Non-enhanced ultrafiltration of iron(III) with commercial ceramic membranes

Non-enhanced ultrafiltration of iron(III) with commercial ceramic membranes

Journal of Membrane Science 334 (2009) 129–137 Contents lists available at ScienceDirect Journal of Membrane Science journal homepage: www.elsevier...

634KB Sizes 0 Downloads 12 Views

Journal of Membrane Science 334 (2009) 129–137

Contents lists available at ScienceDirect

Journal of Membrane Science journal homepage:

Non-enhanced ultrafiltration of iron(III) with commercial ceramic membranes X. Bernat a , A. Pihlajamäki b , A. Fortuny c , C. Bengoa a , F. Stüber a , A. Fabregat a , M. Nyström b , J. Font a,∗ a

Departament d’Enginyeria Química, Escola Tècnica Superior d’Enginyeria Química, Universitat Rovira i Virgili, Av. Països Catalans 26, 43007 Tarragona, Catalonia, Spain Laboratory of Membrane Technology and Technical Polymer Chemistry, Department of Chemical Technology, Lappeenranta University of Technology, P.O. Box 20, FIN-53851 Lappeenranta, Finland c Departament d’Enginyeria Química, EPSEVG, Universitat Politècnica de Catalunya, Av. Víctor Balaguer s/n, 08800 Vilanova i la Geltrú, Barcelona, Catalonia, Spain b

a r t i c l e

i n f o

Article history: Received 1 December 2008 Received in revised form 19 February 2009 Accepted 22 February 2009 Available online 5 March 2009 Keywords: Iron Ceramic membranes Ultrafiltration Streaming potential Hydrolysis

a b s t r a c t Membranes are nowadays being developed as a mature technology to deal with polluted waters containing heavy metals. Nanofiltration membranes have classically been employed for this purpose. However, it was recently shown that a 5 kDa ceramic ultrafiltration membrane could be successfully used to recover iron(III) from aqueous solutions even at acidic pH, although the mechanisms associated with the retention of iron were not clearly established. This paper aims to highlight the phenomena associated with the retention of iron(III) species by commercial ceramic ultrafiltration membranes. The results show that iron(III) retention by ceramic membranes is strongly influenced by the molecular weight cut-off of the membranes and their material, although even 50 kDa molecular weight cut-off membranes are capable to efficiently retain iron species. A retention mechanism based on iron-membrane adsorption seems to be the most likely for explaining iron rejection when using UF membranes. However, charge repulsion phenomena and sieving effects may also contribute to the rejection of iron(III). The occurrence of mononuclear hydrolysed iron species, which can result in the formation of polynuclear species in the neighbourhood of the membrane surface could definitively contribute to the high retention of iron(III) shown by ceramic ultrafiltration membranes. © 2009 Elsevier B.V. All rights reserved.

1. Introduction Heavy metals are frequently present in polluted waters. Applications including electroplating, wood processing, leather tanning, anodising, among many others produce wastewaters containing heavy metals, which therefore need a removal or a recovery step to avoid environmental problems. Moreover, classical wastewater treatments, such as the Fenton process, use the addition of iron salts as catalyst to partially mineralise waters contaminated with organics [1]. These iron salts remain in the treated water, being a potential cause of subsequent environmental problems. Thus, iron leaving the oxidation unit has to be recovered and, if possible, recycled to the oxidation reactor, which could additionally reduce the cost associated with the treatment. Likewise, other salts of metals such as cobalt, manganese, nickel, vanadium, among others, have been tested as effective catalysts for the oxidation of organic pollutants, thus causing the same environmental and cost problems [2]. Dissolved heavy metals can be effectively recovered, or concentrated, by different techniques. Adsorption [3], extraction [4], ion exchange [5] and some other techniques have been successfully tested for this purpose.

∗ Corresponding author. Tel.: +34 977 559646; fax: +34 977 559667. E-mail address: [email protected] (J. Font). 0376-7388/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2009.02.024

Membranes can also be employed for the recovery of heavy metals from aqueous streams. Reverse osmosis membranes, which are able to retain ionic species, usually work at high transmembrane pressures, unfavourably affecting the treatment cost [6]. Nanofiltration is a promising membrane process for the recovery of ionic species [7,8] as well as low molar mass organic species [9]. However, there is a limited availability of either commercial polymeric membranes, resistant enough to the severe acidic pH in long-term operation, or commercial ceramic nanofiltration membranes. In turn, ultrafiltration has been classically employed to recover macromolecules, with molar masses of several thousands, from aqueous effluents. Thus, ultrafiltration membranes have been tested to recover metallic ions from aqueous solutions by adding macromolecules such as soluble polymers [10–13] and surfactants [14–17], which combine with the metals and enhance the separation. However, when dealing with these processes, a recovery step allowing the re-use of the surfactants or polymers is needed not only to decrease the operation costs associated with the additives but also to avoid discharges of organic material with the effluents containing the retained metals. Even though, in principle, ultrafiltration is not appropriate for dealing with isolated salts in solution, it has been previously demonstrated that a 5-kDa ceramic ultrafiltration membrane succeeded in retaining iron(III) species [18]. The pH was found to be a critical variable on iron(III) retention and the speciation of


X. Bernat et al. / Journal of Membrane Science 334 (2009) 129–137

Table 1 Properties of the membranes tested according to the manufacturer, unless otherwise noted.  (mm)




Active layer

PWPa (L/(h m2 bar))

Al2 O3 ––ZrO2 ––TiO2

ZrO2 –TiO2

19.6 58.2 71.5



52.3 76.4


␣-Al2 O3



– –



16.0 39.2

MWCO (kg/mol)

T1 T15 T50

1 15 50

F5 F10

5 10









5 30

– –

15 15

a b c d


Af (cm2 )



PWP was experimentally measured. As T membranes have three channels,  corresponds to the hydraulic diameter of each channel. Polyethersulfone. Polysulfone.

