Coagulation by hydrolysing metal salts

Advances in Colloid and Interface Science 100 – 102 (2003) 475–502

Coagulation by hydrolysing metal salts Jinming Duana, John Gregoryb,* a Ian Wark Research Institute, University of South Australia, Adelaide, SA 5095, Australia Department of Civil and Environmental Engineering, University College London, Gower Street, London WC1E 6BT, UK

b

Received 20 May 2002; accepted 18 July 2002

Abstract Aluminium and iron salts are widely used as coagulants in water and wastewater treatment and in some other applications. They are effective in removing a broad range of impurities from water, including colloidal particles and dissolved organic substances. Their mode of action is generally explained in terms of two distinct mechanisms: charge neutralisation of negatively charged colloids by cationic hydrolysis products and incorporation of impurities in an amorphous hydroxide precipitate (‘sweep flocculation’). The relative importance of these mechanisms depends on factors such as pH and coagulant dosage. Alternative coagulants, based on prehydrolysed forms of aluminium and iron, are more effective than the traditional additives in many cases, but their mode of action is not completely understood, especially with regard to the role of charge neutralisation and hydroxide precipitation. Some basic features of metal hydrolysis and precipitate formation are briefly reviewed and the action of hydrolysing coagulants is then discussed, with examples from the older literature and from some recent studies on model systems. Dynamic monitoring of floc formation and breakage can give useful insights into the underlying mechanisms. Although the results can be reasonably well explained in terms of established ideas, a detailed understanding of the ‘sweep flocculation’ mechanism is not yet available. There are also still some uncertainties regarding the action of pre-hydrolysed coagulants. 䊚 2002 Elsevier Science B.V. All rights reserved. Keywords: Aluminium; Coagulation; Flocculation; Hydrolysis; Iron; Water treatment

*Corresponding author. Tel.: q44-20-7679-7818; fax: q44-20-7380-0986. E-mail address: [email protected] (J. Gregory). 0001-8686/03/$ - see front matter 䊚 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 0 1 - 8 6 8 6 Ž 0 2 . 0 0 0 6 7 - 2

476 J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502

1. Introduction Hydrolysing metal salts, based on aluminium or iron, are very widely used as coagulants in water treatment. ‘Alum’ or aluminium sulfate has been used for water purification since ancient times and was first mentioned by Pliny (approx. 77 AD). An interesting historical account has been given by Cohen and Hannah w1x. Hydrolysing coagulants have been applied routinely since early in the 20th century and play a vital role in the removal of many impurities from polluted waters. These impurities include inorganic particles, such as clays, pathogenic microbes and dissolved natural organic matter. The most common additives are aluminium sulfate (generally known as ‘alum’), ferric chloride and ferric sulfate. Other products based on pre-hydrolysed metals are also now widely used, including a range of materials referred to as polyaluminium chloride. Nearly all colloidal impurities in water are negatively charged and, hence, may be stable as a result of electrical repulsion. Destabilisation could be achieved along DLVO lines, either by adding relatively large amounts of salts or smaller quantities of cations that interact specifically with negative colloids and neutralise their charge. Highly charged cations such as Al3q and Fe3q should be effective in this respect. However, over the normal range of pH values in natural waters (say, 5–8), these simple cations are not found in significant concentrations, as a result of hydrolysis, which can give a range of products. Many hydrolysis products are cationic and these can interact strongly with negative colloids, giving destabilisation and coagulation, under the correct conditions of dosage and pH. Excess dosage can give charge reversal and restabilisation of colloids. At around neutral pH both Al(III) and Fe(III) have limited solubility, because of the precipitation of an amorphous hydroxide, which can play a very important role in practical coagulation and flocculation processes. Positively charged precipitate particles may deposit on impurity particles (heterocoagulation), again giving the possibility of charge neutralisation and destabilisation. A further possibility is that surface precipitation of hydroxide could occur, with similar consequences. More importantly in practice, hydroxide precipitation leads to the possibility of sweep flocculation, in which impurity particles become enmeshed in the growing precipitate and thus effectively removed. These additives can also remove dissolved natural organic matter (NOM), either by charge neutralisation to give insoluble forms, or by adsorption on precipitated metal hydroxide. As well as simple hydrolysing salts, a range of commercial pre-hydrolysed coagulants is available. These contain cationic hydrolysis products and are often more effective than aluminium or iron salts. Although the broad principles of action of these coagulants are reasonably well understood, there are still some uncertainties regarding the nature of the active species, the role of other salts, especially anions, in water, and the nature of the aggregates formed. The mode of action of pre-hydrolysed agents is not yet fully understood. A review of the current state of knowledge will be given, with some examples of recent experimental results on model systems.

J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502 477

2. Hydrolysis of Al(III) and Fe(III) 2.1. Monomeric hydrolysis products All metal cations are hydrated to some extent in water. It is reasonable to think in terms of a primary hydration shell, where water molecules are in direct contact with the central metal ion, and more loosely held water in a secondary hydration shell. In the cases of Al3q and Fe3q, it is known that the primary hydration shell consists of six water molecules in octahedral co-ordination w2x. Owing to the high charge on the metal ion, water molecules in the primary hydration shell are polarised and this can lead to a loss of one or more protons, depending on the solution pH. Effectively, this means that the water molecules in the hydration shell are progressively replaced by hydroxyl ions, giving a lower positive charge, according to the following sequence (omitting co-ordinated water molecules for convenience): y Me3q™Me(OH)2q™Me(OH)q 2 ™Me(OH)3™Me(OH)4

This is an oversimplified scheme, since it is known that dimeric, trimeric and polynuclear hydrolysis products of Al and Fe can form. However, these can often be ignored, especially in dilute solutions, and may not greatly affect the overall metal speciation. Polynuclear hydrolysis products will be considered in Section 2.2. The hydrolysis scheme above will proceed from left to right as the pH is increased, giving first the doubly- and singly-charged cationic species and then the uncharged metal hydroxide, Me(OH)3. In the case of both aluminium and iron, the hydroxide is of very low solubility and an amorphous precipitate can form at intermediate pH values. This is of enormous practical significance in the action of these materials as coagulants. With further increase in pH, the soluble anionic form Me(OH)y 4 becomes dominant. Because of the formation of insoluble hydroxides (and also polynuclear species— see Section 2.2), the determination of hydrolysis constants can be difficult and there are significant differences in some published values (see e.g. Wesolowski and Palmer w3x). Hydrolysis constants can be defined for successive deprotonations in terms of the following equations: M3qqH2OlM(OH)2qqHq M(OH)

2q

q 2

qH2OlM(OH) qH

K1 q

K2

M(OH)2qqH2OlM(OH)3qHq

K3

q M(OH)3qH2OlM(OH)y 4 qH

K4

A solubility constant for the metal hydroxide is also needed: M(OH)3lM3qq3OHy

KS

Although the most stable solids are crystalline forms of metal hydroxides, such as gibbsite and goethite in the case of Al and Fe, respectively, these are usually formed very slowly (typically weeks or months). In the context of coagulation

478 J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502 Table 1 Hydrolysis and solubility constants for Al3q and Fe3q for zero ionic strength and 25 8C. (Values taken from references w3x and w4x)

