH determines the silica species
present in solution. The silicate ion becomes
more predominant as pH increases, up to a
4. Results and discussion maximum at about pH 11, hence removal by
silicate formation would predominantly be the
4.1. Removal by caustic soda with and without removal mechanism at this high pH.
precipitant aids
4.1.2. Iron and manganese removal
Eight jar tests were carried out with Removal of iron was about 96% and did not
increasing amounts of NaOH, starting with a 50 change significantly with increasing doses of
mg/L dose and increasing this by 50 mg/L caustic soda, with the final iron concentration
increments up to 400 mg/L to determine the being of the order of 0.02 mg/L. This result is to
optimum dose rate for precipitation with be expected given that the solubility product of
caustic. The results are shown in Fig. 1. ferric hydroxide is extremely low (log K,, =
-38). Manganese removal increased from
4.1.1. Silica removal 76.5% to 97.3% when caustic dose was
Silica removal increased dramatically (by increased from 50 mg/L to 100 mg/L, and
more than 20%) with each increase in caustic removal approached a plateau after this point
soda dose until it steadied at 200 mg/L caustic where the pH was about 9.6.
262 R. Sheikholeslami and J. Bright/Desalination 143 (2002) 255-267

100 II.0
than that of iron hydroxide and manganese
; hydroxide, which were preferentially removed.
t 80 10.6
x The removal of magnesium constantly
; 60 10.2
increased, although not at a great rate, while the
% e,
a
t
40 9.8 pH of the solution was seen to increase from 9.2
E 20 9.4 to 10.5. Calcium removal is very low
7
e
0 9.0
(7.7-14.5%), and does not improve
0 50 100 150 200 250 300 350 400 significantly with increased caustic doses. This
NaOH dose(mg/L) is most likely due to the fact that calcium
hydroxide is the most soluble of all the
Fig. 1. Removal of Fe (0, initial = 0.58 mg/L), Mg (W,
hydroxides formed, with K,, = -5.19.
initial = 552 mg/L), Mn (+, initial = 6.1 mg/L), Si (A,
initial = 17.4 mg/L), Ca (x, initial = 56 mg/L) and In order to ascertain if any buffering is
solution pH (0) with varied caustic soda doses after 30 occurring and as well possibly assess the
min of settling. mechanism of removal of silica to determine
whether it is predominantly being removed as
2.0 silicate or by physical adsorption on initially
1 I
formed hydroxide, the measured pH was
compared with the calculated pH (based on
6B 1.0
complete dissociation of added sodium
2 0.5 hydroxide). The final concentration of
8
t 0.0 hydroxide in solution was first calculated
E
4 -05 J I assuming that hydroxide is being consumed in
0 50 100 150 200 250 300 350 400 production of Fe(OH)s, Mg(OH)2, Mn(OH)2
NaOH dose (mg/L.)
and Ca(OH)l and that silica is removed by
Fig. 2. Change in millimoles of components removed adsorption onto precipitates only. However,
(Si, x, initial = 0.62 mmol; Mg, n initial = 22.7 mmol; using these assumptions, the calculated final
Ca, o, initial = 1.4 mmol) with increasing doses of hydroxide concentration was negative for all
caustic soda. but the lowest dose of NaOH. Since magnesium
removal increased most dramatically from the
lowest dose of NaOH to the next dose, the final
This is in agreement with literature stating concentration of hydroxide was recalculated
that a pH of about 9.4 is required to achieve assuming magnesium was not removed as
significant manganese removal by precipitation. magnesium hydroxide but as magnesium
silicate. The final hydroxide concentration was
4.1.3. Hardness removal then no longer negative and was in the range of
The initial magnesium concentration was the measured values. This suggests that
552 mg/L, and was the most concentrated magnesium silicate is the form predominantly
component in the model solution, followed by precipitating.
silica at 36.4 mg/L. This is one reason why It was concluded from these results that the
magnesium removal was not significant, as a optimum dose of caustic soda was 200 mg/L,
very large dose of caustic would have been because removal of components did not
required to react with all the magnesium. In increase significantly when caustic soda dose
addition, considering the solubility product of was further increased. Also, it is desirable to
the components monitored (Table 4), the minimize the precipitant dose chosen, because
solubility of magnesium hydroxide is greater higher doses increase chemical costs.
 R. Sheikholeslami and J. Bright/Desalination 143 (2002) 2.55-267 263