iron(III) seemed to be crucial for the efficiency of the recovery process. The aim of this study was to investigate the mechanisms involved in the retention of iron(III) solutions in filtration at acid conditions. Several commercial ceramic membranes made of different materials and with different molecular weight cut-off (MWCO) were selected in order to correlate their properties with the filtration performance. Streaming potential measurements were also conducted to characterise the charge of the membranes in connection with the separation performance. A discussion about the importance of iron(III) speciation by hydrolysis on the retention of iron(III) from aqueous streams and its consequences on the filtration performance is also presented. 2. Materials and methods 2.1. Membranes Six commercial ceramic membranes were selected. They were tubular membranes of 10 mm diameter and 250 mm length. Table 1 summarises some properties of the selected ceramic membranes such as the MWCO, number of channels (nc ) of each membrane element, available filtration area (Af ), internal diameter () and the inorganic material of fabrication. The designated T and F membranes were manufactured by Tami Industries (Nyons, France). The T elements were InsideCéram model whilst the F elements were Filtanium model. Different MWCO (between 1 and 50 kDa) were checked. M5 membrane was a Membralox® membrane manufactured by Pall Exekia (Tarbes, France) exhibiting a nominal MWCO of 5 kDa. Pure water permeabilities (PWP) at 25 ◦ C are also included in Table 1. They were obtained by measuring the permeate flux using deionised water at transmembrane pressures (TMP) of 2, 4, 6 and 8 bar. The plot of permeate flux against TMP was a straight line, as expected from the Darcy’s law. Its slope corresponds to the membrane water permeability. As Table 1 shows, PWP of T and F membranes increase when the MWCO of such membranes is raised. This agrees with the fact that higher MWCO membranes are more permeable to water due to the increased pore sizes. Ceramic membranes (F5 and M5) have quite similar PWP, 52.3 and 61.4 L/h m2 bar, respectively. The PWP of PT polymeric membrane, also with a MWCO of 5 kDa, is 16.0 L/h m2 bar, which differs significantly from those observed with F5 and M5. Thus, the polymeric membranes selected for this study have poorer permeabilities than the ceramic ones. In fact, this difference could have been expected because of the different materials, fabrication processes and manufacturers, which result in different membrane characteristics. In addition, two commercial flat sheet polyethersulfone (PT) and polysulfone (ER) polymeric membranes were selected to compare

their performance with that obtained with the ceramic membranes. PT and ER membranes were manufactured by GE-Osmonics (USA) and their MWCO were 5 and 30 kDa, respectively. The properties of PT and ER membranes are also listed in Table 1. 2.2. Chemicals Deionised water was used for pure water flux measurements. All chemical reagents were of analytical grade and used as received. Streaming potential measurements were done using 1 mM KCl solutions and adjusting their pH by adding suitable small volumes of 0.05 M KOH or 0.1 M HCl. The previous solutions were prepared using RO-filtered (Milli-Q) water with a conductivity of about 0.80 ␮S/cm. Concentrated HCl was also used for both adjusting the pH of the solutions to be filtered containing iron(III) and to dilute aqueous samples for iron analysis. Iron(III) solutions to be filtered were prepared from iron(III) nitrate nonahydrate and deionised water. A 10 g/L oxalic acid solution was used to clean the membranes after the filtration experiments in order to remove any iron adsorbed on the membrane and restore the initial properties. 2.3. Streaming potential measurements The charge of the membranes (apparent zeta potential) was measured through the membrane pores. The influence of the charge of all the layers composing the membrane was thus considered. Zeta potentials were calculated by using the Helmholtz–Schmoluchowski equation without any corrections resulting in relative apparent zeta potentials () [19]. In this study, the obtained apparent zeta potentials of the membranes tested were good enough for comparison purposes because all the measurements were done following the same experimental conditions and protocol. Characterisation of the surface charge of tubular ceramic membranes can be conducted without cutting or crashing the ceramic tubes [20–22]. However, it is also common to employ flat ceramic membranes [23,24] or cutting the membrane tubes for measuring the streaming potential [25]. The streaming potential results presented in this study were obtained without crashing the membrane elements so that, after characterisation, the membranes were used in the subsequent filtration experiments. Solutions containing 1 mM KCl were chosen for the determination of the streaming potential through the pores at 25 ◦ C. Reversible Ag/AgCl electrodes were used to measure the pressure induced potential difference (0.2–1.0 bar) between the retentate and permeate sides of the ceramic membranes. The streaming potential equipment is described in detail elsewhere [26]. However, the membrane module was exclusively designed for this work in order to be able to host tubular ceramic elements of 10 mm diam-

X. Bernat et al. / Journal of Membrane Science 334 (2009) 129–137

eter and 250 mm length. The membrane module was made of PVC and polycarbonate, and it was sealed with o-rings to prevent leaks during the tests. Before the streaming potential measurements, the membranes were kept in deionised water overnight. 2.4. Filtration experiments A detailed scheme of the filtration set-up to test ceramic membranes can be found elsewhere [18]. The filtration experiments were performed in total recycling mode. Thus, both retentate and permeate were recycled back to the feed tank in order to keep the iron(III) concentration constant during the filtration experiments. Before every test, the pure water flux of the clean membrane (Jw ) was measured at TMPs of 2, 4, 6 and 8 bar. At each TMP, the permeate flux was measured three times for five minutes. After that, the feed tank was filled with 3 L of a 0.90 mM iron(III) solution acidified at pH 2.0. Then, the pump was switched on and the TMP was set to 2 bar. The tangential velocity through the ceramic membrane tubes was 0.27 m s−1 in all the experiments. Permeate samples were periodically withdrawn and the permeate flux (JP ) was obtained from the permeate weight measured for a certain time with a balance (A&D Instruments, GF-1200). Once the steady-state was achieved, the TMP was subsequently changed to a higher value. The previous procedure was repeated at 4, 6 and 8 bar. Once the test at 8 bar was finished, the membrane was rinsed with deionised water for 15 min. The permeate water flux was then measured (Ja ) again at 2, 4, 6 and 8 bar as described for the determination of Jw , and compared to permeate fluxes measured before the experiment by means of the ratio Ja /Jw . This ratio is correlated to the permeate flux loss due to interaction between the iron(III) species and the membrane material, someway indicating severe fouling. A decline in the ratio Ja /Jw was usually observed so that membrane cleaning was needed before any further use. The cleaning procedure was conducted in recycling mode using a 10 g/L oxalic acid solution at 50 ◦ C and 2 bar for 1 h. Later, the membrane was rinsed with water until the pH of both permeate and retentate was neutral. Finally, the permeate flux was measured again in order to assure that the initial permeate flux was completely restored. If not, cleaning and rinsing steps were repeated until the permeate flux was completely recovered. The values of Jw and JP were used, besides for the abovementioned purposes, to calculate the membrane resistance (Rm ) and the created resistance (Ra ) occurring during the filtration of the ferric solutions. Thus, Ra includes all the phenomena causing the layer formation over the membrane material, which would include concentration polarisation phenomena and/or gel layer formation. Ra was calculated as the difference between the total resistance (Rt ) obtained during the filtration experiment and Rm . Darcy’s law has been adapted to obtain the previous parameters according to Eq. (1) and (2) where  is the dynamic viscosity of the aqueous media. The ratio of Ra /Rm has been calculated and used to infer the importance of the formed layer. Jw =