3q

Al Fe3q

pK1

pK2

pK3

pK4

pKSam

4.95 2.2

5.6 3.5

6.7 6

5.6 10

31.5 38

mechanisms, it is more relevant to consider the solubility of the amorphous precipitates that form initially. However, solubility constants for the amorphous forms, KSam are not known precisely and only estimated values can be quoted. Table 1 gives values for hydrolysis and solubility constants (in pK form), taken from Wesolowski and Palmer w3x for Al and from Flynn w4x for Fe. The values are for conditions of zero ionic strength and 25 8C. Reference w3x gives extensive data for Al at other temperatures and ionic strengths. Using the values in Table 1, it is possible to plot, as a function of pH, the concentrations of the various species in equilibrium with the amorphous hydroxide precipitate. Such diagrams are shown in Fig. 1 for Al and Fe. The total amount of soluble species in equilibrium with the amorphous solid is effectively the solubility of the metal and it can be seen that in each case there is a minimum solubility at a certain pH value. For Al this is approximately pH 6, at which the solubility is of the order of 1 mM. Measurements of Al solubility as a function of pH give reasonable agreement with such calculations w5x, even though only monomeric species are included. For Fe, the minimum solubility is much lower—less than 0.01 mM, and the corresponding region is broader than for Al. Martin w6x has pointed out a significant difference in the hydrolysis behaviour of Al and Fe, which is apparent from the values in Table 1 and the computed results in Fig. 1. The hydrolysis constants for Al cover a much narrower range than those for Fe. The latter are spaced over approximately 8 pH units, whereas all of the Al deprotonations are ‘squeezed’ into an interval of less than 1 unit. Martin explained this feature by the transition from the octahedral hexahydrate Al3q.6H2O to the tetrahedral Al(OH)y 4 . This makes the successive hydrolysis steps co-operative in nature. All of the hydrolysed species for Fe3q retain the octahedral co-ordination and the stepwise deprotonations show the expected spread of pK values. This difference is clearly seen in the plots of species distributions in Fig. 2, which show the mole fraction of the various soluble hydrolysis products in equilibrium with the amorphous precipitate. The ferric species each attain significant relative concentrations in solution at appropriate pH values, whereas for Al, apart from a narrow pH region approximately 5–6, the dominant soluble species are Al3q and Al(OH)y 4 at low and high pH, respectively. 2.2. Polynuclear species As well as the simple monomeric hydrolysis products discussed above, there are many possible polynuclear forms that could be considered w2x. For Al, these include

J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502 479

Fig. 1. Concentrations of monomeric hydrolysis products of Fe(III) and Al(III) in equilibrium with the amorphous hydroxides, at zero ionic strength and 25 8C.

Al2(OH)4q and Al3(OH)45q and there are equivalent species for Fe. Formation 2 constants for the dimers and trimers are known, but, for practical purposes, they do not significantly affect the speciation shown in Figs. 1 and 2. Martin w6x showed that the Fe dimer, Fe2(OH)4q 2 , could become significant in acid solutions (pH-3), but the corresponding Al dimer does not occur to any significant extent in saturated solutions of Al(OH)3. From the standpoint of coagulation with simple Al and Fe salts, only monomeric hydrolysis products and the amorphous hydroxide precipitate need be considered. Polynuclear hydrolysis products can be prepared in significant amounts under certain conditions. The best known of these is Al13O4(OH)7q 24 or ‘Al13’, which can be formed by controlled neutralisation of aluminium salt solutions or by several other methods. This tridecamer has the so-called ‘keggin’ structure, consisting of a central tetrahedral AlO5y unit surrounded by 12 Al octahedra with shared edges. 4 The tetrahedral and octahedral Al sites can be easily distinguished in the 27Al NMR spectrum w7x. The structure has also been confirmed by small angle X-ray methods

480 J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502

Fig. 2. Proportions (mole fractions) of dissolved hydrolysis products in equilibrium with amorphous hydroxides.

w8x and by potentiometric titration w9x. Under appropriate conditions, Al13 forms fairly rapidly and essentially irreversibly, remaining stable in aqueous solutions for long periods. The tridecamer has been detected in the natural aquatic environment, an acid forest soil water w10x. The Al13 unit has an ionic radius in solution of approximately 1.3 nm w8x. Other polynuclear species, such as the octamer, Al8(OH)4q 20 , have been proposed, based on coagulation data w11x. However, there is no direct evidence for the octamer and it is unlikely to be significant in practice. The titration of an aluminium salt solution with base typically gives a curve like that in Fig. 3, in which four regions can be distinguished, based on the amount of added base B (sOHyAl). In Region 1 the base neutralises free acid produced by spontaneous hydrolysis, giving a rapid increase of pH. Above approximately pH 4, hydrolysed species are formed and there is an extended phase (Region 2) where pH increases only slowly, since added base is consumed by hydrolysis. In this region,

J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502 481

Fig. 3. Titration curves for neutralisation of aluminium salt solutions, showing variation of pH with added base B (sOHyAl)

large quantities of polymeric species can be formed and become the predominant soluble species at high B values. During titration, depending on mixing conditions, added base can give local supersaturation and precipitation of the amorphous hydroxide, or excess formation of the Al(OH)y 4 ion. These conditions may favour the formation of the Al13 polymer, although the details are not clear. Further increase of B, to the range 2.4–2.8, gives Region 3, in which a small shoulder appears in the curve, following a sharp rise in pH. In this region, supersaturation of the solution with respect to amorphous Al(OH)3 and rapid precipitation occurs. This shoulder disappears in the presence of highly charged anions, such as sulfate, which can promote hydroxide precipitation w12x. In Region 4, the added base reduces the positive surface charge of the colloidal hydroxide particles and visible precipitates are formed. Further addition of base gives a rapid increase in pH. The speciation of Al solutions can be conveniently studied by a timed colorimetric reaction with ferron reagent (8-hydroxy-7-iodo-5-quinoline sulfonic acid). This method was introduced by Smith w13x and is based on the observation that different forms of Al react at varying rates with ferron. Mononuclear Al species (Ala) react almost instantaneously and polynuclear species (Alb) much more slowly. Colloidal or precipitated Al (Alc) shows practically no reaction with ferron. The proportion of polynuclear species (Alb) determined by the ferron technique corresponds quite well with results from membrane filtration tests w14x and with the proportion of Al13 from NMR studies w15x. Typically, the proportion of the various species is determined as a function of the degree of neutralisation, B. For 0.1 M AlCl3 solution, titrated with 0.5 M NaOH w13x, the proportion of Alb increased from approximately 33% to 83% as B was raised from 1.0 to 2.5. For values of B above 2.5, the amount of precipitate (Alc) increased significantly. In the same study, it was shown that the Alb fraction was quite stable to dilution and to changes in pH,

482 J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502

and only slowly converted to other forms. This is in marked contrast to monomeric hydrolysis products, which respond very rapidly to changes in chemical conditions. It is doubtful whether Al13 forms under conditions where aluminium salts are added to water at around neutral pH, to give low Al concentrations (typical of water treatment conditions). In this case, it is thought that monomeric hydrolysed species predominate in solution and that amorphous precipitates form without the involvement of Al13 species (see below). Polymeric hydrolysis products of ferric salts can also be prepared and characterised by ferron analysis w14,16x. In partially neutralised ferric chloride solutions, it has been suggested that, in the range of B (OHyFe) up to 1.0, monomeric and di- and trimeric species are predominant in solution w17x. The iron trimer, Fe3(OH)45q , was postulated as the nucleus for phase transition in supersaturated ferric chloride solutions w18x, which may build up larger polymers such as Fe6 and Fe9 with size approximately 1.66 nm w19x. Bottero et al. w20x reported the formation of Fe24 polycation in hydrolysed ferric chloride solution and they observed that fractal aggregates form with Fe24 as subunits, by using a small angle X-ray scattering (SAXS) technique. In such solutions, polycations are first produced, and further neutralisation causes aggregation of these polycations leading to formation of fractal polymers. 2.3. Precipitate formation Precipitated metal hydroxides can be formed in various ways, such as by neutralisation of the metal salt to B (OHyM) ratios of approximately 3, as mentioned above. However, the precise mechanisms and structure of the precipitate have been the subject of much debate and there is a rather extensive and confusing literature on the topic. We shall restrict attention here mainly to aluminium. In the case of iron, there are redox as well as hydrolytic reactions to consider w21x and the subject becomes quite complex. An early hydrolysis–precipitation model for aluminium involved the two-dimensional growth of hexameric ring units w22x, although there is no direct experimental evidence for this so-called ‘coreqlinks’ model. Several studies w23,24x have suggested that the initial stage in the formation of gels by neutralisation of aluminium solutions consists of the aggregation of tridecamer (Al13) units, although detailed mechanisms are still not clear. Bottero et al. w24x concluded, from NMR, infra-red and small angle X-ray studies, that the nature of the clusters formed depends very much on the OHyAl ratio. For Bs2.5 at short times, chain-like clusters of low fractal dimension and size of approximately 40 nm are formed. With increasing B the clusters become larger and more compact and there is a progressive loss of the tetrahedral Al. The initially precipitated form undergoes rearrangement on ageing or heating and eventually attains long-range crystalline order. However, in the context of coagulation by hydrolysing metal salts, it is the rapidly-formed amorphous precipitate, which is of most interest.

