3.1 Ultrafiltration process

Ultrafiltration is based on the use of porous membranes with a MWCO in the range of 1–300 kDa (pore sizes between ∼2 and 100 nm) and requires the application of transmembrane pressure differences of between 3 and 10 bars. UF is widely used to concentrate solutions of macromolecules by only letting the molecules of solvent and solutes of low MW through the membrane. Steric effects (dependent on the relative sizes of the pores and the molecular species) are central to the mechanisms of exclusion by the UF membrane. However, the effects of electric charge resulting from the chemical properties of the membrane material when the molecular species in solution are charged must also be taken into account. Indeed, most membranes, whether organic or inorganic acquire an electrical surface charge when brought into contact with a polar medium such as an aqueous solution. The origin of this surface charge can be the acidic or basic properties of the surface moieties, this is the case of organic membranes bearing, for instance, amide, carboxyl or sulphone groups, or amphoteric properties of the surface sites including hydroxyl functions at the surface of the metal oxides used to form ceramic membranes (Szymczyk and Fievet, 2007; Larchet and Pontié, 2008). These interactions can be put to good use to separate species of similar size but with different charges.

UF has become the norm in many industrial sectors such as water treatment (clarification–disinfection of drinking water) (Maurel, 1989), depollution of waste water by membrane bioreactors (Van Kaam et al., 2006; Hernandez, 2006), the dairy industry (standardisation of the level of proteins in milk (Nyström et al., 1998), concentration of whey proteins (Glover, 1985)), but also in the automotive industry (treatment of paint baths by electrophoresis (Maurel, 1989), treatment of degreasing baths (Li et al., 2006), destabilisation of machining oil emulsions). Note that applications involving the separation or concentration of ionic species are lacking – this can be understood by considering the size of the pores in UF membranes. However, the range of utilisation of UF can be enlarged towards solutes smaller than the pores in the membrane by means of pretreatments such as micellisation (Yurlova et al., 2002) or complexation (Kryvoruchko et al., 2002; Mimoune et al., 2007).

3.2 Complexation–ultrafiltration processes

The notion of complexation–ultrafiltration was first introduced in 1968 by Michaels (1968a,b). As an UF membrane stops macromolecules but lets ions through easily, the novel idea was to “artificially” increase the size of the ions by complexing them onto a polymer that would be retained by the membrane. Thus, by adding macromolecular chains bearing ligand functions (called macroligands) to a metal-containing solution, the product of UF is a concentrated solution of complexed metal ions in the retentate and a metal-free permeate. This assumes that the macroligand–metal ion complexes are stable and rejected by the membrane, since cation rejection assumes not only rejection of the polymer but also stability of the complex. This technique appears under different names: PAUF, polymer or polyelectrolyte enhanced ultrafiltration (PEUF), polymer supported ultrafiltration (PSU), liquid-phase polymer-based retention (LPR), enhanced ultrafiltration, or simply polymer filtration (PF) (Geckeler et al., 1988; Canizares et al., 2002; Palencia et al., 2009; Rivas et al., 2009, 2011).

The main advantages of this hybrid process are that (i) the installations are compact with low operating costs since the energy consumption of the process is reduced compared to that of the alternatives (i.e. evaporation, NF, etc.), (ii) separation is efficient and selectivity good, (iii) the complexing agents are chosen so as to be easily regenerated. Table 1 reports other advantages and the main disadvantages. The performance of the double technique complexation–ultrafiltration depends on the rejection of both the macromolecules and the metal ions. In general, for a given MW, the retention of the flexible macromolecules in solution decreases as the pressure is increased. This can be explained by the fact that when the macromolecules are subjected to a transversal pressure gradient in the vicinity of the pore entrances, they are stretched out along the direction of flow and can then enter the pores. Sudareva et al. (1991) suggested that the existence of different conformations and/or the effect of the polarisation layer would contribute to this effect. Rejection is therefore not simply dependent on MW. Staub et al. (1984) showed that the charge of the macromolecule strongly influences rejection and the closer the size of the macromolecule to the size of the pore, the greater the influence of the charge. Concerning metal ion binding, the stability of the complexes (at equilibrium) is closely linked to various parameters such as pH, ligand concentration, ionic strength of the solution (presence of salts or other ligands) and temperature. In particular, the pH is important since its increase can lead to the formation of metal hydroxides and this reaction will be in competition with the complexation of the cations by the polymer. In addition, the complexing properties of the polyelectrolytes depend on the pH. In general, an excess of ligand is used with respect to the metal cation since this improves binding. The limiting factors here are then the viscosity of the solution and the cost of the process as a whole.