TMP Rm · 


JP =

TMP TMP = Rt ·  (Rm + Ra ) · 


The filtration performance of PT and ER membranes was tested in a commercial batch stirred cell (Sterlitech Corporation, model HP4750). An external stirrer (Selecta, model Agimatic REV-S) was used to fix the stirring speed during the experiments. Finally, a nitrogen cylinder and a backpressure valve were used to maintain the TMP during the filtration tests. Before use, the membrane was immersed in water overnight and subsequently pressurised at 12 bar in order to avoid further compaction. The experimental pro-


cedure was the same as with the ceramic membranes. Thus, Jw , JP and Ja were also measured during the experiments and several samples were withdrawn from the permeate for iron analysis. However, the experiments were done at a fixed TMP, 6 bar, and the stirring speed was fixed to 314 × 10−1 rad/s (300 rpm). The initial volume of the solution to be filtered was 240 mL. As the operating mode was different than that used with the ceramic membranes, the filtration with PT and ER membranes was stopped when 40 mL of solution remained inside the filtration cell. 2.5. Analyses The iron concentration in both feed solution (Cf ) and permeate samples (Cp ) was analysed by atomic absorption spectrometry (PerkinElmer spectrophotometer, model 3110) in an acetylene-air oxidising flame at a wavelength of 248.8 nm. 1% (wt/vol) HCl was used for diluting samples prior to iron analysis in order to prevent the hydrolysis of iron in the sampling vials. Iron(III) overall retention (R) was calculated according to Eq. (3). R (%) =


Cp Cf

· 100


3. Results and discussion 3.1. Streaming potential characterisation In this section the streaming potential of the different ceramic membranes used in this study is presented. Fig. 1 shows the effect of pH on the apparent zeta potential of T1, T15 and T50 membranes. It must be noted that, regardless the membrane, their charge changes from positive to negative at a pH between 4.5 and 5.0. Fievet et al. [20] measured the streaming potential of a T membrane, made by the same supplier and with the same ceramic materials, having a MWCO of 1.5 kDa, using an 1 mM KCl solution for the streaming potential tests. They found that the isoelectric point was located at pH 6. Taking into account that the experimental set-up was not identical, the streaming potential results are reasonably comparable. The streaming potential trends exhibited by these ceramic membranes as function of the pH can be explained by the equilibrium dissociation of the metal oxides forming the membrane [23]. When metal oxides are exposed to an aqueous media, the amphoteric surface groups (MOH) may dissociate. Thus, the reactions shown in Eqs. (4) and (5) can take place at acid or basic conditions, respectively. MOH + H+ → MOH2 + ↔ M+ + H2 O −

MOH + OH → M(OH)2 ↔ MO + H2 O

Fig. 1. Apparent zeta potential of Tl, T15 and T50 membranes at several pH.

(4) (5)


X. Bernat et al. / Journal of Membrane Science 334 (2009) 129–137

Fig. 2. Apparent zeta potential of F5 and F10 membranes at several pH.

Fig. 3. Apparent zeta potential of M5 membrane at several pH.

As it can be deducted from Eqs. (4) and (5), the dissociation of MOH surface groups is strongly related to the pH in the vicinity of the membrane surface. In addition, either when particles have no charge or when they have the same number of positive and negative charges, the surface charge becomes zero. Thus, the previous explanation serves to understand that ceramic membranes have positive apparent zeta potentials at low pH, but they become negative at higher pH. As Fig. 1 shows, the MWCO of the membrane does not seem to affect the charge density when the pH is higher than about 5.2 because the apparent zeta potentials are quite close. However, the higher the MWCO of the membrane, the larger the charge density in the positive charge zone. Fig. 2 shows the evolution of the membrane charge at different pH for the F membranes. For F membranes, the membrane charges are smaller than those of T membranes in the negative zone, i.e. at pH higher than about 5.5. It must be noted that F membranes are entirely made of titania, whereas T membranes are made of mixtures of alumina, titania and zirconia. In addition, the F10 membrane has practically no charge after the isoelectric point. Labbez et al. [27] found that the isoelectric point of a titanium oxide membrane (also from Tami Industries) was at pH 6, which is in agreement with the presented characterisation. Unlike T membranes, the higher the MWCO, the lower the membrane charge density beyond pH 6. Obviously, the membrane material, membrane configuration and/or manufacturing process seem to have a significant effect on the apparent zeta potential of ceramic membranes. The apparent zeta potential trend of M5 membrane is similar to that obtained with T and F membranes. As it can be seen in Fig. 3, the membrane has the isoelectric point at a pH of about 5.4. In addition, when comparing the apparent zeta potentials of this membrane to that observed with F5 membrane, which has the same MWCO, the values are quite similar. Overall, the isoelectric point is found to be at a pH between 4.5 and 5.5 for all the membranes. Moritz et al. [23] stated that the isoelectric points of titania, zirconia and alumina flat membranes were at pH 4.2, 4.6 and 4.3 respectively, which is mostly in agreement with the results obtained for the membranes used in the present study. The small differences between the isoelectric points presented in this study and those shown by Moritz et al. [23] may be due to the different manufacturing processes and membrane configurations, as demonstrated by the same authors in a previous work [28]. In addition, membrane pre-treatment, history, aging, storage and so on may lead to different streaming potential values. On the other hand, the isoelectric points of the oxides used in the composition of the tested membranes were found to be