93.5%, but for 25 mg/L alum dose, silica

removal drops to 81.7%.
The effect of micro and ultrafiltration of
supernatant was investigated. Iron removal was
marginally improved (less than 1%) by
e 20 94 filtration; however, calcium removal was
S affected quite substantially, particularly by
0 90
0 5 IO IS 20 25
ultrafiltration. Generally, microfiltration did not
Alum dose (mg/L) improve calcium removal, however, ultra-
Fig. 3. Effect of alum addition to optimum caustic soda
filtration improved calcium removal from
dose of 200 mg/L (Fe, o, initial = 0.74 mg/L; Mg, m, 1.48% to 18.06% for zero alum dose.
initial = 523 mg/L; Mn, +, initial = 6.3 mg/L; Si, A, Therefore, it is clear that calcium was present as
initial = 17.6 mg/L; Ca, X, initial = 58.9 mg/L; and a suspension in the supernatant that was smaller
solution pH (0). than micron size but larger than ultrafiltration
pore size.
Manganese removal, like iron removal, was
100 II 0
;
not significantly improved by filtration. Silica
2
z
80 106
removal was slightly improved for zero alum
E 60 102 dose (from 91.5 to 93.6%), however, silica
I
t
= 40 98
a
removal was not improved by filtration for the
t other jar tests. This seems to indicate that very
E 20 ) 9.4
little, if any, colloidal silica existed in the
Sos90
supernatant. However, it is possible that the size
0 5 IO 15 20 25
FeCI,dose(mg/L)
range of colloidal silica present was less than
0.2 micron and passed through the
Fig. 4. Addition of ferric chloride to optimum caustic
microfiltration, which was accurately measured
soda dose of 200 mg/L (Fe, o, initial = 0.74 mg/L; Mg,
n , initial = 523 mg/L; Mn, +, initial = 6.3 mg/L; Si, A, but was not picked up by the ultrafiltration
initial = 17.8 mg/L; Ca, x, initial = 58.9 mg/L; and pH since the experimental technique used for
(0). ultrafiltration was not very accurate. The
filtration apparatus was rinsed with distilled
water and a sample of supernatant was filtered
4.1.4. Use ofprecipitant aids through, the flask was swirled around, and the
The effect of addition of precipitant aid to filtrate disposed of before another volume of
the optimum caustic soda dose of 200 mg/L was supernatant was filtered and analysed. It would
investigated. The precipitant aids tested were have been preferable to rinse the apparatus with
alum and ferric chloride. Alum doses of 10 to distilled water and allow it to dry and use a
25 mg/L were tested and the results obtained clean membrane for each sample. This is
are shown in Fig. 3. The removal of most of the particularly important since the concentrations
components is relatively unchanged with the being analysed are so small. This will be done
addition of alum, which indicates that there is for future experiments. It would appear from
not a great deal of colloidal matter present this set of results that filtration (by
which alum is effective at removing. Silica microfiltration or ultrafiltration) of sample does
removal is seen to decrease with alum addition. not significantly improve removal, so the
This could be due to the lower pH with alum precipitates settle sufficiently in the given
addition. Without alum, silica removal is settling time (30 minutes) and there is a
264 R. Sheikholeslami and J. Bright/Desalination 143 (2002) 255-267

negligible amount of colloidal material present. really affected when the sample was filtered.
However, for shorter settling times, filtration With 20 mg/L of ferric chloride, silica removal
may be essential to prevent colloidal fouling in was 93.38%. These results would indicate
RO. perhaps addition of ferric chloride as a
Ferric chloride doses tested were 10 to 25 precipitant aid or filtration would be necessary;
mg/L, and the results obtained are shown in Fig. filtration following ferric chloride addition for
4. These results show that the addition of ferric silica is not particularly necessary.
chloride does not significantly improve removal Manganese removal is improved with
for any of the components, except for calcium. microfiltration of supernatant for all doses.
Calcium removal remained almost the same for However, this improvement is small and
the two lower doses of ferric chloride, but unnecessary, as manganese removal is
improved for the higher doses of 20 and 25 substantial without filtration (greater than
mg/L. Calcium removal was 3.42% with caustic 98.5% for all doses).
soda alone, and improved to 6.09 and 7.14% for
ferric doses of 20 and 25 mg/L, respectively. 4.2. Lime and soda ash addition with and
Therefore calcium removal was effectively without precipitantaids
doubled when a ferric chloride dose of 25 mg/L
was used compared to caustic soda alone. Eight jar tests were conducted with lime and
Indeed, removal of silica is decreased by soda doses that ranged from 50 mg lime and
addition of ferric chloride. This could be due to 150 mg soda to 200 mg lime and 800 mg soda.
the formation of colloidal particles in a The removal obtained for these tests is shown in
suspension, which will be determined in the Fig. 5.
following when we examine the effect of Removal of iron is almost complete (96.2%)
filtration of supernatant. Iron removal also at the lowest dose, and does not improve
decreases, which is due to the addition of ferric significantly with increased doses. Manganese
chloride, however, the additional ferric added removal is even more complete, with removal at
should precipitate out. 200 mg lime/600 mg soda dose reaching 99.9%.
The supernatant for ferric chloride tests were Clearly, the increase in the calcium
analyzed without filtration as well as with MF concentration is due to the addition of lime
and UF. Microfiltration improved iron removal, {Ca(OH)z} which has not precipitated to form
particularly for the 15 mg/L dose of ferric CaCOj. This must be due to the fact that there
chloride, where removal was 87.1% for was not enough carbonate present for it to react
unfiltered supernatant and 96.2% for with, as it can be seen from the figure that for
microfiltered supernatant. Calcium removal did the same lime dose, when soda is increased,
not significantly improve with filtration, but as calcium removal increases.
with alum, improvements were quite significant Silica removal improved as lime dose
in some cases with ultrafiltration. For a 15 mg increased, and reached a maximum of 76.9%
dose of ferric chloride calcium removal was when treated with 200 mg lime and 600 mg
improved from almost 0% with no filtration to soda. Magnesium removal improved slightly
14.3% with ultrafiltration; this indicates a high with increased doses of lime and soda,
degree of colloidal sized particles. indicating that insoluble MgC03 and Mg(OH)z
Silica removal was improved from 91.47% have been formed and removed, which may
without filtration to 93.64% with ultrafiltration explain improved silica removal since soluble
for zero dose of ferric chloride. However, with silica is adsorbed onto the surfaces of these
ferric chloride added, silica removal was not precipitates.
 R. Sheikholeslami and J. Bright/Desalination 143 (2002) 255-267 265