3.3 Polymers used in PAUF processes

There are three main categories of water-soluble polymer able to form tight complexes with metal cations, namely (i) polyelectrolytes with amine functions [poly(ethyleneimine) (PEI), poly(allylamine) (PAA)], carboxyl functions [poly(acrylic acid) (PA), carboxymethylcellulose (CMC)] or hydroxyl functions [poly(vinylalcohol) (PVA_l)] (Barron-Zambrano et al., 2002; Pastor et al., 2002; Islamoglu and Yilmaz, 2006; Petrov et al., 2002; Kryvoruchko et al., 2004; Molinari et al., 2004a,b, 2006a,b, 2008; Kuncoro et al., 2005; Cojocaru and Zakrzewska-Trznadel, 2007), (ii) polyelectrolytes modified by complexing ligands [modified PEI, diethylaminoethylcellulose, etc.] (Trivunac and Stevanovic, 2006) and (iii) synthetic copolymers obtained by copolymerisation of vinyl monomers (Juang and Chiou, 2000a,b; Molinari et al., 2004b, 2006b). Additional novel natural macroligands such as pectin and chitosan have appeared in the last few years (Table 2).

3.4 Why use chitosan as polyelectrolyte?

The majority of commercial water-soluble polymers are derived from petroleum-based raw materials synthesised using processing chemistry that is not always safe or environmentally friendly. Today, there is growing interest in developing natural low-cost alternatives to synthetic polymers. In addition, in a global context of seeking to reduce the mass of waste generated, recycling of manufacturing by-products is one of today’s priorities in sustainable development.

Chitosan is obtained from chitin, a natural waste from shellfish (e.g. shrimp, crab, squid, lobster shells). It is cheap, abundant and ecologically relevant (Crini and Badot, 2008; Crini et al., 2009). Fig. 1 gives the stereochemical structure of commercial chitosan. Chitosan presents multiple physical–chemical and polyelectrolytic properties at acid pH (Table 3). Use is made of all these properties in PAUF.

Close

Figure 1 Chemical structure of chitin [poly(N-acetyl-β-d-glucosamine)], chitosan [poly(d-glucosamine)] and commercial chitosan (a copolymer characterised by its average degree of acetylation (DA)).

Export to PPT

Each commercial batch of chitosan is characterised by its molecular weight (MW), degree of acetylation (DA) or degree of deacetylation (DD), and crystallinity. As with PEI, the most suitable polymer is chosen for each separation. The structure of chitosan presents amine and acetamide moieties (Fig. 1) as well as, of course, numerous hydroxyl groups conferring a strong hydrophilic character. In acid media, the protonation of the amine functions makes the polymer soluble and thus it behaves as a polyelectrolyte (pKa ∼6.3 depending on the DD). Note that it is the only cationic polymer known among all biopolymers. The advantages of using chitosan in its dissolved form arise from the availability and accessibility of the amine functions. Dissolving the polymer leads to the appearance of repulsion forces between the macromolecular chains, allowing them to unravel and make the internal sites accessible (Juang and Chiou, 2000a; Guibal, 2004; Crini and Badot, 2008). The intermolecular hydrogen bonding that occurs in the solid state is repressed in solution: this also enhances the availability of the internal amine moieties for interaction with metal cations.