at pH 7–8.1 for alumina, pH 8 for zirconia and pH 5–6.1 for titania [29,30]. Hence, it is evident that the isoelectric point of titania, oxide found in the composition of all the selected membranes, is closer to the isoelectric point of the membranes than those of zirconia and alumina particles are. 3.2. Iron(III) filtration with T membranes T1, T15 and T50 membranes exhibited a Rm of 2.06 × 1016 , 6.94 × 1015 and 5.65 × 1015 m−1 , respectively, for pure water flux. The iron solution filtration results obtained with T membranes are presented in Fig. 4, which shows the permeate flux evolution at several TMP. Regardless the membrane, JP decreases along the time. However, practically all the flux decline occurred during the first thirty minutes of filtration and a steady-state is observed beyond this point. Fig. 4 also shows that, when increasing the TMP after the steady-state is achieved, JP is temporarily increased regardless the experimental conditions. However, as long as the filtration is carried out, JP decreases again achieving the steady-state. Generally, regardless the membrane used, there is a limiting flux beyond which an increase of pressure does not cause a gain on the JP . Thus, flux is limited by fouling and a TMP increase is only invested either in compacting the deposited fouling layer and/or increasing its thickness. As it can be seen in Table 2, permeate flux of T1 membrane is limited at around 65 L/h m2 whilst for T15 and T50 membranes, permeate flux is limited at around 95 L/h m2 . In Table 2, it can also be

Fig. 4. Permeate flux evolution of T membranes at pH 2 and several TMP.

X. Bernat et al. / Journal of Membrane Science 334 (2009) 129–137


Table 2 Steady-state permeate fluxes and resistances of T membranes. TMP (bar)


T15 2

2 4 6 8


JP (L/h m )

Ra (m


Ra /Rm

36.1 ± 1.1 56.8 ± 2.6 64.5 ± 3.5 66.4 ± 4.7

1.79 × 10 7.85 × 1015 1.70 × 1016 2.81 × 1016 15

0.09 0.38 0.82 1.36

T50 2

JP (L/h m ) 83.8 94.0 96.0 93.7

± ± ± ±

2.8 14.7 13.3 13.5

observed that for T1 membrane, Ra is increased from 1.79 × 1015 to 2.81 × 1016 m−1 when the TMP is increased from 2 to 8 bar. Table 2 also demonstrates that the formed layer becomes increasingly significant as the MWCO is higher. At 2 bar, Ra /Rm was 0.09 for T1, 0.39 for T15 and 0.78 for T50. As shown in Fig. 5, the TMP does not exhibit almost any influence on the iron(III) retention, which only depends on the membrane used. The retentions are quite significant, between 60 and 80%, and cannot be exclusively attributed to sieving mechanisms. In a previous paper, the occurrence of the iron retention was attributed to the presence of soluble charged iron hydroxides that could interact with the ceramic material, forming a stable layer over the membrane surface [18]. It is believed that this new layer is mainly responsible for the iron retention capacity of these UF membranes. The influence of the MWCO on the retention measured could depend on the pore size distribution. Larger pores leave a higher free pass area to the flux that makes this layer less stable because of the larger flux, leaving defects where the solution can freely pass through. Initially, the observed retention may actually be explained by size exclusion, charge repulsion, and iron interactions with the ceramic material or a combination of them. Regarding the charge of T membranes, at pH 2, these T membranes are positively charged. In addition, the iron species appearing in solution at this pH are soluble charged hydroxides as Fig. 6 shows. The speciation diagram presented in Fig. 6 has been obtained with the Medusa software, a freely available software for the obtention of speciation diagrams, and is valid for a 0.9 mM Fe(III) solution at variable pH [31]. Thus, repulsion between positive charged hydroxides and the positively charged surface groups of the membranes does occur, but it is not expected to be mainly responsible of the ion rejection because of the large MWCO in comparison with the molecular weights of the species to be retained, mainly for the highest MWCO membranes. To distinguish the previous effects, the hydrolysis of iron(III) in solution and the molar masses of soluble hydroxides must be taken into account.

Fig. 5. Iron retention of T membranes at pH 2 and several TMP.


Ra (m


2.70 × 10 1.03 × 1016 1.83 × 1016 2.76 × 1016 15

Ra /Rm

JP (L/h m2 )

0.39 1.48 2.64 3.98

80.2 95.8 95.2 94.8

± ± ± ±

6.2 9.3 8.4 9.2

Ra (m−1 )

Ra /Rm

4.42 × 1015 1.12 × 1016 1.98 × 1016 2.84 × 1016

0.78 1.98 3.50 5.03

Soluble iron charged hydroxides present in solution occur as a result of the hydrolysis of iron(III). The hydrolysis process of ferric ion in aqueous media is classically explained by the following reactions [32]: Fe3+ ↔ FeOH2+ ↔ polynuclear species ↔ Fe2 O3 ·nH2 O (s)