not as important as silica, iron and manganese

removal and it is important to keep the dose as
low as possible to minimise both chemical costs
and volume of sludge produced.
The pH of the treated solutions was quite
high, even for the lowest dose of lime and soda
where the pH was 9.7. This explains the
efficient removal of iron and manganese. For
Lime dose / Soda (mg/Lj the same dose of lime, pH was seen to increase
with the higher dose of soda (carbonate), as
Fig. 5. Removal of components (Fe, o, initial = 0.59 expected. The optimum dose of lime and soda
mg/L; Mg, n , initial = 560 mg/L; Mn, +, initial = 5.9 was chosen to be 150 mg/L lime and 450 mg/L
mg/L; Si, A, initial = 17.1 mg/L; Ca, x, initial = 52
soda ash, since after this point, removal of
mg/L) and pH (0) of solution obtained with varied
doses of lime and soda, after 30 mins settling time. components did not improve significantly.
Lime, although an inexpensive and obviously
effective precipitant, was quite difficult to work
with accurately because of its low solubility and
z’ooz”o
% 80 106
consequent need to utilize it as a suspension.
; Sodium hydroxide has the advantage of being
8 60 102
B a soluble, however it is more expensive.
F 40 98 P
Effect of addition of precipitant aid was
E
r 20 ++I 9.4 investigated for lime/soda ash runs. The
9

- o-90
optimum precipitant dose of 150 mg/L lime and
0 5 IO 15 20 25 450 mg/L soda ash was used with two different
Alum dose (mg/L] precipitant aids, namely alum and ferric
chloride, to see if these would improve
Fig. 6. Effect of alum addition to optimum lime and removal. Results are shown in Fig. 6.
soda dose of 150 mg/L lime and 450 mg/L soda (Fe, o, It is clear that alum addition only affects
initial = 0.67 mg/L; Mg, W, initial = 529 mg/L; Mn, +,
silica and calcium removal, as the removal of
initial = 6.3 mg/L; Si, A, initial = 18 mg/L; Ca, x,
initial = 59 mg/L; and solution pH (0). the other metals does not increase with alum
addition. Silica removal is marginally improved
with alum addition, from 66.3% to 69.6% with
alum dose of 15 mg/L. This improvement may
The very low magnesium removal, which be attributed to the coprecipitation of silica with
ranged from 0.04-6.7% may be explained by aluminium hydroxide, or adsorption onto
the pH, which should be between 10.5-l 1 for aluminium hydroxide surface. This increase in
effective magnesium removal. At this pH, removal (3.3%) is not substantial, however, and
magnesium non-carbonate hardness (as we have does not justify this additional step of alum
here in magnesium sulfate) is removed by addition.
precipitation as magnesium hydroxide. The Interestingly, calcium removal decreased
residual calcium ions (from the lime) can then quite substantially with alum addition. This
be removed by precipitation with carbonate could be due to the accompanying decrease in
(soda ash). Magnesium removal could be pH, since the optimum pH range for calcium
increased by increasing the dose of lime and removal is between 10-l 1, usually about 10.4.
soda ash further, however, hardness removal is This pH decrease is due to the fact that the alum
266 R. Sheikholeslami and J. Bright/Desalination 143 (2002) 255-267