J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502 483

Details of the neutralisation procedure, such as temperature, stirring conditions and base injection rate can have a very important influence. Ohman and Wagberg w25x showed that different routes to a neutralised Al solution could give significantly different precipitate properties. They used a stopped-flow ‘flash’ neutralisation technique with AlCl3 solution, a partially prehydrolysed AlCl3 solution and a solution of sodium aluminate. These showed broadly similar titration curves, but significant differences in the particle size at B values of 3 or more. In all cases the particles showed a dramatic increase in size at pH values of approximately 7 or greater. The effect of mixing conditions on aluminium precipitation has been considered in some detail by Clark et al. w26x. They studied the neutralisation of AlCl3 solutions in a stirred tank reactor and considered the kinetics of hydrolysis reactions in relation to characteristic mixing time scales. This work showed that there is a competition between the formation of polynuclear hydrolysis products and precipitated solid. With more intense mixing, the results indicated that precipitation would be favoured. Neutralisation of fairly concentrated metal solutions by added base is not directly relevant to the use of hydrolysing metal coagulants in practice, where the metal salt is added to water, usually containing excess alkalinity, to give a final concentration of approximately 0.1 mM or less. In this case, neutralisation would occur rapidly and it is very likely that precipitation would occur without the formation of significant polynuclear species such as Al13 w5x. It has been shown that precipitates formed by addition of aluminium sulfate and polyaluminium chloride (a prehydrolysed solution containing Al13) give different solid phases. In the latter case, the polymeric structure is maintained in the precipitate and after re-dissolution in acid w5x. Similar behaviour has been found for ferric sulfate and a pre-hydrolysed form (polyferric sulfate) w27x. The surface charge characteristics of precipitated metal hydroxides are of great importance in coagulation. In common with oxides and other minerals they show an isoelectric point (i.e.p.) at which the apparent (electrokinetic) surface charge is zero. At pH values below the i.e.p. the precipitate is positively charged and at higher pH values it has a negative charge. The value of the i.e.p. depends on the preparation details and on the solution composition, so that there are significant differences in values reported in the literature. Precipitation from aluminium chloride solutions gives a solid with an i.e.p. in the region of 9 w25x, whereas from aluminium sulfate the value is closer to 8 w5x. For amorphous ferric hydroxide the i.e.p. is somewhat lower. For both aluminium and ferric salts, pre-hydrolysed forms give precipitates with i.e.p. values shifted upwards by one or more pH units. The varying charge with pH can greatly affect the precipitation process. At around neutral pH for aluminium, the initially formed colloidal precipitate is positively charged and, hence, is colloidally stable. As the pH is increased towards the i.e.p., the stability decreases and the particles can aggregate into large, settleable flocs. Hayden and Rubin w28x carried out a light scattering study on precipitation from an

484 J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502

aluminium nitrate solution (1 mM) as a function of pH. At low pH, the solution appeared clear, but showed a Tyndall beam, indicating the presence of very small, colloidal particles. As the pH increased, the particle size increased, giving higher turbidity. Above approximately pH 7 much larger particles were formed, which settled rapidly to give a reduced turbidity. The i.e.p. in this case was approximately 8, in the middle of the pH range where settleable flocs were produced. The results of Ohman and Wagberg w25x are consistent with these findings, although their precipitates were from AlCl3 solutions and showed a rather higher i.e.p. value. It is worth noting that the i.e.p. for Al(OH)3 occurs at a pH value well above that of minimum solubility (Fig. 1), so that the largest flocs do not correspond with the maximum amount of precipitate. The presence of highly charged anions, such as sulfate, can have a large effect on hydroxide precipitation. Sulfate can reduce the positive charge of the precipitate in the acid region, so that large flocs are formed over a wider pH range. This was clearly shown by Hayden and Rubin w28x and others. The sulfate effect has been known since the 1930s w1x and is very important in practice since aluminium and ferric sulfates are commonly used as coagulants and natural waters can contain significant amounts of sulfate. 3. Mechanisms of coagulation 3.1. General Natural waters contain a very wide variety of particulate impurities. These include inorganic substances such as clays and metal oxides, various organic colloids and microbes such as viruses, bacteria, protozoa and algae. Aquatic particles cover a broad range of particle size, from nm to mm dimensions and present a significant challenge in water treatment technology. For smaller particles, separation efficiency can be greatly enhanced by aggregation to give an increased size (coagulationy flocculation). Over the usual range of natural water pH (say, 5–9) particles nearly always carry a negative surface charge. This may be because the water pH is above the isoelectric point, which is usually the case for particles of biological origin. Even for mineral particles with a fairly high i.e.p., adsorption of natural organic matter usually gives a negative surface charge w29x. Because of their surface charge, aquatic particles are often colloidally stable and resistant to aggregation. For this reason, coagulants are needed to destabilise the particles. According to the classical ideas of colloid stability, destabilisation can be brought about by either: ● an increase in ionic strength, giving some reduction in the zeta potential and a decreased thickness of the diffuse part of the electrical double layer, or ● specific adsorption of counterions to neutralise the particle charge. In both cases, additives effective for negative particles should be salts with highly charged cations. It is unlikely that a sufficient increase in ionic strength would be a

J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502 485

practical destabilisation method, but counterion adsorption is much more promising, since quite small amounts of additive would usually be sufficient. Aluminium and iron salts give cationic hydrolysis products that are strongly adsorbed on negative particles and can give effective destabilisation. Polymeric additives can also be used to cause aggregation of particles and they may act either by polymer bridging or charge neutralisation (including ‘electrostatic patch’ effects) w30x. The action of hydrolysing metal coagulants can involve similar mechanisms. We shall initially consider the action of coagulants with respect to particulate impurities in water with only ‘simple’ salts of Al and Fe. Later sections will deal with pre-hydrolysed coagulants and dissolved organic matter. 3.2. Charge neutralisation At very low concentrations of metal, only soluble species are present—the hydrated metal ion and various hydrolysed species, which, assuming only monomeric forms will depend on solution pH, as shown in Fig. 1. It is generally thought that hydrolysed cationic species such as Al(OH)2q are more strongly adsorbed on negative surfaces than the free, hydrated metal ion w31x. Adsorbed metal ions may be in the form of outer sphere or inner sphere complexes w32x. In the former case there is at least one water molecule separating the cation from the surface, i.e. the cation retains its hydration shell. Inner sphere complexes involve the direct coordination of the metal ion to surface groups, with no intervening water. Models for surface complex formation (e.g. Stumm w32x), are mainly for metal oxide surfaces, and involve specific parameters, such as binding constants. The process is analogous to complex formation in solution and only monolayer coverage can occur. However, it is known that hydrolysing coagulants can neutralise the negative surface charge of many types of particle, including bacteria and clays, and it is unlikely that a specific complexation reaction will provide an explanation of all the observed effects. Generally, charge neutralisation with aluminium salts occurs at quite low metal concentrations—typically a few mM at around neutral pH. Letterman et al. w33x found that, for several inorganic suspensions at pH 6, the amount of aluminium needed to bring the electrophoretic mobility to zero was in the region of 5 mM Al per m2 of particle surface. Inspection of Fig. 1 shows that, even at very low concentrations, the solubility limit of the hydroxide may be exceeded. Also, at neutral pH, cationic hydrolysis products should represent only a tiny fraction of the total soluble Al (Fig. 2), the dominant form being the aluminate ion. This suggests that the effective charge-neutralising species may be colloidal hydroxide particles, which should be positively charged up to approximately pH 8. Even when the bulk hydroxide solubility is not exceeded, a form of surface precipitation may take place. James and Healy w34x among others have suggested that adsorption of soluble hydroxide can lead to a layer of amorphous hydroxide precipitate, by surface nucleation and precipitation.