However, Eq. (6) does not include water linked on iron species, as well as water and protons involved in the reactions according to the stoichiometry. In acidic solutions, taking into account water ligands, ferric ion exists as the hexaaquo ion, Fe(H2 O)6 3+ (M = 163.93 g/mol) [33]. Accordingly, FeOH2+ and Fe(OH)2 + are present as Fe(OH)(H2 O)5 2+ (M = 162.93 g/mol) and Fe(OH)2 (H2 O)4 + (M = 161.92 g/mol), respectively. Thus, taking into account these molecular weights and the membrane MWCO, which is at least one order of magnitude greater, the retention cannot merely be assigned to a size-exclusion mechanism. Polynuclear species can also exist in solution. Iron hydrolysis has been considered as a polymerisation reaction because of the formation of polynuclear species of iron in aqueous solution, where Fe(H2 O)6 3+ can be taken as the monomer [33]. Although different polymerisation processes may occur in aqueous media, the olation process seems to be the predominant, at least in the first stages of polymerisation. When olation occurs, hydroxide interactions are responsible for the formation of nearly all the polynuclear species of the M3+ cations. The polynuclear species formed from Fe(H2 O)6 3+ can be Fe2 (HO)2 4+ and Fe3 (HO)4 5+ [34]. Thus, mononuclear and polynuclear species appearing in solution could actually contribute to the retention in the filtration of iron(III), because of the interaction (adsorption) with the membrane material thus forming a stable layer of these species on the membrane. Nevertheless, the speciation diagram of 0.9 mM Fe(III) solution at pH 2, shown in Fig. 6, predicts the presence of mononuclear hydrolysed species in solution but not polynuclear ones at this solution pH. It has been demonstrated that iron(III) cations are adsorbed onto ceramic materials forming a monolayer and their adsorption is closely related to the metal speciation [35]. A maximal iron adsorp-

Fig. 6. Chemical speciation diagram of a 0.90 mM Fe(III) solution.


X. Bernat et al. / Journal of Membrane Science 334 (2009) 129–137

Fig. 8. Permeate flux evolution of F and M5 membranes at pH 2 and several TMP.

Fig. 7. Ja /Jw of T membranes at pH 2 and several TMP.

tion capacity is observed when hydrolysed species are present in the solution [36]. Concretely, a range of pH between 1.5 and 2.6 has been found to be optimal for adsorbing Fe(III) species onto ceramic materials due to the appearance of iron(III) hydrolysed species. In addition, adsorption studies of Fe(III) onto titania at pH 2 have shown that polymerisation of Fe(III) species favours the adsorption capacity of the material [35,37]. Thus, the initial mechanism allowing the retention of Fe(III) by ceramic membranes can be explained by the adsorption of the hydrolysed species of Fe(III) onto the ceramic material of the membrane, forming a deposit layer. The proposed mechanism is supported by the sharp decrease on the permeate flux as well as by the high resistances of the formed layers compared to the ones of the membranes before being used. After the Fe(III) species adsorption, the retention may be even enhanced by charge repulsions between these positively charged hydrolysed Fe(III) species adsorbed onto the membrane and the positively charged species present in the bulk solution. In addition, the trend shown by the retention at several TMP using T1, T15 and T50 suggests that, at the steady-state, a size effect could be decisive on the rejection of iron species by ceramic ultrafiltration membranes. An accumulation of iron soluble hydroxides, i.e. an increase of the local iron concentration, in the neighbourhood of the membrane, caused by concentration polarisation and deposit layer formation, can be expected. Thus, the chemical properties of the ferric solution would change, allowing the formation of polynuclear species, which can be more easily retained by size-exclusion phenomena although charge repulsions and adsorption phenomena may simultaneously occur. In fact, polynuclear hydrolysed species are more easily adsorbed onto ceramic material or even they can directly grow on the ceramic material after the adsorption of their mononuclear predecessors [38]. The Ja /Jw revealed by T membranes, due to the filtration of ferric solutions, is shown in Fig. 7. TMP does not significantly affect the permeate flux loss of T membranes. Nevertheless, the higher the membrane MWCO, the higher the ratio Ja /Jw and, in consequence, the lower the permeate flux loss. As the MWCO increases,

the Rm decreases so that the additional resistance given by the layer formed has a stronger impact on the overall resistance, becoming even dominant. This could be ascribed to the fact that the higher flux keeps a greater iron concentration near to the membrane as the retention is maintained, therefore leaving more chance to the formation of a thicker layer. However, the increased thickness of this layer favours its partial removal after cleaning the membranes with deionised water in crossflow mode. 3.3. Iron(III) filtration with F and M5 membranes Filtration of 0.9 mM Fe(III) solution at pH 2 was also tested using F membranes, which are fully made of TiO2 . Also, M5 membrane was tested for the same purpose. As stated in Table 1, M5 membrane support is made of ␣-Al2 O3 whereas its active layer is TiO2 . Thus, filtration results of the previous membranes are useful to ascertain the effect of the ceramic membrane material on the filtration performance. Fig. 8 shows the JP evolution for F and M5 membranes at several TMP. As for T membranes, JP falls during the first 30 min of operation after setting the TMP and a constant JP is observed after this initial period. Thus, iron species are instantaneously deposited on the membrane material from the beginning of the filtration. As for T membranes, Table 3 demonstrates that, when using F and M5 membranes, permeate flux becomes limited when the TMP is raised. JP of F5 and F10 membranes becomes limited at around 75 L/h m2 . In consequence, an increase on the TMP only serves to thicken and/or compact the deposit layer. M5 permeate flux is already limited at around 44 L/h m2 , which is significantly lower than those observed for T and F membranes. Thus, higher adsorption density occurs when using M5 membrane, which is probably related to the different composition of the membrane material. Thus, a difference in the membrane material strongly affects the filtration performance. Concerning the study of the resistances, the clean membrane resistances of F5, F10 and M5 membranes are 7.72 × 1015 , 5.29 × 1015 and 6.58 × 1015 m−1 , respectively. Table 3 also shows that, when TMP and MWCO are larger for F membranes,

Table 3 Steady-state permeate fluxes and resistances of F and M5 membranes. TMP (bar)

2 4 6 8



JP (L/h m2 )

Ra (m−1 )

63.7 ± 5.7 78.4 ± 13.0 76.5 ± 6.1 76.8 ± 6.5

4.96 × 10 1.29 × 1016 2.40 × 1016 3.44 × 1016

Ra /Rm 15

0.64 1.67 3.11 4.46


JP (L/h m2 ) 60.5 74.8 74.7 75.0

± ± ± ±

6.7 7.7 6.2 8.4

Ra (m−1 ) 8.07 × 10 1.63 × 1016 2.72 × 1016 3.78 × 1016 15

Ra /Rm

JP (L/h m2 )