solution added is itself quite acidic. It may be 100 II 0

concluded from these results that the addition of 1
g 80 106
alum is not justified as it has very little effect on 2
z 60 102
removal of the components studied. % X
7 40 98 0.
As seen in Fig. 7, the addition of ferric
z
chloride to the optimum dose of lime and soda r 20 94
L
ash had a negative effect on calcium and iron
0 9.0
removal, as removal decreased with ferric 0 5 IO 15 20 25

chloride addition for these components. The F&I, dose (mg/L)

reason for iron removal decreasing is due to the

Fig. 7. Effect of ferric chloride addition to optimum
fact that iron is being added, and clearly not all
lime and soda dose of 150 mg/L lime and 450 mg/L
of this iron is being precipitated. The final iron soda (Fe, o, initial = 0.67 mg/L; Mg, n , initial = 529
concentration was 0.024 mg/L when no ferric mg/L; Mn, +, initial = 6.3 mg/L; Si, A, initial = 18
was added, and 0.081 mg/L when 25 mg/L of mg/L; Ca, x, initial = 59 mg/L; and solution pH (0).
ferric chloride was added. Calcium removal did
not display a clear trend with ferric chloride
addition, but it is clear that ferric chloride Table 6
addition does not improve calcium removal. Removal by optimum dose of precipitants
Silica removal was improved from 66.29%
Optimum dose % Removal of component
with no ferric chloride to 71.05% with 20 mg/L
Fe Mg Mn Si Ca
ferric chloride, however with further ferric
increases, removal did not further improve. It is Caustic soda 96.18 10.21 99.21 90.59 12.81
quite clear from these results that ferric chloride 200mg/L
does not substantially improve removal of the Lime 150 mg/L 95.56 2.23 99.68 68.01 -40.64
components studied. This indicates that the and soda ash
main removal mechanism is precipitation, 450 mg/L
rather than coagulation, rendering coagulant (or
precipitant aid) futile. Coagulants are used for
colloidal particles, and since they have very
little effect here we can assume that there is The addition of precipitant aids to the
very little, if any colloidal matter present. optimum caustic soda dose does not
substantially improve removal. Alum addition,
both without filtration and with microfiltration
5. Conclusions and recommendations or ultrafiltration did not improve removal for
any of the components by more than 2%, and
The optimum precipitant dose of caustic indeed silica removal decreased by almost 2%
soda was 200 mg/L and the optimum dose of for 25 mg/L alum. Ferric chloride did not
lime and soda ash was 150 mg/L and 450 mg/L. improve the removal of any of the components
In general caustic had a superior performance regardless of filtration of supernatant, except
for Mg, Si, and Ca removal. For Mn and Fe, calcium, which was increased by 3.7% when 25
they behaved comparably at about 99% and mg/L of ferric chloride was added compared to
96%, respectively. The amount of iron, caustic soda alone. Alum addition to the
magnesium, manganese, silica and calcium optimum lime and soda ash dose also did not
removed from the model solution is significantly improve removal. Silica removal
summarized in Table 6. increased slightly (by 2-3%), however calcium
 R. Sheikholeslami and J. Bright/Desalination 143 (2002) 255-267 267

removal Ferric had similar determine more accurately the optimum dose.
effect, improving removal but Decreasing precipitant doses would reduce
decreasing calcium The chemical cost and the volume of sludge
indicated microfiltration did not provide produced.
significant improvement, if any at all, to justify
its inclusion at a step in the pretreatment. This
however might not be the case if the settling
References
time is not sufficient for particle settlement. The
results for ultrafiltration were not as conclusive
Ul G. Holmes, B.R. Singh, and L. Theodore,
due to the experimental technique and needs to Handbook of Environmental Management and
be further investigated. Technology, 1993.
The economic feasibility of adding a PI R. Sheikholeslami and S. Tan, Desalination, 126
pretreatment stage before reverse osmosis will ( 1999) 267-280.
vary from plant to plant; however, the main [31 E.F. Bellow, Chem. Eng., (1978) 85-91.
[41 R. Sheikholeslami and S. Zhou, Desalination, 132
advantage of pretreatment is the reduced
(2000) 337.
environmental liability and reduced membrane A. Amokrane, C. Comel, and J. Veron, Water
[51
fouling costs. The optimum dose will also Res., 3 1 (1997) 2775-2782.
depend on the recoveries desired, and it may 161 R.K. Iler, The Chemistry of Silica: Solubility,
not be necessary to remove components to the Polymerization, Colloid and Surface Properties,
extent that was achieved in this study. and Biochemistry, Wiley, New York, 1979.
More jar tests should be conducted, with a [71 S.D. Faust and O.M. Aly, Chemistry of Water
Treatment, Butterworths Publishers, 1983.
wider variety of doses around the range of the
PI M.J. Collie, Industrial Water Treatment
optimum precipitant dose, as it would be
Chemicals and Processes, Developments since
preferable to reduce the dose used and 1978, Nayes Data Corp., Park Ridge, NJ, 1983.