486 J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502

A surface precipitation model was proposed by Farley et al. w35x to explain cation adsorption on oxide surfaces. In this model, when cations adsorb to the surface of a mineral a precipitate of the cation with the constituent ions of the mineral surface may form at high surface coverage. This allows sorption to occur through a continuum between surface complex formation and bulk solution precipitation of the sorbing ion. As the cation is complexed at the surface, a new hydroxide surface is formed. In the model, cations at the solid (oxide) water interface are treated as surface species, while those not in contact with the solution phase are treated as solid species forming a solid solution. In this case, at low adsorbing cation concentration, surface complexation is the dominant mechanism. However, as the sorbate concentration increases, both the surface complex concentration and the amount of the surface precipitate increase until the surface sites become saturated. Surface precipitation then becomes the dominant sorption mechanism. Furthermore, as bulk solution precipitation is approached, the amount of the surface precipitate becomes large. It follows that, as metal is adsorbed on a surface, a new hydroxide surface will be formed, allowing further mass transfer of metal to the solid phase. This can give multilayer sorption, in contrast to surface complex formation. In practice it is often quite difficult to distinguish between surface precipitation and the deposition of colloidal hydroxide particles which have been precipitated in bulk solution. A combination of these effects is included in the Precipitation Charge Neutralisation (PCN) model, which was introduced by Dentel w36x to explain coagulation by hydrolysing metal salts in water treatment. A schematic illustration of the processes involved is given in Fig. 4. The PCN model has been presented in a quantitative form w37x although this aspect will not be covered here. According to the PCN model, coagulation with aluminium or iron salts involves three steps: 1. Destabilisation begins after addition of a dose of coagulant that exceeds the operational solubility limit of aluminium (or iron) hydroxide. 2. Aluminium or iron hydroxide species are then deposited onto colloidal surfaces, as shown in Fig. 4. This figure shows that metal hydroxide could end up on particle surfaces by several possible pathways. 3. Under typical conditions, metal hydroxide is positively charged, while the original colloidal particles are negatively charged. So the deposition process can result in charge neutralisation or charge reversal of the colloidal particles at certain doses, as shown in a simplified manner in Fig. 5. If the positively charged adsorbed species are in the form of isolated regions, then a form of ‘electrostatic patch’ attraction may be important, as in the case of polyelectrolytes w30x, although this does not seem to have been examined systematically for hydrolysing coagulants. It is important to note that the PCN model does not consider bulk hydroxide precipitation and ‘sweep flocculation’, which will be discussed in the next section. There is no doubt that, at the correct dosage, charge neutralisation by adsorbed hydrolysis products andyor hydroxide precipitate can cause negatively charged particles to become destabilised and hence to coagulate. When electrophoretic

J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502 487

Fig. 4. Schematic illustration of the concept of the Precipitation Charge Neutralisation (PCN) model. (After Dentel w36x.)

mobility (EM) measurements are carried out, it is evident that the optimum coagulation dosage corresponds with the condition where the zeta potential of the particles is close to zero. Some results for kaolin suspensions (50 mgyl) coagulated with aluminium sulfate (‘alum’) at pH 6, are shown in Fig. 6 w38x. This shows the residual turbidity of the suspensions after standard stirring and settling conditions (jar test) as well as the electrophoretic mobility of the particles, soon after coagulant addition. Since the minimum residual turbidity occurs at approximately 8 mM Al, this is the ‘optimum dosage’ for particle separation. The EM value is very close to zero at this point, indicating that charge neutralisation is responsible for the destabilisation of the clay particles. At slightly higher alum dosages, the EM becomes positive and the residual turbidity increases, indicating that charge reversal causes restabilisation of the particles. Very similar results were obtained for kaolin by Letterman and Vanderbrook w39x. According to Fig. 1, 8 mM Al at pH 6 is slightly above the solubility limit of amorphous Al(OH)3. At higher pH values the optimum alum dosage increases because of the decreased positive charge of the adsorbed species. In such systems it appears that the electrokinetic properties of the particles are very like those of the amorphous hydroxide precipitate w33x, with an isoelectric point in the pH region 8–9, depending

Fig. 5. Deposition of metal hydroxide species on oppositely-charged particles, showing charge neutralisation and charge reversal. (After Dentel w36x.)

488 J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502

Fig. 6. Electrophoretic mobility (EM) and residual turbidity for kaolin suspensions (50 mgyl) with low dosages of aluminium sulfate (‘alum’) at pH 6. (Replotted from data of Duan w38x.)

on the anions present in solution. At around the i.e.p. the particles do not become positively charged, even at high Al dosages and so no restabilisation is observed. If charge neutralisation is the predominant destabilisation mechanism, then there should be a stoichiometric relationship between the particle concentration and the optimum coagulant dosage w40x. At low particle concentrations, low coagulant dosages should be required. Under these conditions coagulation rates can be very low, which causes problems in water treatment. Another practical difficulty is that the optimum coagulant dosage range can be quite narrow, which means that rather precise dosing control is needed. Both of these difficulties can be overcome by using higher coagulant dosages, where extensive hydroxide precipitation occurs, giving sweep flocculation. 3.3. Sweep flocculation It has long been recognised w41x that, in many cases, optimal removal of particles from water is achieved under conditions of rapid and extensive hydroxide precipitation. In the case of aluminium coagulants, optimum pH values are approximately 7, close to the minimum solubility (Fig. 1) but close enough to the i.e.p. to give fairly rapid aggregation of the colloidal precipitate particles. Although details are not fully understood, it seems clear that impurity particles are enmeshed in a growing hydroxide precipitate and are effectively removed from suspension. This process has become known as ‘sweep flocculation’ since particles are ‘swept out’ of water by an amorphous hydroxide precipitate. Sweep flocculation generally gives considerably improved particle removal than when particles are destabilised just by charge neutralisation. At least part of the reason is the greatly improved rate of aggregation, because of the increased solids concentration. Hydroxide precipitates tend to have a rather open structure, so that even a small mass can give a large effective volume concentration and, hence, a

J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502 489

high probability of capturing other particles. It is also possible that binding (‘bridging’) of particles by precipitated hydroxide may give stronger aggregates. Increasing the coagulant dosage in the sweep region gives progressively larger volumes of sediment w42x but, beyond the operational optimum dosage, there is little further improvement in particle removal. The different mechanisms outlined above have led to the definition of four zones of coagulant dosage, with the following consequences for negatively charged particles: Zone Zone Zone Zone

1: 2: 3: 4:

Very low coagulant dosage; particles still negative and hence stable. Dosage sufficient to give charge neutralisation and hence coagulation. Higher dosage giving charge neutralisation and restabilisation. Still higher dosage giving hydroxide precipitate and sweep flocculation.

The example in Fig. 7 shows the results of jar test and EM measurements for kaolin suspensions with alum at pH 7 w38x. Below approximately 8 mM Al there is essentially no reduction in turbidity, since the particles are negatively charged and colloidally stable (Zone 1). There is a narrow range of lowered turbidity in the region of 15 mM Al, which is close to the dosage where the EM is reduced to zero (Zone 2). By 20 mM Al, the particles are positively charged and completely restabilised, since the residual turbidity is no lower than that for the original clay suspension (Zone 3). Beyond approximately 60 mM Al the turbidity falls again as a result of sweep flocculation (Zone 4). It is very significant that a substantial change in residual turbidity occurs in a region of alum dosage where the EM of the particles is still positive and shows no appreciable reduction. Although there is a gradual reduction in EM as the alum dosage is increased, this is not obviously related to the degree of turbidity removal. It is also worth noting that the residual turbidity in Zone 4 is significantly lower than in Zone 2, indicating a much greater degree of clarification by sweep flocculation. These experiments were supplemented by dynamic measurements of floc growth, using a simple optical monitoring technique w43x, based on the principle of ‘turbidity fluctuations’. This gives a semi-empirical Flocculation Index (FI), which is strongly correlated with floc size. Fig. 8 shows the change in FI with time for two different alum dosages at pH 7: 5 and 40 mM Al2(SO4)3 (10 and 80 mM Al). The first of these is well within Zone 2, where charge neutralisation is the operative mechanism and the higher dosage is that where the onset of optimal sweep flocculation occurs. There are very significant differences between the two curves in Fig. 8. For 5 mM alum, the FI value begins to increase very soon after dosing (the first minute corresponds to the ‘rapid mixing’ phase of the jar test, where little floc growth occurs). Flocs then grow quite slowly and the FI reaches a plateau, corresponding to a limiting floc size, which depends on the stirring rate. This is consistent with the fairly rapid adsorption of charge-neutralising species on the kaolin particles, followed by quite slow coagulation. The aggregates (flocs) formed are quite weak and grow only to a rather small size. The rapid charge neutralisation is also indicated by the fact that all of the EM reduction occurs during the initial rapid mix phase.

