1

Hydrophilicity and Charge of NF Membranes and their Effect on the Filtration Efficiency at Different pH Mika Mänttäri∗, Arto Pihlajamäki# and Marianne Nyström Laboratory of Membrane Technology and Technical Polymer Chemistry, Department of Chemical Technology, Lappeenranta University of Technology, P.O. Box 20, FIN-53851 Lappeenranta, Finland. [email protected], [email protected], [email protected], fax: +358-5-6212199 Abstract It is well known that the hydrophilicity of membranes affects their flux and fouling. It may also be assumed that hydrophilic membranes have a higher charge than hydrophobic membranes. In this work the effect of pH on membrane structure, its permeability and retention was studied. In addition, we studied whether the changes in the membrane properties due to the pH change were reversible. This is important for understanding the performance of nanofiltration membranes at different conditions and for the selection of cleaning processes. Moreover, the results facilitate the choice of membrane for specific applications. Several commercial NF membranes were studied at different pH values. Their retention and flux were explained by the charge and the hydrophilic characteristics of the membranes. The filtrations were made with uncharged sugar and salt solutions. The flux of more hydrophilic membranes changed more than hydrophobic ones when the pH was increased in the feed solution, i.e. these membranes became significantly more open at high pH. This was explained by the chemical nature of the polymer chains in the membrane skin layer, i.e., dissociable groups in the polymer made the surface more hydrophilic and looser when charges of the polymer chains started to repel each other at elevated pH. The changes in permeabilities and retentions were mostly found to be reversible in the pH range studied.

1.

Introduction

Nanofiltration is a relatively new pressure driven membrane separation process which offers higher fluxes than reverse osmosis and significantly better retentions than ultrafiltration for lower molar mass molecules such as sugars, natural organic matter (NOM) and even ions. The separation process in nanofiltration is a combination of sieving and diffusion of molecules through the selective layer of a membrane. Also, surface charges play a more important role in separation by NF than with other pressure driven membrane processes. Thus, the membrane selective layer could be thought to be a three-dimensional network of polymer chains. Therefore, NF membranes often do not have real pores but just free space inside the polymer chain network. The most attractive applications for NF are in treatment of different industrial effluents especially for reuse when removal of organics is desired, but retention of ions is not needed, in separation of NOM from surface waters (low ionic content) and in very specific separations of organic molecules of similar sizes. The retention in NF of different species in solution at different pH has been investigated intensively. However, most of the studies are limited to retention of different types of electrolytes in various pHs. The retentions of charged species, e.g. ions, have been explained to depend on the valency, on the concentration and chemical nature of the compounds in solution and on the surface charge, charge density and the chemical nature of the groups on the membrane surface. The retention of uncharged species like organics depends on size and shape, the chemical nature and the hydrophilicity/-phobicity of the compounds. In our previous study it was found that different NF membranes showed very different retentions when only a small change in pH of the feed was done [1]. This was probably due to the different surface chemistries of the membranes studied. The aim of this study was to test NF membranes in the whole pH range the manufacturers recommend. Not only monovalent ion retention was analysed but also the retention of a small, ∗ #

Corresponding author Presenting author

2 uncharged organic molecule (glucose) was detected along with pure water permeability. The filtration runs of dilute sodium chloride (NaCl) and glucose solution were conducted at different pH for several commercial NF membranes. For closer study three membranes of different skin layer material were chosen. The ampholytic nature of the membrane surfaces was confirmed by streaming potential measurements and their hydrophilicity at different pH was studied by contact angle measurements. 2.

Materials and methods

2.1. Membranes The NF membranes studied were NTR 7450 (Hydranautics), NF 200 (Dow Deutschland) and Desal-5 DL (Osmonics). The membrane samples were kindly supplied by the manufacturers. The NTR 7450 membrane has a sulphonated polyethersulphone skin layer on top of a polysulphone support. The NF 200 membrane has a semi-aromatic piperazine based polyamide skin layer on a polysulphone microporous support reinforced with polyester nonwoven backing. The Desal-5 DL membrane has a proprietary polyamide skin layer on a polysulphone/polyester support. In Table 1 some properties of these membranes are presented. Table 1 Some properties of membranes studied according to manufacturers unless otherwise stated. PWP = pure water permeability, R = retention, ∆pH = recommended pH range (long term), θ = contact angle with pure water. Membrane PWP @ 25°C (-) (L/m2hbar) MgSO4 NTR 7450 13 a NF 200 11 a Desal-5 DL 7.6 96 a Own measurements b Ikeda et al. [2] c http://www.somicon.com 14.07.2003 d Own measurements, sessile drop method