The affinity of chitosan for transition metal ions has been widely studied, and examples of selected affinity series are reported in Table 4. Irrespective of the type of material used, the results can mainly be explained by the interactions occurring between the amine functions of the chitosan and the metal ions. For instance, Krajewska (2001), studying the diffusive permeability coefficients of 15 metal ions through chitosan membranes, showed that the rank order of the coefficients (Cu²⁺ < Ni²⁺ < Zn²⁺ < Mn²⁺ < Pb²⁺ < Co²⁺ < Cd²⁺ < Ag⁺) corresponds to the metal ion–polymer affinity sequence, confirming chelation of these ions by chitosan. Kaminski et al. (2008), studying interactions of Cu²⁺, Zn²⁺ and Cr(VI) adsorbed onto chitosan beads, reported that at low concentrations the sequence of adsorption was Cu²⁺ > Zn²⁺ > Cr(VI), confirming that chitosan exhibited a strong selectivity towards copper, but at high concentrations the series of metal selectivity was different (Zn²⁺ > Cu²⁺ > Cr(VI)). Similar results were previously reported by Vold et al. (2003).

However, some issues do arise with the use of chitosan. In a PAUF process, its use requires addition of acid to enable complete dissolution in water because a minimum amount of −NH₃⁺ groups is needed to prevent the formation of particles and/or flocculation. Thus, it is important to find the practical conditions needed to prepare solutions. Moreover, a further disadvantage is related to the difficulty of finding reproducible sources of chitosan, with a suitable MW (depending on the membrane MWCO).

4.1 Free metal cations

Chitosan-based PAUF is currently in rapid expansion. The three main advantages often mentioned are (i) the good chelating capacity of the chitosan chains for numerous metal ions, especially those used in the surface treatment industry (the capacity is due to its high charge density) (ii) the excellent selectivity of the separations (Volchek et al., 1993; Taha et al., 1996; Chauffer and Deratani, 1998; Juang and Chiou, 2001; Rivas et al., 2003; Aroua et al., 2007), and (iii) the ability of the biopolymer to be regenerated without suffering degradation (Llorens et al., 2004).

Most transition metals can form complexes with chitosan albeit with differing binding strengths. Chitosan can separate numerous metals (Cu²⁺, Ni²⁺, Zn²⁺, Mn²⁺, Pb²⁺, Cd²⁺, Co²⁺, Hg²⁺, etc.) or recover them selectively from complex mixtures. Only alkali metals and alkali earths have no affinity for chitosan (Krajewska, 2001). This property is particularly useful for industrial applications involving decontamination of polycontaminated mixtures. It can also be noted that the selectivity can be related to the binding mechanisms and thus to the pH of the medium.

Not only must the pH be taken into account but also other parameters such as the quantity of amine present on the macromolecular chains and thus DD, MW and crystallinity. DD and MW are important characteristics since they control a lot of the properties of chitosan including its behaviour in water and hence its binding capacity. The degree of crystallinity controls the properties of hydration and swelling of the chitosan in solution which in turn influence the diffusion properties (Guibal, 2004; Crini and Badot, 2008). Chitosan is often considered as a flexible linear cationic copolymer with a higher binding capacity than PEI. However, the flexibility is dependent on the conformation of the macromolecule in solution. Likewise, the conformation and thus also the viscosity of the solutions also depend on the DD and on the pH which both alter the charge density of the polymer chain. When the charge density is high, the macromolecules are unrolled (owing to charge repulsion phenomena) and the viscosity is high. In contrast, when the polymer charge is low, the macromolecule chains are rolled up into statistical bundles which implies low viscosity (owing to lack of chain interaction phenomena). Hence, the lower the pH, the higher the viscosity. Viscosity also increases with DD as the more the polymer is deacetylated, the more the free amine groups can become protonated and the greater its solubility. Viscosity rises too with concentration and it is also dependent on the molecular weight of the polymer chains, but it falls as the temperature rises.

Generally high-molecular-weight biopolymers are preferred as they prevent losses of material in the form of low MW chains but also because they enable the use of UF membranes with higher MWCO values which leads to higher permeate flow rates (Juang and Chiou, 2001; Tabatabai et al., 1995). The main advantage of chitosan nevertheless remains its high binding capacity – but even this point has been a subject of debate. Indeed, while several studies report the good retention of bivalent metal cations by chitosan (Juang and Chiou, 2000a,b; Llorens et al. 2004; Bal et al., 2006; Verbych et al., 2006; Miretzky and Fernandez, 2009), the binding constant of chitosan with metal cations remains low. We can recall that its affinity for metal cations is variable and particularly dependent on the pH.