490 J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502

Fig. 7. As Fig. 6, but over a wider range of alum dosages and at pH 7. (Replotted from data of Duan w38x.)

For the dosage of 40 mM Alum, the onset of floc formation is considerably delayed—significant rise in the FI value does not begin until approximately 5 min after dosing. However, a very rapid rise then occurs and the FI reaches a value nearly four times that in the other case, indicating much larger flocs. The lag time observed is related to the time required to form relatively large amorphous hydroxide precipitate particles. In the same study w38x it was shown that particles of aluminium hydroxide in the same solution at pH 7, but without kaolin, took several minutes to grow to a detectable size and the delay was of the same order as that observed for the onset of flocculation in Fig. 8. The delay can be considerably reduced by increasing the alum dosage or by increasing the pH to approximately 8, which is the i.e.p. of the hydroxide precipitate in this system. However, even with a more

Fig. 8. Dynamic monitoring of kaolin suspensions at two dosages of alum, under the same conditions as for Fig. 7. (Replotted from data of Duan w38x.)

J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502 491

rapid onset of flocculation, the rate of growth is not significantly greater and the final FI value is about the same. These results confirm that there are very important differences between destabilisation by charge neutralisation and sweep flocculation. In particular, flocs form more rapidly (perhaps after an initial delay) and can become much larger in the case of sweep flocculation, so that a greater degree of separation can be achieved. It seems that these effects are closely connected with the formation of a bulk hydroxide precipitate, initially in the form of very small colloidal particles (a few nm in size), which are positively charged at around neutral pH. It is likely that some of these particles form a coating on the impurity particles, reversing their charge. Subsequently, aggregation of the colloidal hydroxide particles occurs, either on the particle surfaces (a form of heterocoagulation) or in bulk solution. Details of this process are still not clear, but microscopic observation of flocs produced under ‘sweep’ conditions show the original impurity particles embedded in an amorphous precipitate. A schematic diagram showing a possible sequence of events in sweep flocculation with aluminium salts is given in Fig. 9. The Smoluchowski theory for particle aggregation in shear fields (orthokinetic flocculation) leads to the conclusion that flocculation rate is directly proportional to the effective particle volume w44x. Growing hydroxide precipitate consists of very small primary particles in a rather open, fractal structure and it is easy to show that the effective floc volume can become quite large, even for low coagulant doses. It is found that the settled floc volume increases in proportion to the dosage of hydrolysing coagulants under sweep conditions w42x. This probably accounts for the enhanced aggregation rate in sweep flocculation. Simply neutralising the particle charge with a rather thin layer of adsorbed species would not give a significantly increased collision radius. Although the broad principles are reasonably well understood, there are several complications with hydrolysing coagulants, which can be important in practice. These will be discussed in the next sections. 3.4. Floc strength and breakage Practical applications of hydrolysing coagulants are nearly always under conditions of turbulent fluid motion. Mixing of coagulant involves quite intense agitation (‘rapid mix’) for a short time and this is usually followed by a longer period of gentler mixing, either in a stirred tank or some form of hydraulic flocculator. The purpose of the second phase is to promote orthokinetic collisions of particles and hence floc growth. Flocs grow initially at a rate that depends on the energy dissipation (or applied shear), as well as on the particle concentration and collision efficiency. As flocs become larger, further growth is restricted by the applied shear for essentially two reasons. Existing flocs may be broken as a result of disruptive forces w45x and the collision efficiency of particles in a shear field becomes lower as particle size increases w46x. A dynamic balance between floc growth and breakage often leads to a steady-state floc size distribution, where the limiting size depends on the applied shear rate w47x.

492 J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502

Fig. 9. Schematic diagram showing the interaction of aluminium species with initially negatively charged particles in water. The particles on the right hand side are initially stable and then become destabilised by charge neutralisation. At higher coagulant dosages they can become restabilised by charge reversal and incorporated in a flocculent hydroxide precipitate (‘sweep flocculation’).

If the effective shear rate is increased, pre-formed flocs can be broken, in a manner, which depends on the floc size relative to the turbulence microscale w47x. Flocs formed by hydrolysing coagulants tend to be rather weak, so that breakage occurs readily. In the case of sweep flocculation, this breakage is not fully reversible, so that flocs do not completely re-form when the original shear conditions are restored. This effect is well known in practice, but has received rather little systematic attention w48x. Recent work w49x has given more detailed information on the subject, although the underlying mechanisms are still not well understood. Using the same dynamic monitoring method mentioned earlier, Yukselen and Gregory w49x showed that flocs formed with kaolin suspensions and aluminium sulfate were broken irreversibly at high stirring speeds and that the degree of irreversibility depended on the time of breakage. An example of their results is shown in Fig. 10. In this case a kaolin suspension (50 mgyl) was flocculated by alum at a dosage of 130 mM Al at around neutral pH (i.e. well within the sweep floc regime). Flocs were formed by stirring for 10 min at 50 rpm and then the stirring rate was increased to 400 rpm. These stirring rates correspond to mean shear rates (G values) of approximately 25 and 520 sy1, respectively. The high stirring speed was maintained for between 10 and 300 s and was then reduced to the original 50 rpm. From the change in Flocculation Index with time, it is clear that, under slow stirring conditions, flocs grew fairly rapidly to a limiting size. When the stirring speed was increased, there was an immediate and rapid drop in the FI value, by a factor of nearly 10, showing very substantial breakage of flocs. Most of the breakage was achieved after approximately 10 s. When the stirring rate was returned to 50 rpm, some re-growth of flocs occurred, but not back to the

J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502 493

Fig. 10. Dynamic monitoring of kaolin suspensions, showing formation, breakage and partial re-formation of flocs. In all cases suspensions were dosed with alum (130 mM Al) and stirred at 50 rpm for 10 min. A higher stirring speed (400 rpm) was then applied for times of 10–300 s (as indicated on curves), followed by further stirring at 50 rpm. (From w49x.)

previous FI value. Furthermore, the degree of recovery decreased for longer breakage times. This is especially apparent for the 300 s breakage case. As yet, there is no adequate model to explain these findings. The effect seems to be specific to the sweep flocculation case, since flocs formed using cationic polyelectrolytes, which destabilise the clay particles by charge neutralisation and electrostatic patch effects, showed almost complete re-formation after breakage. It is likely that breakage of metal hydroxide flocs involves rupture of chemical bonds, which are unable to re-form. 3.5. Effect of anions Solution chemistry has considerable influence on coagulation by hydrolysing metal ions. This influence will depend on how strongly anions can co-ordinate with aluminium in terms of replacement of hydroxyl ion w50x, or how they affect the kinetics of precipitation w51x. Matijevic w31x suggested that it is the solution chemistry, particularly the pH and destabilising anion concentrations that determine whether or not the precipitate coated particles will be flocculated. Letterman and Vanderbrook w39x suggested that Zone 4 coagulation would be controlled by the solubility of the adsorbed aluminium hydroxide precipitate and surface ionisation and sulfate complex formation reactions. Among the anions, nitrate has very little tendency to co-ordinate with metal ions, and does not have a significant influence on destabilisation with metal coagulants. However, anions such as bicarbonate, chloride, sulfate, etc. have considerable effects on coagulation by aluminium salts.