NaCl 50 b 70

R (%) Na2SO4 92 b

CaCl2 < 45

∆pH (-) 2-11c 3-10 2-11

θ (°) 58 d 30 d 41 d

2.2. Chemicals All chemicals used were of analytical grade. NaCl and glucose were used in the filtration runs and the pH was adjusted by 0.1 M sodium hydroxide (NaOH) or hydrochloric acid (HCl). The streaming potential measurements were conducted in 1 mM potassium chloride (KCl) solution where pH was adjusted by small amounts of 1 M potassium chloride (KOH) or HCl or 36% HCl. The same acids or bases were used to change the pH in the contact angle measurements. The pure water used in all filtration runs and surface characterisations was RO-filtered (Milli-Q) and its conductivity was less than 1 µS/cm. 2.3. Filtration runs The filtration experiments to study the retention properties of the studied membranes as a function of pH were made using a filtration apparatus, which contains three flat-sheet modules in parallel. The apparatus is described in detail elsewhere [3]. The cross-flow velocity (about 4 m/s) was clearly in the turbulent region (the Reynolds number about 12000). The temperature, 40°C, used was near the upper limit of most membranes. The filtrations at 8 bar pressure were made in a total recycling mode by circulating the permeate and the feed back to the feed vessel. The model solution contained 225 ppm glucose and 225 ppm NaCl in pure water. The volume of the feed tank was about 20 L. Before the filtration of model solution the membranes were dipped in 50% ethanol solution for about 5 seconds. After that they were flushed and installed into the modules. The area of each membrane was 0.0046 m2. The

3 experiments were started by pressurising the membranes at a pressure of 20 bar for about 10 minutes. After that the pure water flux was measured over night (about 16 hours) at a pressure of 8 bar. Then salt and glucose were added and the sample (about 30 mL) at the first pH was collected after 15 minutes of filtration. Then the pH was decreased stepwise using HCl and after the lowest pH the apparatus was flushed with water and new salt and sugar solutions were added. The filtration near the original pH was continued over night and the samples were taken after 30 minutes and after about 16 hours of filtration. Then the pH was increased stepwise using NaOH until the highest pH (about pH 11) was obtained and again the apparatus was flushed with water and new salt and sugar were added. The filtration was continued over night and samples were collected after 30 minutes and after about 16 hours of filtration. 2.4. The analyses methods The surface charges (zeta potential) of the membranes were analysed along their surfaces since the membranes studied are tight and do not allow for a free passage of the electrolyte through the membrane. The zeta potentials were calculated from the HelmholtzSchmoluchowski equation without any corrections. The streaming potential measurements were performed in 1 mM KCl solution at 25°C and using reversible ionselective Ag/AgCl electrodes to measure the pressure induced (0.2 – 1.0 bar) potential difference between the ends of the membrane slit (height about 0.4 mm). The streaming potential apparatus is described in detail elsewhere [4, 5]. The contact angles of the membranes with pure water at different pH were analysed using the modified Wilhelmy plate method (wet surface). The apparatus and method are described in more detail elsewhere [6]. First the receding contact angle (velocity of the membrane surface from water was adjusted to 1 mm/min) was measured in pure water without pH adjustment. Then the pH was shifted up or down with 1 M KOH or HCl or 36% HCl. A normalised contact angle was calculated by dividing the contact angle for the changed pH by the one for the unchanged pH, i.e., the first value. The membrane sample was let to reach equilibrium for 30 minutes under mixing before the contact angle was measured. Normally measurements were done shifting pH up first and then shifting it down. A new sample was prepared between the series. The contact angles of dry and clean membrane surfaces against pure water were also measured with the sessile drop method. The results were averages of 10 – 17 independent measurements. The membrane samples for both streaming potential and contact angle measurements were prepared as follows. First the sample was cut to the right size, then it was cleaned in an ultrasonic bath in pure water (about 200 mL) for 3 times 10 minutes changing the water every time and finally the sample was dipped in 20% ethanol and rinsed with pure water to help the surface to wet. Surfaces were not let to dry any more after this preparation (except for the samples for the sessile drop contact angles). Glucose concentration in the samples was analysed as total dissolved carbon (TDC) using a Shimadzu TOC-5050A analyser (standard SFS-ISO-8245). The salt content was analysed measuring the chloride ions by ion chromatography or sodium by atomic absorption spectroscopy. The retentions were calculated as: R% = (1-Cp/Cf)·100%, where Cp and Cf are the concentrations in permeate and feed samples, respectively. 3.