Juang and Chiou (2000b) demonstrated the selectivity of chitosan for Cu²⁺ in a Cu²⁺, Co²⁺, Ni²⁺ and Zn²⁺ mixture. The presence of chitosan at a concentration twice as high as that of Cu²⁺ improves copper retention 6- to 10-fold at pH < 6 (pH of Cu²⁺ precipitation). On filtration of a solution of Cu²⁺ in the presence of 2 eq. of chitosan, metal retention is seen to be around 50–75% in slightly acid medium (pH 3–6) and reaches 100% for pH values over 8 (Juang and Chiou, 2000b). These authors also reported the ability of chitosan to separate Cu²⁺ and Zn²⁺ ions unlike PEI. In another article, the same authors Juang and Chiou (2001) showed that chitosan is much more efficient than PEI at extracting and concentrating the ions Na⁺, K⁺, Ca²⁺, Mg²⁺, Cl⁻, NO₃⁻ and SO₄²⁻.

It is therefore in alkaline medium that divalent metal-ion containing solutions are best processed by metal complexation with the polymer and precipitation of the metal hydroxides and secondly optimum permeate fluxes. The downside is that chitosan is hydrolysed in alkaline media. Ni²⁺ was found to present characteristics similar to Cu²⁺ whereas for Zn²⁺, retention only became total at around pH 10 owing to the formation of hydroxides. The presence of the cations Na⁺ and Ca²⁺, even at high concentrations, had very little influence on Cu²⁺ uptake by chitosan in acid media. This suggests that the metal–polymer interaction likely takes the form of a coordination bond. For instance Verbych et al. (2006) showed that the introduction of NaCl into the solution did not influence copper uptake but that high concentrations of CaCl₂ can hinder chitosan-metal exchanges. Moreover, increasing the salinity causes a sharp drop in permeate flux, probably due to a reduction in the electrostatic repulsion between polycations.

Taha et al. (1996), Juang and Chiou (2000a,b), Llorens et al. (2004) and Kuncoro et al. (2005) demonstrated that chitosan’s efficiency varies with its concentration (there are also contradictory reports on this point) and the pH of the medium. The viscosity of chitosan solutions can affect their UF performance as they alter the hydrodynamics of the system. Kuncoro et al. (2005) reported that the performance of the process mainly depends on the pH of the solution. They showed that recovery of mercury increases with the pH to stabilise at about pH 5.5. Mercury uptake was also found to increase with the excess of polymer before it reached a plateau over a molar ratio of 5–7 (amine/mercury). The excess does not have to be too large before binding is optimal. Problems of fouling also occur in conditions that favour aggregation (Juang and Chiou, 2001) or when the polymer starts to degrade in basic medium (Juang and Chiou, 2000a,b).

The studies of Verbych et al. (2006) indicate that chitosan/metal complexes are very stable and can easily form deposits at the surface of membranes and in the pores. In addition, it is not necessary to use high concentrations of polymer. The authors demonstrate the high exchange capacity of chitosan in neutral medium. Similar results had been published earlier for Cd²⁺ (Llorens et al., 2004).

Two metal cations that bind very easily to chitosan are Cu²⁺ and Hg²⁺ (Guibal, 2004). However, the binding/uptake capacity also varies with the authors. For Hg²⁺, Cu²⁺ and UO₂²⁺ a binding capacity of up to 2 mmol metal g⁻¹ has been reported. It is 0.5–1 mmol metal g⁻¹ for ions Pb²⁺, Cd²⁺ and Zn²⁺ and under 0.5 mmol metal g⁻¹ for Ni²⁺ and Co²⁺. For metal anions (e.g. vanadate or molybdate), the uptake is higher, reaching 7–8 mmol metal g⁻¹ (Guibal, 2004). These differences in uptake capacity suggest that in complex mixtures some ions will be retained preferentially. The process will be dependent on the pH but in general will not depend on the ionic strength of the solution, unlike with PEI.