1.53 3.08 5.14 7.15

47.6 45.7 43.3 43.8

± ± ± ±

14.3 11.2 12.7 11.0

Ra (m−1 )

Ra /Rm

1.04 × 1016 2.88 × 1016 4.94 × 1016 6.72 × 1016

1.58 4.38 7.51 10.2

X. Bernat et al. / Journal of Membrane Science 334 (2009) 129–137

Ra /Rm is also greater. At 8 bar, Ra /Rm is 4.46 for F5 membrane whilst it is 7.15 for F10 membrane. In addition, Table 3 also indicates that the two membranes with the same MWCO but made of different materials (F5 and M5) show different resistances. The Ra /Rm of M5 is always higher than that of F5 membrane, regardless of the TMP. For example, at 8 bar, Ra /Rm of M5 is about 2.3 times higher than that of F5. Thus, the composition of the ceramic membranes undoubtedly influences the filtration process. As Fig. 9 shows, iron retentions are slightly increased when the TMP increases. Instead, no differences of iron retention are observed when changing the MWCO of the membranes with F membranes. This effect can be attributed to the material of the membrane that would influence the adsorption between soluble charged hydroxides and the membrane material. Moreover, the retentions achieved with M5 membranes are slightly lower than those with F5. As abovementioned, when comparing F5 and M5 membranes, lower overall resistances are obtained when working with F5. Thus, a denser or thicker layer could be formed when dealing with M5 membranes causing a higher permeate flux decrease than with the F5. Once more, the influence of the membrane material and the manufacturing process on iron retention is demonstrated. Comparing the apparent zeta potentials of M5 and F5, the former has a slightly higher charge than the latter. Hence, if repulsion was the dominant factor in retention of iron species, a significant difference on M5 iron retention should be expected. However, when inspecting Fig. 9, M5 retentions are only slightly lower than F5 retentions. Thus, these results support that charge repulsion does not have an essential role on the iron(III) retention by ceramic ultrafiltration membranes. Fig. 10 shows that the ratios Ja /Jw measured with F membranes are higher than those observed with T membranes, previously shown in Fig. 7, which indicates that lower permeate flux loss occurs when using F membranes. Consequently, when using F membranes, less fouling remains after cleaning the membranes with deionised water. As Fig. 10 shows, the ratios Ja /Jw at the tested TMPs when filtering with M5 membrane are the highest observed. Thus, despite the increased resistance of the formed layer when using M5 membrane, less severe fouling remains after cleaning the membrane with deionised water, which agrees with the results obtained with T membranes. 3.4. Iron(III) filtration with PT and ER membranes In order to assess the key role of the ceramic nature of the membranes, the polymeric PT and ER membranes were used to filter 0.9 mM Fe(III) solutions at pH 2, as with the six ceramic


Fig. 10. Ja /Jw of F and M5 membranes at pH 2 and several TMP.

membranes tested. Previous streaming potential characterisation of polyethersulfone ultrafiltration membranes, carried out with the same streaming potential device as in this study, has shown that their surfaces are neutral at all pH spectra [39,40]. On the contrary, the isoelectric point of polysulfone membranes was found at a pH between 3 and 4 [39,41,42]. Furthermore, the surface charges of ER membrane were higher than those observed with any of the ceramic membranes tested in this study [39]. Thus, if repulsion was the dominant factor on the filtration of ferric solutions by ultrafiltration membranes, significant rejection should be observed when dealing with ER membrane. However, no retention was observed at any TMP with PT membrane, whereas meagre retentions, below 3%, were achieved with ER membrane. The fact that the retention of iron species was almost nil when using polymeric membranes clearly indicates that the membrane charge had no effect on the rejection of iron species by ultrafiltration. Hence, the membrane material affinity for adsorbing iron species is decisive for their retention, as explained in the previous sections. In fact, it has been proven that iron ions are highly adsorbed on ceramic materials and no iron adsorption is detected when polysulfone fibres are tested [43]. Thus, because of the lack of inorganic oxides, there are no adsorption phenomena taking place in polymeric membranes and the subsequent formation of a deposit layer does not occur, which prevents polymeric membranes from being able to retain iron species. It is worth mentioning that in this study, the ceramic membranes had to be cleaned, after the filtration tests, with an oxalic acid solution, as explained before, because it is an iron-complexing agent (as recommended by the membranes’ manufacturers) and it was able to remove the iron fouling. Hydrochloric acid and phosphoric acid were also tested as cleaners and no fouling removal was achieved. Thus, the necessity of using a complexing agent for the total removal of iron from the used ceramic membranes confirms that the layer formed over the ceramic material was significantly stable. 4. Conclusions

Fig. 9. Iron retention of F and M5 membranes at pH 2 and several TMP.

Ceramic ultrafiltration membranes can be successfully employed to separate iron(III) from aqueous solutions at pH 2 achieving retentions up to 98%. The transmembrane pressure does not strongly influence the iron retention. Nevertheless, permeate flux becomes limited when increasing the TMP for all the tested membranes. For T1 membrane, permeate flux is limited at around 65 L/h m2 whilst for T15 and T50 membranes, permeate flux is limited at around 95 L/h m2 . In addition, permeate fluxes of F5 and F10 membranes are limited at around 75 L/h m2 whilst M5 permeate flux becomes limited at 44 L/h m2 . Thus, the ceramic