494 J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502

The effects of bicarbonate ion on coagulation of kaolin particles by aluminium salts have been studied w51,52x. The traditional view of the significance of the bicarbonate ion in the coagulation or flocculation of suspensions using hydrolysing coagulants has been associated with its contribution to the alkalinity (buffer capacity) of the water. The importance of maintaining a buffered solution is related to the chemical and physical characteristics of hydroxide precipitates w51x. If a suspension does not have sufficient alkalinity, addition of the hydrolysing metal salt may significantly depress solution pH, so that electrical charge andyor colloidal properties of the precipitate would be greatly affected. Precipitation might even be prevented if the solution pH is reduced beyond the solubility range of the hydroxide. Because ferric hydroxide is much less soluble than aluminium hydroxide (Fig. 1) the former is precipitated over a much broader pH range. For this reason, sweep flocculation by ferric salts is less sensitive to pH w41x. Generally, bicarbonate, sulfate and chloride, have little or no effect on the pH of aluminium precipitation. However, they may exert great influence on the range of pH values where the initial precipitate can aggregate to settleable flocs w28x. Sulfate is a moderately strong co-ordinator with aluminium, and the presence of sulfate ion extends the pH range of coagulation towards the acid side under normal coagulation condition w39,52x. Near to the isoelectric pH of freshly precipitated aluminium hydroxide, coagulation and destabilisation of particles is due to the coating of the inherently unstable aluminium hydroxide possibly resulting from ionisation of the precipitate surface or adsorption of sulfate anions. The presence of sulfate in solution can reduce significantly the positive charge of aluminium hydrolysis products. It has been shown that dissolved silica (silicic acid) can exert significant effect on coagulation with alum and ferric chloride w38x. The presence of dissolved silica can promote or inhibit coagulation of kaolin clay suspensions depending on concentration of silica and solution pH at a normal alum dosages. At pH 6, dissolved silica can promote alum coagulation greatly at some low concentrations, while at neutral or alkaline pH, a strong detrimental effect on coagulation was observed. This was found to be related to the effect of dissolved silica on aluminium precipitation w53x. If the presence of dissolved silica can promote aggregation of initial crystallites of aluminium hydroxide, it promotes coagulation of kaolin suspension by alum, and vice versa. 3.6. Temperature effects It is commonly found that hydrolysing metal coagulants perform less well at low temperatures. Temperature effects may be due to physical or chemical factors. Physically, water temperature may affect particle transport processes or particle collision rates, primarily through the effect on viscosity, and thus on the mixing energy dissipated in water. It is known that the orthokinetic collision rate greatly exceeds the perikinetic rate due to Brownian diffusion for particles whose size is greater than approximately 1 mm. In orthokinetic coagulation, where particle collisions are provided by fluid shear, it has been suggested that, as the viscosity

J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502 495

increases with decreasing temperature, the poor rapid-mixing conditions caused by low water temperature might lead to inhomogeneous distribution of coagulant species in the water, which results in a poor coagulation w54x. However, Hanson and Cleasby w55x concluded that the effects of temperature on coagulation could not be explained by the effect on parameters such as energy dissipation and turbulence microscale. Furthermore, much of the difference observed between alum coagulation at 20 and 5 8C is related to floc strength and not to turbulent flow field characteristics. It was found that the system chemistry was more important than the choice of energy input parameters at different water temperatures. Chemical influence on coagulation by hydrolysing metal salts due to variation of water temperature may be related to the effect on hydrolysis reactions, precipitation and solubility of the metal hydroxide. Kang and Cleasby w56x reported that, decreasing water temperature from 25 to 5 8C results in lowering the minimum solubility of Fe(OH)3 by 0.2 log unit and shifting it approximately 0.4 pH unit to the alkaline side. Dempsey w57x showed that, for a temperature change from 25 to 1 8C, the theoretical solubility minimum of Al(OH)3(s) shifts 0.6–0.8 pH units to the alkaline side and is lowered by approximately 0.7 log unit. Similar observations were reported by Driscoll and Letterman w58x and Hem and Roberson w59x. Water temperature may also affect rate of the metal ion hydrolysis reactions and establishment of equilibrium of solid phase with dissolved species in solution. With increasing temperature and pH, the rate of hydrolysis of Fe(III) salts is accelerated and the formation time of soluble polymeric iron species is reported to decrease rapidly w4,60x. In addition, the rate of approach to the equilibration concentration of aluminium hydroxide is significantly enhanced with increasing temperature w59x. However, Morris and Knocke w61x reported that rate of aluminium or iron(III) precipitation was not significantly affected over a temperature range of 1–23 8C. It has been observed that, at low temperature, optimal coagulation pH shifts to a higher value when using Fe(III) or Al(III) coagulants w5,55,62x. If pOH was kept constant, coagulation kinetics with ferric sulfate at 20 and 5 8C were nearly identical w55x. Van Benschoten and Edzwald w5x found that the pH at which Al precipitation occurs increased from 4.6 at 25 8C to 5.5 at 4 8C; and that the isoelectric point of Al precipitates shifted from pH 7 at 25 8C to pH 9 at 4 8C. It appeared that Al precipitates at 4 8C maintained a positive charge at higher pH than at 25 8C. It is possible that the temperature dependence of hydrolysis equilibria can also influence the species adsorbed on particles, and thus the coated particle surface properties. Gray et al. w14x suggested that slowing the rate of hydrolysis and precipitation reactions of metal coagulants in lower-temperature water is beneficial to some conditions, probably due to permitting hydrolysed species to react more extensively with particles. In contrast, Morris and Knocke w61x concluded that the effect of low temperature on coagulation efficiency in terms of turbidity removal was not related to reduced metal hydroxide precipitation rates. Instead, the poor turbidity removal at low temperature may be attributed to the floc characteristics. In cold water conditions, flocs are formed rather slowly and are smaller than those

496 J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502

formed at normal water temperature w61,63x. Hanson and Cleasby w55x found that both iron and alum flocs formed at 5 8C with kaolin clay, even at constant pOH, were much weaker that those at 20 8C. It was found that, under low-temperature conditions, Fe(III) coagulants can produce much better turbidity and colour removal than alum w61,64x. The better performance of Fe(III) over alum is believed to be due to the faster rate of precipitation with Fe(III) and the formation of larger flocs under low-temperature conditions. Haarhoff and Cleasby w65x found that FeCl3 is a better coagulant than alum for turbidity removal at low temperature in direct filtration at the same molar dosage of Fe3q and Al3q. Hanson and Cleasby w55x also observed that, at 5 8C, ferric sulfate coagulation of kaolin suspensions yielded better efficiency than alum, the iron floc being significantly stronger than the alum floc at temperatures 5 and 20 8C. Pre-hydrolysed products are also more effective than conventional coagulants at low temperatures (see below). 4. Pre-hydrolysed coagulants As well as traditional coagulants, based on Al and Fe salts, there are now many commercial products that contain pre-hydrolysed forms of the metals, mostly in the form of polynuclear species (see Section 2.2). In the case of Al, most materials are formed by the controlled neutralisation of aluminium chloride solutions and are generally known as polyaluminium chloride (PACl). It is believed that many of these products contain substantial proportions of the tridecamer Al13. Some information on the preparation of such materials and those based on iron is available w15x, but some important details are commercially sensitive and not easily found. In the case of aluminium sulfate, it is difficult to prepare pre-hydrolysed forms with high B values, because sulfate encourages hydroxide precipitation. The presence of small amounts of dissolved silica can substantially improve the stability up to B values of approximately 1.5 w66x. The resulting product is known as polyaluminosilicate sulfate (PASS). Pre-hydrolysed materials are often found to be considerably more effective than the traditional coagulants w67x. PACl products seem to give better coagulation than ‘alum’ at low temperatures and are also claimed to produce lower volumes of residual solids (sludge). Because they are already partially neutralised, they have a smaller effect on the pH of water and so reduce the need for pH correction. However, the mechanisms of action of PACl and similar products are still not well understood. Most explanations are in terms of the high charge associated with species such as Al13 and the consequent effectiveness in neutralising the negative charge of colloids in water. The relatively high stability of Al13 means that it should be more readily available for adsorption and charge neutralisation at around neutral pH, whereas conventional ‘alum’ undergoes rapid hydrolysis and precipitation. However, charge neutralisation cannot be the only mechanism of destabilisation, otherwise only ‘Zone 2’ coagulation would occur, which is less effective than sweep flocculation. It is still not clear what role hydroxide precipitation plays in the action