Monovalent cations, especially ions of alkali metals, generally show a low binding affinity for chitosan. Trivalent ions are particularly well eliminated since they generally hydrolyse at acid pH. They also have a higher affinity for oxygen-rich polymers, such as pectin, which are better chelators (Aroua et al., 2007). Uptake of gold III by cheap biopolymers such as chitosan and sericin has been studied (Chen et al., 2011), and it is sericin, which presents amide functions, that proved to be the more efficient.

All these data on chelation can be mainly explained by the behaviour of chitosan in solution. As with other water-soluble polymers, its complexing capacity for numerous metals is higher when the polymers are dissolved because the macromolecular chains are better deployed and thus their surface of interaction increases (Guibal, 2004). Metal uptake increases with the polymer concentration following the law of mass action. However, increasing the chitosan/metal ratio rapidly affects the flow of permeate in acid and neutral media, following a hyperbolic law in acid medium and extremely abruptly at pH 6 (Verbych et al., 2005, 2006). In general, the polymer concentration increases exponentially in the vicinity of the membrane/solution interface over a thickness of a few tens of microns (concentration polarisation phenomenon). A simple 2-phase model provided a satisfactory description of the experimentally observed rejection (Llorens et al., 2004).

In most cases, the interaction of metal cations with a polymer (especially with a polycationic polymer like chitosan) does not lead to the formation of flocs and sedimentation, the metals thus remain in suspension. Ultrafiltration thus offers a physical barrier suited to the size of the polymers, enabling filtration at lower pressures and with higher permeate flows than nanofiltration. This also offers the possibility to recycle the polymer, releasing the metals in acid medium to then precipitate them by electrolysis. Another advantage that is often mentioned when using chitosan concerns this regeneration step (Llorens et al., 2004; Juang and Chiou, 2000a). Once complexed on the chitosan chains, the metals can be easily “decomplexed” after their separation. The metals can then be concentrated justifying their recovery by conventional processes (e.g. precipitation) which cannot be economically applied to the initial low-concentration solutions. Rivas et al. (2003) reported that the pollutants can be concentrated 100-fold. After elimination of the metal from the concentrate, the polymer can be regenerated to its active form by addition of base. This technique also enables the separation of metals of the same valency on the basis of the chemical affinity for the polymer. It can therefore compete with electrodialysis or ion exchange resins, especially when the solution to be treated requires a disinfection step. Molinari et al. (2008) reported the successful separation of Cu²⁺ and Ni²⁺ ions by UF in the presence of PEI while Verbych et al. (2006) described the separation of Cu²⁺ and Ca²⁺ ions using chitosan. This chemical selectivity is complementary to that of NF (selectivity achieved by charge effects).

For metal cations, regeneration is often performed by acidification or electrolysis, which reverses the complexation of the metal. The risk is the degradation of the polymer although this point is still unclear. While Llorens et al. (2004) reported easy regeneration of the polymer without degradation, the ability to undergo numerous regeneration cycles was a critical point of the process (especially reduction of the polymer’s molecular weight) for other authors (Volchek et al., 1993; Geckeler and Volchek, 1996; Juang and Chiou, 2001). Some metals such as palladium (Guibal, 2004) and cadmium (Llorens et al., 2004) are more easily desorbed than others such as platinum. Optimising the binding and regenerations steps can also significantly enrich the solution with respect to one of the metals (Llorens et al., 2004).

Modification of the chitosan chains is also possible and has been proposed to improve the capture of cations (Emara et al., 2011). The chelation properties (binding and selectivity) can indeed be improved by chemically altering the polymer, grafting specific functional groups onto the amine functions. Chitosan is a highly reactive polyamine and its chemical alteration can (i) increase the number of chelation/complexation sites (and hence the uptake capacity of the polymer chains), (ii) improve the intrinsic binding selectivity by grafting functions that can change the mechanism of binding, broadening the range of action of the polymer by shifting its workable pH range. Numerous ligands can thus be grafted onto the glucose units of chitosan such as derivatives with carboxylic acid functions, phosphates, or sulphur-containing moieties (Varma et al., 2004; Wu et al., 2010; Emara et al., 2011). For instance, the selective recovery of precious metals or of very toxic substances requires a modification of the amine functions of chitosan giving rise to elaborated structures with easily accessible sites (Arrascue et al., 2003; Chang and Chen, 2006).