Results and discussion

The retentions and permeabilities of three different NF membranes in different pHs are displayed in Fig. 1. As it can be seen from the figure, the membranes behave somewhat differently. The salt retention of the NF 200 membrane has a sharp minimum at around pH 4. This is also close to the isoelectric point of this membrane (see Fig. 2). The retention of

4 uncharged glucose stays very stable but starts to decrease when the pH reaches a value of about 8. At the same pH the permeability also starts to increase significantly. Also at pH 9 and above a significant change can be seen in the surface hydrophilicity of the NF 200 membrane (see Fig. 3). The same trend with the permeability cannot be seen with the other membranes. The NTR 7450 membrane permeability slightly increases at a more basic pH but the Desal-5 DL membrane has a very stable permeability in the whole pH range. The salt retention of the NTR 7450 membrane shows a most interesting development. The retention is less than 20% when the pH is below 6 and more than 70% if the pH is above 6. The change is striking and sharp. This behaviour can be explained by the charge effect too. NF membranes can retain monovalent ions only due to their surface charge. Moreover, a weak charge may not be enough to cause ions to be retained. The ion retentions with every membrane studied here gains the maximum value when the surface charge is the strongest, i.e., at rather high pH. At lower pH values the surface charges are too weak to cause the ions to be retained except with the NF 200 membrane, which also has a positive charge strong enough at acidic condition. The NTR 7450 and Desal-5 DL membranes have a surface charge strong enough only at pH values about 6 and above. However, the change of the ion retention of the NTR 7450 membrane at pH 6 is surprisingly sharp. 100

50

100

50

20

80

40

60

30

60

30

40

20

40

20

10 Na + C l-

Gluc o s e P e rm e a bility

0

0 4

6

8

pH

Fig. 1.

10

12

14

Retention, %

Permeability, L/(m

Retention, %

20

20

10

20

10

Na+ Cl-

0 0

2

Glucose Permeability

0

0

4

6

pH

8

10 12 14

Permeability, L/(m2h bar)

40

40

Retention, %

30

80

Permeability, L/(m2h bar)

60

h bar)

40

2

50

Desal-5 DL

2

80

0

100

NTR 7450

NF 200

Na+ Cl0

2

Glucose Permeability 4

6

8

0

10 12 14

pH

Glucose and ion retentions as well as permeabilities of the membranes studied at different pH. 40°C, 8 bar.

The changes in permeabilities and retentions were found to be reversible. If the conditions were changed back to neutral after the basic conditions (increased permeability and higher retention of ions or lower retention of glucose), the permeability and retentions were restored. The reversibility was also observed with contact angle measurements. The apparent zeta potentials of the clean surfaces of the three NF membranes are shown in Fig. 2. The NF 200 membrane has much more surface charge than the other membranes. This membrane skin layer is of polyamide material, which possesses dissociable carboxylic and amine groups and therefore can exhibit negative or positive surface charge depending on pH. The NTR 7450 membrane skin layer is sulphonated polyethersulphone that has undissociable sulphone groups in the chemical structure and appears therefore to be much more neutral. This surface can gain some charge through unbalanced adsorption of cations and anions from the surrounding solution. Freger et al [7] suggested that the outmost layer of a polyamide skin layer appears to be carboxylic-rich and separated from a support amine-richer layer. Streaming potential measurements support this conclusion since the NF 200 membrane has a much stronger negative charge at basic conditions than positive charge at acidic condition. Thus, the surface would have higher carboxylic group density than amine group density. The NF 200 and Desal-5 DL membrane skin layers both are from polyamide but probably different monomers or differences in the polymerisation process result in somewhat different

5 surface chemistries. This could be seen in dissimilar ion retention and permeability behaviours as a function of pH as well as they have slightly different surface charges and hydrophilicities. The NF 200 membrane probably has a higher density of dissociable groups on the skin layer or the distribution of the carboxylic and amine groups is different from that on the Desal-5 DL membrane skin layer.