X. Bernat et al. / Journal of Membrane Science 334 (2009) 129–137

material has an important effect on the filtration performance and is crucial to reject iron species. A retention mechanism based on the formation of a stable layer of iron species adsorbed on the membrane material seems to be the key factor allowing the retention of iron by ultrafiltration. The existence of the layer is demonstrated by the occurrence of severe fouling that can only be completely eliminated by cleaning with oxalic acid, which is well-known to complex iron. After the layer formation, charge repulsions between the adsorbed iron species (positively charged) and the iron species present in the bulk solution (also positively charged) can coexist with adsorption phenomena. The effect of the MWCO of the membranes on the iron retention and the low surface charges exhibited by the tested membranes reveals that a sieving mechanism probably takes place during the ultrafiltration of the ferric solutions. Thus, mononuclear soluble iron hydroxide species occurring at pH equal to 2 may initiate the formation of polynuclear species in the vicinity of the membrane surface due to the local increase of the iron concentration in this zone. However, although sieving effects can explain the retention results, adsorption of polynuclear species onto the membranes and charge repulsions between the adsorbed positive species and those present in the bulk solution can also contribute to the iron retention. The fact that polymeric membranes, with similar MWCO as those of the ceramic, cannot retain iron(III) or only slightly retain it indicates that the ultrafiltration membranes must be inorganic, i.e. based on metal oxides. The ceramic material ensures the feasibility of the separation process, mainly attributed to adsorption phenomena, not occurring when polymeric membranes are used. Moreover, the different charge densities of the polymeric membranes examined, which exhibit very low retention, confirms that the electrostatic repulsion phenomenon by itself does not play a key role on the retention of ferric species by means of ultrafiltration membranes. Acknowledgements The Spanish Ministry of Education and Science is gratefully acknowledged for its funding for this project (CTM2005-01873). Universitat Rovira i Virgili is also thanked for providing a Ph.D. scholarship to carry out this research work.

Nomenclature Af Cf Cp Ja JP Jw MWCO nc PWP R Ra Rm Rt TMP   

available filtration area (cm2 ) iron concentration in the feed solution (mM) iron concentration in the permeate (mM) permeate flux after the filtration of ferric solutions (L/h m2 ) permeate flux during the filtration of ferric solutions (L/h m2 ) clean membrane permeate flux (L/h m2 ) molecular weight cut-off (kDa) number of channels pure water permeability (L/h m2 bar) iron retention additional resistance (m−1 ) clean membrane resistance (m−1 ) total resistance during the filtration of ferric solutions (m−1 ) transmembrane pressure (bar) internal diameter (mm) dynamic viscosity (Pa s) zeta potential (mV)

References ˜ [1] M. Pera-Titus, V. García-Molina, M.A. Banos, J. Jiménez, S. Espulgas, Degradation of chlorophenols by means of advanced oxidation processes: a general review, Appl. Catal. B-Environ. 47 (2004) 219–256. [2] G.P. Anipsitakis, D.D. Dionysiou, Radical generation by the interaction of transition metals with common oxidants, Environ. Sci. Technol. 38 (2004) 3705–3712. [3] U.K. Garg, M.P. Kaur, V.K. Garg, D. Suda, Removal of hexavalent chromium from aqueous solution by agricultural waste biomass, J. Hazard. Mater. 140 (2007) 60–68. [4] K. Sarangi, P.K. Parhi, E. Padhan, A.K. Palai, K.C. Nathsarma, K.H. Park, Separation of iron(III), copper(II) and zinc(II) from a mixed sulphate/chloride solution using TBP, LIX 84I and Cyanex 923, Sep. Purif. Technol. 55 (2007) 44–49. [5] T. Panayotova, M. Dimova-Todorova, I. Dobrevsky, Purification and reuse of heavy metals containing wastewaters from electroplating plants, Desalination 206 (2007) 135–140. [6] Z. Ujang, G.K. Anderson, Performance of low pressure reverse osmosis membranes (LPROM) for separating mono- and divalent ions, Water Sci. Technol. 38 (1998) 521–528. [7] S. Alami-Younssi, A. Larbot, M. Persin, J. Sarrazin, L. Cot, Rejection of mineral salts on a gamma alumina nanofiltration membrane. Application to environmental process, J. Membr. Sci. 102 (1995) 123–129. [8] K.-H. Choo, D.-J. Kwon, K.-W. Lee, S.-J. Choi, Selective removal of cobalt species using nanofiltration membranes, Environ. Sci. Technol. 36 (2002) 1330–1336. [9] B. Van der Bruggen, J. Schaep, D. Wilms, C. Vandecasteele, Influence of molecular size, polarity and charge on the retention of organic molecules by nanofiltration, J. Membr. Sci. 156 (1999) 29–41. [10] M.S. Diallo, S. Christie, P. Swaminathan, J.H. Johnson, J.R.W.A. Goddard, Dendrimer enhanced ultrafiltration. 1. Recovery of Cu (II) from aqueous solutions using PAMAM dendrimers with ethylene diamine core and terminal NH2 groups, Environ. Sci. Technol. 39 (2005) 1366–1377. [11] O. Sanli, G. Asman, Removal of Fe (III) ions from dilute aqueous solutions by alginic acid-enhanced ultrafiltration, J. Appl. Polym. Sci. 77 (2000) 1096–1101. [12] R. Molinari, S. Gallo, P. Argurio, Metal ions removal from wastewater or washing water from contaminated soil by ultrafiltration–complexation, Water Res. 38 (2004) 593–600. ˜ [13] P. Canizares, A. Pérez, R. Camarillo, J.J. Linares, A semi-continuous laboratoryscale polymer enhanced ultrafiltration process for the recovery of cadmium and lead from aqueous effluents, J. Membr. Sci. 240 (2004) 197–209. [14] P. Reiller, D. Lemordant, A. Hafiane, C. Moulin, C. Beaucaire, Extraction and release of metal ions by micellar-enhanced ultrafiltration: influence of complexation and pH, J. Colloid Interface Sci. 177 (1996) 519–527. [15] M. Aoudia, N. Allal, A. Djennet, L. Toumi, Dynamic micellar enhanced ultrafiltration: use of anionic (SDS)-nonionic (NPE) system to remove Cr3+ at low surfactant concentration, J. Membr. Sci. 217 (2003) 181–192. [16] R.-S. Juang, Y.-Y. Xu, C.-L. Chen, Separation and removal of metal ions from dilute solutions using micellar-enhanced ultrafiltration, J. Membr. Sci. 218 (2003) 257–267. [17] J. Sabaté, M. Pujolà, J. Llorens, Comparison of polysulfone and ceramic membranes for the separation of phenol in micellar-enhanced ultrafiltration, J. Colloid Interface Sci. 246 (2002) 157–163. [18] X. Bernat, A. Fortuny, F. Stüber, C. Bengoa, A. Fabregat, J. Font, Recovery of iron (III) from aqueous streams by ultrafiltration, Desalination 221 (2008) 413–418. [19] M. Nyström, M. Lindström, E. Matthiasson, Streaming potential as a tool in the characterization of ultrafiltration membranes, Colloids Surf. 36 (1989) 297–312. [20] P. Fievet, M. Sbaï, A. Szymczyk, C. Magnenet, C. Labbez, A. Vidonne, A new tangential streaming potential setup for the electrokinetic characterization of tubular membranes, Sep. Sci. Technol. 39 (2004) 2931–2949. [21] S. Condom, M. Cretin, M. Persin, J. Sarrazin, A. Larbot, Characterization of three low UF mineral membranes by streaming potential measurements, Desalination 149 (2002) 447–451. [22] S. Condom, A. Larbot, S. Alami-Younssi, M. Persin, Use of ultra- and nanofiltration ceramic membranes for desalination, Desalination 168 (2004) 207–213. [23] T. Moritz, S. Benfer, P. Árki, G. Tomandl, Influence of the surface charge on the permeate flux in the dead-end filtration with ceramic membranes, Sep. Purif. Technol. 25 (2001) 501–508. [24] M. Sbaï, P. Fievet, A. Szymczyk, B. Aoubiza, A. Vidonne, A. Foissy, Streaming potential, electroviscous effect, pore conductivity and membrane potential for the determination of the surface potential of ceramic ultrafiltration membrane, J. Membr. Sci. 215 (2003) 1–9. [25] P. Narong, A.E. James, Sodium chloride rejection by a UF ceramic membrane in relation to its surface electrical properties, Sep. Purif. Technol. 49 (2006) 122–129. [26] M. Nyström, A. Pihlajamäki, N. Ehsani, Characterisation of ultrafiltration membranes by simultaneous streaming potential and flux measurements, J. Memb. Sci. 87 (1994) 245–256. [27] C. Labbez, P. Fievet, A. Szymczyk, F. Thomas, C. Simon, A. Vidonne, J. Pagetti, A. Foissy, A comparison of membrane charge of a low nanofiltration ceramic membrane determined from ionic retention and tangential streaming potential measurements, Desalination 147 (2002) 223–229. [28] T. Moritz, S. Benfer, P. Arki, G. Tomandl, Investigation of ceramic membrane materials by streaming potential measurements, Colloid Surf. A-Physicochem. Eng. Asp. 47 (2001) 25–33. [29] M. Grün, A.A. Kurganov, S. Schacht, F. Schüth, K.K. Unger, Comparison of an ordered mesoporous aluminosilicate, silica, alumina, titania and zirconia in