The choice of the MWCO threshold of the membrane is extremely important, and severe permeate flux reductions are observed for higher MWCO thresholds which can allow a significant fraction of the polymer to enter the pores. It has been reported that the most suitable MWCOs are situated around 10–30 kDa (Verbych et al., 2005, 2006).

The same group (Verbych et al., 2005) studied the effect of humic and fulvic acids in water containing divalent cations during UF treatment. They noted that the presence of humic acids did not reduce the rejection of metal cations and that it increased their rejection when the concentrations of chitosan and humic acid increased together. However, increasing the concentrations of the macromolecules (chitosan and humic acid) leads to a sharp reduction of the permeate flow due to the formation of chitosan-humic acid complexes, with accumulation on the membrane leading to charge neutralisation and concentration polarisation.

In conclusion, numerous metal cations can bind to chitosan, although the affinity varies. However, the fact that the formation constants are relatively weak is an advantage, facilitating polymer regeneration after complexation (Llorens et al., 2004). In some cases, chemical modification of the polymer is necessary, especially to increase the binding capacity or the selectivity. The modifications can also obviate the need for chemical reagents to release the concentrated metal. To avoid serious and rapid loss of permeate flow, it is important to be able to run the filtration at a low trans-membrane pressure and high tangential velocity. Filtration is more efficient in alkaline media but there is a risk of chitosan degradation. In spite of the advantages that chitosan presents for the chelation of metals, it has not yet been used at an industrial scale. The following reasons for this were given by Guibal (2004). Firstly, the chemical stability of the polymer is low when in solution. It is easily degraded by microorganisms (bacteria, fungi) which can also lead to biological contamination of the solution being processed. In addition, it becomes hydrolysed in strongly acid or alkaline media, decreasing its mean molecular mass and adversely affecting its rejection by the PAUF membrane. Currently, the absence of large units of production and the lack of standardisation of the production processes lead to a high variability in the properties of the polymer obtained. Finally, the raw material is still quite costly, 10–15 Euros/kg for environmental applications, owing to the very limited market. Moreover, when the affinity of the metal cation for the chitosan remains moderate, as is the case for Cd²⁺, competition with the proton is not favourable and the cost of acid and base to regenerate the polymer remain dissuasively high (Sabaté et al., 2006).

4.2 Anionic species

As a polycation, at pH ⩽ 6 chitosan is associated to polyanions including humic and fulvic acids (Verbych et al., 2005), tartrate, citrate and polyaminocarboxylates such as ethylenediaminetetracetic acid (EDTA) (Gyliené et al., 2006, 2008). The stability of these complexes is strongly dependent on the polymer concentration and the pH. The interactions involve electrostatic attraction but the presence of numerous hydrogen bonds can also stabilise the associations. Metals with a high valency, which become hydrolysed when present as anions, can be captured by the chitosan in acid medium. Aroua et al. (2007) reported for instance that at acid pH filtration with a PES membrane led to a 70% abatement of Cr(VI) with chitosan, and almost 100% elimination with PEI.

4.3 Metal-based complexes

Several associations of anionic complexes with chitosan have been reported. For instance UF of As(V) in water that contains humic acids at pH 7.6 showed extensive fouling of the cellulose membrane due to the formation of a gel, which is greatly reduced upon addition of chitosan (Lin et al., 2008). It appeared that the humic acids bind preferentially to chitosan and are then more soluble. At pH 7.6 the chitosan is unable to take up anionic As(V) directly but becomes associated to the arsenic-humic acid complex. In addition, chitosan is able to retain numerous anionic complexes in which the metal cation is stabilised by a ligand. Thus, the Cu-citrate (Guzman et al., 2003) and the Cu-EDTA (Gyliené et al., 2006) complexes bind to chitosan in slightly acid medium. The driving forces underlying the binding are electrostatic interactions between polyions of opposite charges, direct interactions between the metal and the chitosan, and numerous hydrogen bonds. These observations were confirmed by Zamariotto et al. (2010) with PEI on copper and nickel chelated with nitrilotriacetic acid (NTA) and EDTA. An apparent rejection rate of 98% was noted between pH 4 and 9. However, the more the metal is enveloped in the chelating agent, the less it is retained by the membrane. This is because the number of coordination bonds between the polymer and the metal decreases. The excess of free anionic ligand bound by the polycation is little by little displaced by the metal, which forms a more stable coordination complex (Zamariotto et al., 2010).