J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502 497

of pre-hydrolysed coagulants. Sulfate also plays an important role in precipitation with PACl w68x. It has been shown recently w42x that the volume of sediment produced in coagulation of clay suspensions by commercial PACl products is proportional to the coagulant dosage. This implies that some form of sweep flocculation is operating, since the volume of hydroxide precipitate would be expected to depend on the amount of coagulant added. From dynamic studies with a range of hydrolysing coagulants w42x it has been shown that PACl products give more rapid flocculation and stronger flocs than for ‘alum’ at equivalent dosages. However, in all cases, floc breakage was irreversible to some extent, as found previously for alum (Fig. 10). This is further evidence that the pre-hydrolysed products do not act simply by charge neutralisation and that some form of sweep flocculation is involved. It is very likely that the nature of the precipitate differs in the case of PACl materials, as has been shown for precipitation from solutions of Al13 (see Section 2.3), but a detailed understanding of flocculation mechanisms with these coagulants is still lacking. 5. Interaction with dissolved organic matter So far, we have only considered the removal of particles from water by hydrolysing coagulants. However, many natural waters contain dissolved organic substances, which also need to be removed. Natural organic matter (NOM) in water may impart undesirable colour to water and some constituents can form carcinogenic substances when water is chlorinated. NOM consists of a huge variety of organic compounds including simple sugars, amino acids, organic acids, proteins and many others. In most cases, so-called ‘humic substances’ are major components of aquatic NOM. This term covers a range of complex organic materials that exist in all soil and water environments and are thought to originate from decomposition of plant and animal remains. They are classified according to their aqueous solubility, with fulvic acids being more soluble than humic acids. Fulvic acids predominate in most waters and have lower molecular weight, typically in the region of 500–2000. Humic acids are reported to have a wide range of molecular weights, from a few thousand up to as high as 100 000, although the high values may be a result of aggregation. Essentially, humic substances can be regarded as natural anionic polyelectrolytes, of rather indeterminate structure. They have various functional groups, including carboxylic and phenolic, and a framework of randomly condensed aromatic rings. A hypothetical structure is shown in Fig. 11. Because of the ionisation of carboxyl groups, humic substances will have anionic charge at pH values higher than approximately 4 and are generally soluble under these conditions. It has long been known that humic substances can be effectively removed from water by hydrolysing coagulants and there have been many studies on this subject w69,70x. Humic substances adsorbed on aquatic particles can give enhanced stability and increased coagulant dosages may be needed w71x. In fact, for waters with high concentrations of organic matter, there is often a stoichiometric relation between the

498 J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502

Fig. 11. Hypothetical molecular structure of humic acid, showing important functional groups.

organic content (usually measured as dissolved organic carbon, DOC) and the required coagulant dosage w72x. Optimum pH for removal of dissolved organics is usually rather less (typically pH 5–6) than that for removal of suspended particles. When the coagulation process in water treatment is specially modified to ensure good removal of organic matter it is often known as enhanced coagulation. There are two likely mechanisms for the removal of humic substances by hydrolysing metal coagulants: ● Binding of metal species to anionic sites, thus neutralising their charge and giving a reduced solubility. For fairly large molecules, this can lead to precipitation of the metal–humic complex, to form particles that can be removed by sedimentation or filtration. ● Adsorption of humic substances on amorphous metal hydroxide precipitate. At pH values approximately 5–6, the humic substances are negatively charged and Al and Fe hydroxides are positively charged, which would give strong adsorption and some charge neutralisation. Pre-formed ferric floc has been shown to be a good adsorbent for humic substances w73x. In many practical cases it is not easy to distinguish between the precipitation and adsorption mechanisms. In a recent study w74x both were shown to operate, depending on pH and coagulant dosage. Humic substances isolated from lake sediments in Xi’an, China were coagulated with aluminium sulfate at different pH values and over a range of dosages. The removal was monitored by the reduction in UV absorbance (at 254 nm) after a standard coagulation and sedimentation procedure. The electrophoretic mobility of the destabilised particles was determined immediately after rapid mixing of the coagulant. In this case the EM values were converted to zeta potentials. The results in Fig. 12 show the residual UV absorbance and zeta potentials as functions of alum dosage at pH 5.0. The alum dosage is expressed on the basis of

J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502 499

Fig. 12. Removal of aquatic natural organic matter by coagulation with alum at pH 5, showing residual UV absorbance at 254 nm and zeta potentials of coagulated material. Alum dosages are shown relative to organic carbon content of the water sample (mg Alymg TOC).

humic substance concentration as mg Alymg of total organic carbon (TOC). There is a region of significant reduction in UV absorbance between approximately 0.1 and 0.2 mg Alymg TOC, which corresponds very well with the point at which the zeta potential reverses sign. This is strong evidence for a simple charge neutralisation mechanism under these conditions. At higher alum dosages there is another region where the absorbance is reduced, but this is not correlated with any significant change in zeta potential. The implication is that adsorption on a hydroxide precipitate is responsible for the removal. At pH 7, the results are quite different. Fig. 13 shows that there is an appreciable reduction in UV absorbance at approximately 0.2 mg Alymg TOC, which gradually improves at higher dosages. The zeta potential remains negative over the whole dosage range, but approaches zero at the highest dosages. It is very likely that the removal of humic substances under these conditions is entirely by adsorption on precipitated aluminium hydroxide. The reduction in absorbance is slightly greater at the higher pH, especially at the higher alum dosages. Pre-hydrolysed coagulants are also effective in removing dissolved organic matter from water, but in this case they often show no significant improvement over traditional coagulants w14x. It is known that polynuclear Al species, such as Al13, can be depolymerised in the presence of organic matter in natural waters w75x. 6. Conclusions Although the broad principles governing the action of hydrolysing coagulants are reasonably well understood, there are several important gaps in knowledge, which are of both fundamental and practical interest. For simple aluminium and ferric salts at low dosages, it is well established that charge neutralisation can be an effective means of destabilising colloidal particles. The precise nature of the cationic species is not known in detail, but it is likely that some form of surface precipitation is

500 J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502

Fig. 13. As for Fig. 12, but for coagulation at pH 7.

involved. Charge neutralisation can only be achieved at pH values below the effective i.e.p. of the metal hydroxide (approx. 8–9 in the case of Al). As the i.e.p. is approached, larger coagulant dosages are needed to achieve charge neutralisation. At excess coagulant dosage particle charge is reversed and restabilisation occurs and in many cases the range of effective coagulant dosages is quite narrow. When charge neutralisation is the only destabilisation mechanism, and for rather dilute suspensions, then the rate of coagulation can be low and relatively small aggregates are formed, giving poor removal efficiency. At higher coagulant dosages, bulk precipitation of metal hydroxide occurs, which can give large flocs of rather open structure. Impurity particles originally present in water become incorporated in these flocs and can be very effectively removed. This ‘sweep flocculation’ mechanism is generally more rapid than coagulation by charge neutralisation and gives larger flocs. There is often no clear correlation between particle charge and the onset of sweep flocculation. In some cases, there is a significant lag time between dosing the coagulant and the onset of appreciable floc growth, which is thought to be related to the time required for aggregation of the primary precipitate particles. This process is greatly dependent on the pH of the solution and the presence of certain anions. Flocs formed by hydroxide precipitates are rather weak and breakage is not fully reversible. Pre-hydrolysed coagulants are often more effective than simple Al and Fe salts. Part of the reason has to do with highly charged cationic species, such as Al13, which are rather stable and have a better opportunity to adsorb on negative colloids and neutralise their charge. However, at practical dosages of these coagulants, the solubility of the metal hydroxides is greatly exceeded and it is highly likely that precipitation plays an important role. The improved performance of these materials is probably due to the different nature of the precipitate formed, although more detailed studies are needed. As well as effectively removing colloidal particles, hydrolysing coagulants can also be used to remove dissolved natural organic matter from water. In this case, a