X. Bernat et al. / Journal of Membrane Science 334 (2009) 129–137

[30] [31] [32]


[34] [35] [36]


normal-phase high-performance liquid chromatography, J. Chromatogr. A 40 (1996) 1–9. J. Winkler, S. Marmé, Titania as a sorbent in normal-phase liquid chromatography, J. Chromatogr. A 888 (2000) 51–62. I. Puigdomenech, Medusa Software, 2004, J.J. Pignatello, E. Oliveros, A. MacKay, Advanced oxidation processes for organic contaminant destruction based on the Fenton reaction and related chemistry, Crit. Rev. Environ. Sci. Technol. 36 (1) (2006) 1–84. M.A. Blesa, Phase transformations of iron oxides, oxohydroxides, and hydrous oxides in aqueous media, Adv. Colloid Interface Sci. 29 (1989) 173– 221. C.F. Baes, R.E. Mesmer, The Hydrolysis of Cations, John Wiley & Sons, New York, 1976. M.A. Anderson, A. Rubin, Adsorption of Inorganics at Solid–Liquid Interfaces, Ann Arbor Science, MI, 1981. R.O. James, T.W. Healy, Adsorption of hydrolyzable metal ions at the oxidewater interface. I. Co(II) adsorption on SiO2 and TiO2 as model systems, J. Colloid Interface Sci. 40 (1972) 42–52. E. Piera, M.I. Tejedor-Tejedor, M.E. Zorn, M.A. Anderson, Relationship concerning the nature and concentration of Fe(III) species on the surface of TiO2 particles

[38] [39]






and photocatalytic activity of the catalyst, Appl. Catal. B-Environ. 46 (2003) 671–685. R.O. James, P.J. Stiglich, T.W. Healy, Analysis of models of adsorption of metal ions at oxide/water interfaces, Faraday Discuss. Chem. Soc. 52 (1975) 142–156. A. Weis, M.R. Bird, M. Nyström, The chemical cleaning of polymeric UF membranes fouled with spent sulphite liquor over multiple operational cycles, J. Membr. Sci. 216 (2003) 67–79. D. Delaunay, M. Rabiller-Baudry, L. Paugam, A. Pihlajamäki, M. Nyström, Physico-chemical characterisations of a UF membrane in dairy application to estimate chemical efficiency of cleaning, Desalination 200 (2006) 189–191. A. Martín, F. Martínez, J. Malfeito, L. Palacio, P. Prádanos, A. Hernández, Zeta potential of membranes as a function of pH. Optimization of isoeelectric point evaluation, J. Membr. Sci. 213 (2003) 225–230. K.S. Kim, K.H. Lee, K. Cho, C.E. Park, Surface modification of polysulfone ultrafiltration membrane by oxygen plasma treatment, J. Membr. Sci. 199 (2002) 135–145. D.L. Jones, A.C. Edwards, Evaluation of polysulfone hollow fibres and ceramic suction samplers as devices for the in situ extraction of soil solution, Plant Soil 150 (1993) 157–165.