4.4 Mechanisms

In UF processes, metal cations cannot be retained by the membrane since the average ion size is smaller than the membrane pore diameter. With the PAUF process, separation is dependent on the interaction strength between metal ions and the polymer functional groups. When these interactions are strong and when the complex generated is sufficiently large, it is retained by the membrane. Those metal ions that are not fixed to polymer cross through the membrane. Data on the general principles and fundamental concepts of the PAUF process using different water-soluble functional polymers have been discussed in detail by Rivas et al. (2011).

Chitosan-based PAUF also associates the two mechanisms for the recovery of metals, i.e. complexation in solution and separation by the membrane (Verbych et al., 2006; Aroua et al., 2007). The water-soluble chitosan polymer and metal ion solution are contacted on the feed-side in the filtration system. The macromolecular chains interact with metal ions which are bound to the polymer resulting in polymer–metal chains, which are then retained by a size exclusion mechanism; whereas unbound species, smaller than the pores, pass through the membrane into the permeate stream. The principle of the process is to increase the size of the target solute by complexation with the macromolecular chains of chitosan. This step corresponds to the binding of the metal onto the macromolecular chain in specific conditions. The reaction is pushed to the right owing to the high affinity of metal ions for chitosan.

Two mechanisms are clearly established for the interpretation of metal complexation on chitosan, i.e. electrostatic interactions in acid media and metal chelation (coordination), although the formation of ion pairs has also been reported (Guibal, 2004; Crini and Badot, 2008). Metal ion adsorption is assumed to occur through single or mixed mechanisms including coordination on amino groups in a pendant fashion or in combination with vicinal hydroxyl groups, and ion-exchange with protonated amino groups through proton exchange or anion exchange, the counter ion being exchanged with the metal anion. Other interactions are also involved such as hydrogen bonding, hydrophobic interactions and physical interactions due to the structure of the network of macromolecular chains.

The nature of the reaction depends upon several parameters related to the polymer (ionic charge, DA, MW, chemical modification), to the solution (pH, ionic strength, polymer dose, metal concentration) and to the chemistry of the metal ion (ionic charge, ability to be hydrolysed and to form polynuclear species).

In spite of the numerous studies focusing on the uptake of metal ions by chitosan, there is still debate as to the actual mechanisms involved (Monteiro and Airoldi, 1999; Rhazi et al., 2002; Llorens et al., 2004; Debbaudt et al., 2004; Verbych et al., 2006). Depending on the interactions between the metal species and the chitosan, processes observed include rejection of metal anions (e.g. Cr(VI), As(V)) in acid medium involving the polycation and rejection of metal cations at pH ⩾ 3. As UF is also able to eliminate colloidal metal hydroxides very efficiently (Juang and Chiou, 2000a,b), sorting out exactly what happens in neutral and alkaline media with respect to the chitosan is often difficult and few studies provide enlightenment on this subject (Llorens et al., 2004).

Studying the chelation of copper by chitosan, Domard (1987) was the first to propose a mechanism known as the “pendent model” which considers that the metal is bound to an amine function just like a pendent jewel (the rest of copper’s coordination shell is occupied by hydroxyl functions of chitosan or of the hydration shell). Other authors (Muzzarelli et al., 1980; Schlick, 1986; Rhazi et al., 2002) propose a bridging model involving two amine functions, either from two different chains or from a single chain (the coordination shell being completed by the contribution of two hydroxyl groups form the chitosan or from the hydration shell (Guibal, 2004)). It is still today difficult to make a decision in favour of either of the models (Monteiro and Airoldi, 1999; Debbaudt et al., 2004). By means of molecular modelling Sabaté and co-workers (Llorens et al., 2004; Sabaté et al., 2006) proposed two complexes for cadmium involving the metal ion and the amine group of the chitosan chains (R-NH₂) as follows: Cd(R-NH₂)²⁺ and Cd(R-NH₂)₂²⁺.