J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502 501

precipitation–coagulation mechanism can be important, in which anionic sites on the organic molecules are neutralised by hydrolysed metal species, so that the solubility is greatly reduced. Adsorption of organic matter on amorphous hydroxide precipitate can also be effective, although this generally requires higher coagulant dosages. The relative importance of these mechanisms depends primarily on pH. References w1x J.M. Cohen, S.A. Hannah, In Water Quality and Treatment, 3rd ed, McGraw Hill, New York, 1971, pp. 66–122. w2x D.T. Richens, The Chemistry of Aqua Ions, Wiley, Chichester, 1997. w3x D.J. Wesolowski, D.A. Palmer, Geochim. Cosmochim. Acta 58 (1994) 2947. w4x C.M. Flynn, Chem. Rev. 84 (1984) 31. w5x J.E. Van Benschoten, J.K. Edzwald, Water Res. 24 (1990) 1519. w6x R.B. Martin, J. Inorg. Biochem. 44 (1991) 141. w7x J.W. Akitt, N.N. Greenwood, B.L. Khandelwal, G.D. Lester, J. Chem. Soc. Dalton Trans. (1972) 604. w8x J.Y. Bottero, D. Tchoubar, J.M. Cases, F. Fiessinger, J. Phys. Chem. 86 (1982) 3667. w9x L.-O. Ohman, W. Forsling, Acta Chem. Scand. A 35 (1981) 795. w10x T. Wagatsuma, M. Kaneko, Soil Sci. Plant Nutr. 33 (1987). w11x E. Matijevic, K.G. Mathai, R.H. Ottewill, M. Kerker, J. Phys. Chem. 65 (1961) 826. w12x H. De Hek, R.J. Stol, P.L. De Bruyn, J. Colloid Interface Sci. 64 (1978) 72. w13x R.W. Smith, ACS Adv. Chem. Ser. 106 (1971) 250. w14x K.A. Gray, C.H. Yao, C.R. O’Melia, J. Am. Water Works Assoc. 87 (1995) 136. w15x H.X. Tang, Z.K. Luan, in: H.H. Hahn, E. Hoffmann, H. Odegaard (Eds.), Chemical Water and Wastewater Treatment IV, 1996, pp. 83–93. w16x H.X. Tang, W. Stumm, Water Res. 21 (1987) 115. w17x P.J. Murphy, A.M. Posner, J.P. Quirk, J. Colloid Interface Sci. 56 (1976) 284. w18x W. Schneider, B. Schwyn, in: W. Stumm (Ed.), Aquatic Surface Chemistry, Wiley, New York, 1987, pp. 167–196. w19x D. Tchoubar, J.Y. Bottero, P. Quienne, M. Arnaud, Langmuir 7 (1991) 398. w20x J.Y. Bottero, A. Manceau, F. Villieras, D. Tchoubar, Langmuir 10 (1994) 316. w21x Y. Deng, Water Res. 31 (1997) 1347. w22x P.M. Bertsch, in: G. Sposito (Ed.), The Environmental Chemistry of Aluminum, CRC Press, Boca Raton, 1989, pp. 117–168. w23x S.M. Bradley, R.A. Kydd, R.F. Howe, J. Colloid Interface Sci. 159 (1993) 405. w24x J.Y. Bottero, M. Axelos, D. Tchoubar, J.M. Cases, J.J. Fripiat, F. Fiessinger, J. Colloid Interface Sci. 117 (1987) 47. w25x L.-O. Ohman, L. Wagberg, J. Pulp Pap. Sci. 23 (1997) J475. w26x M.M. Clark, R.M. Srivastava, R. David, Environ. Sci. Technol. 27 (1993) 2181. w27x J.Q. Jiang, N.J.D. Graham, Water Res. 32 (1998) 930. w28x P.L. Hayden, A.J. Rubin, in: A.J. Rubin (Ed.), Aqueous-Environmental Chemistry of Metals, Ann Arbor Science Publishers, Ann Arbor, 1974, p. 180. w29x K.A. Hunter, P.S. Liss, Limnol Oceanogr. 27 (1982) 322. w30x J. Gregory, in: C.A. Finch (Ed.), Ind. Water Soluble Polym., Royal Society of Chemistry, Cambridge, 1996, pp. 62–75. w31x E. Matijevic, J. Colloid Interface Sci. 43 (1973) 217. w32x W. Stumm, Chemistry of the Solid-Water Interface, Wiley Interscience, New York, 1992. w33x R.D. Letterman, S.G. Vanderbrook, P. Sricharoenchaikit, J. Am. Water Works Assoc. 74 (1982) 44. w34x R.O. James, T.W. Healy, J. Colloid Interface Sci. 40 (1972) 53.

502 J. Duan, J. Gregory / Advances in Colloid and Interface Science 100 – 102 (2003) 475–502 w35x w36x w37x w38x w39x w40x w41x w42x w43x w44x w45x w46x w47x w48x w49x w50x w51x w52x w53x w54x w55x w56x w57x

w58x w59x w60x w61x w62x w63x w64x w65x w66x w67x w68x w69x w70x w71x w72x w73x w74x w75x

K.J. Farley, D.A. Dzombak, F.M.M. Morel, J. Colloid Interface Sci. 106 (1985) 226. S.K. Dentel, Crit. Rev. Environ. Control 21 (1991) 41. S.K. Dentel, J.M. Gossett, J. Am. Water Works Assoc. 80 (1988) 187. J. Duan, Influence of Dissolved Silica on Flocculation of Clay Suspensions with Hydrolysing Metal Salts. PhD Thesis, University of London. 1997. R.D. Letterman, S.G. Vanderbrook, Water Res. 17 (1983) 195. W. Stumm, C.R. O’Melia, J. Am. Water Works Assoc. 60 (1968) 514. R.F. Packham, J. Colloid Interface Sci. 20 (1965) 81. J. Gregory, V. Dupont, Water Sci. Technol. 44 (2001) 231. J. Gregory, D.W. Nelson, Colloids Surf. 18 (1986) 175. M. Elimelech, J. Gregory, X. Jia, R.A. Williams, Particle Deposition and Aggregation. Measurement, Modelling and Simulation, Chapter 6, Butterworth-Heinemann, Oxford, 1995. T. Matsuo, H. Unno, J. Environ. Eng., ASCE 107 (1981) 527. L.B. Brakalov, Chem. Eng. Sci. 42 (1987) 2373. ¨ K. Muhle, in: B. Dobias (Ed.), Coagulation and Flocculation, Decker, New York, 1993, pp. 355–390. R.J. Francis, Water Res. 21 (1987) 1023. M.A. Yukselen, J. Gregory, Environmental Engineering Science, In press (2002). S.P. Marion, A.W. Thomas, J. Coll. Sci. 1 (1946) 221. R.D. Letterman, M. Tabatabaie, R.S. Ames, J. Am. Water Works Assoc. 71 (1979) 467. G.P. Hanna, A.J. Rubin, J. Am. Water Works Assoc. 62 (1970) 315. J. Duan, J. Gregory, J. Inorg. Biochem. 69 (1998) 193. E.A. Vik, B. Eikebrokk, ACS Symp. Ser. 219 (1989) 385. A.T. Hanson, J.L. Cleasby, J. Am. Water Works Assoc. 82 (1990) 56. L.S. Kang, J.L. Cleasby, J. Env. Eng. Sci. ASCE 121 (1995) 893. B.A. Dempsey, in: Seminar Proceedings: Influence of Coagulation on the Selection, Operation and Performance of Water Treatment Facilities, American Water Works Association, Denver, 1987, pp. 19–30. C.T. Driscoll, R.D. Letterman, J. Environ. Eng., ASCE 114 (1988) 21. J.D. Hem, C.E. Roberson, ACS Symp. Ser. 416 (1990) 429. J.H.A. Van der Woude, P.L. De Bruyn, Colloids Surf. 8 (1983) 55. J.K. Morris, W.R. Knocke, J. Am. Water Works Assoc. 76 (1984) 74. T.R. Camp, B.A. Root, B.V. Bhoota, J. Am. Water Works Assoc. 32 (1940) 1913. W. Hutchison, P.D. Foley, J. Am. Water Works Assoc. 66 (1974) 79. W.R. Knocke, S. West, R.C. Hoehn, J. Am. Water Works Assoc. 78 (1986) 189. J. Haarhoff, J.L. Cleasby, J. Am. Water Works Assoc. 80 (1988) 168. J.-P. Boisvert, C. Jolicoeur, Colloids Surf. A 155 (1999) 161. J.Q. Jiang, N. Graham, Chem. Br. 34 (1998) 38. D.S. Wang, H.X. Tang, J. Gregory, Environ. Sci. Technol. 36 (2002) 1815. E.S. Hall, R.F. Packham, J. Am. Water Works Assoc. 57 (1965) 1149. G.A. Edwards, A. Amirtharajah, J. Am. Water Works Assoc. 77 (1985) 50. M.R. Jekel, Water Res. 20 (1986) 1543. H. Bernhardt, H. Schell, J Water SRT-Aqua 42 (1993) 239. G. Cathalifaud, M.T.W. Mossa, M. Mazet, Water Sci. Technol. 27 (1993) 55. X.C. Wang, P.K. Jin, J. Gregory, Water Sci. Tech., In press (2002). A. Masion, A. Vilge-Ritter, J. Rose, W.E.E. Stone, B.J. Teppen, D. Rybacki, J.Y. Bottero, Environ. Sci. Technol. 34 (2000) 3242.