Zeta potential, mV

10 5

pH

0 0

4

8

-5

NF 200

-10

NTR 7450

-15

Desal-5 DL

-20

Fig. 2.

Surface charges (apparent zeta potentials) of the NF membranes studied.

NF 200

1.5 1.0 0.5 0.0

Fig. 3.

0

2

4

6 8 10 12 pH

2.0 1.5 1.0 0.5

NTR 7450

0.0 0

2

4

6 8 10 12 pH

Normalised contact angle

2.0

Normalised contact angle

Normalised contact angle

The contact angle of a surface against water reflects its wettability. When water wets the surface, i.e., a small contact angle, the surface has the ability to interact with water molecules (dipoles). Dissociable groups on a surface help it to interact with water molecules and make the surface more hydrophilic. This is confirmed in Fig. 3 and Table 1 where the sessile drop contact angle of the NF 200 membrane is much smaller than with the Desal-5 DL membrane. The NTR 7450 membrane shows the most hydrophopic surface. Sulphone groups should also have some capability to interact with water molecules. Also, in Fig. 3 it can be seen that the dissociation of carboxylic and amine groups on the surface of the NF 200 membrane results in an increase in hydrophilicity when pH is increased or decreased. Polyethersulphone is a rather hydrophobic material and adding sulphone groups should make it more hydrophilic. The sessile drop contact angle of the NTR 7450 membrane surface shows a rather hydrophobic character despite of sulphonation. The contact angle of the NTR 7450 membrane indicates a hydrophobic surface also as a function of pH (see Fig. 3). 2.0 1.5 1.0 0.5

Desal-5 DL

0.0 0

2

4

6 8 10 12 pH

Normalised contact angles of the NF membranes with pure water at different pH. Receding contact angles (speed of rod movement = 1 mm/min) of the modified Wilhelmy method.

The contact angle has often been criticised as a surface characterisation method of membranes since membrane surfaces usually have pores and they are rough. The surfaces of NF membranes have proven to be very smooth and practically without pores. Therefore, the contact angle measurements would reflect well the hydrophilicity of an NF membrane surface. Strongly charged surfaces like the polyamide surface of the NF 200 membrane have a weaker charge at acidic pH and the ions can pass the membrane where as uncharged glucose is retained due to size. In a basic solution this surface has a more negative charge and ions are

6 retained better. When the pH is shifted to strongly basic the carboxylic groups at the surface dissociate fully and the surface gains its strongest negative charge. However, the negative charges on the polymer chains in the three-dimensional network of the surface start to repel each other and make the skin layer looser. The free volume in the network increases and the ‘pore size’ of the membrane skin layer increases. Also, flux increases and the glucose molecules can utilise the free space between polymer chains and pass the membrane. If the membrane skin layer does not have dissociable groups like the NTR 7450 membrane, then this kind of behaviour cannot be observed. 4.

Conclusions In this paper filtration results of different commercial NF membranes with different skin layer materials is discussed when pH was changed between 2 and 11. The retentions of ions and uncharged glucose changed significantly especially when the pH was shifted to more basic. Also, fluxes developed very differently with different NF membranes when pH was gradually increased. This surprising behaviour was explained by the different surface chemistries of the membranes. If a membrane surface possesses dissociable groups then the membrane can become more open at elevated pHs. Normally a significant decrease in retention would be undesirable but can be an advantage in some cases such as in separation of small, uncharged molecules from ions with strongly charged membranes. Also the fluxes of these kinds of membranes can often be improved significantly by adjusting the pH to more basic. The fouling tendency might decrease at the same time since the surface of a charged membrane would become more hydrophilic. Thus increasing pH could be used as an in situ surface modification method. Moreover, two general guidelines for the membrane choice for different filtrations could be given. If the feed solution contains uncharged organic molecules, the retention of which is desired, then a less charged membrane could be chosen. However, if the membrane needs to retain ions, a membrane with a strong surface charge should be used or the conditions should be adjusted so that the membrane surface gains its strongest charge. Acknowledgements Technicians Anne Kokko and Helvi Turkia are thanked for their help and hard work with various analyses. MSc-student Heli Kumpulainen is acknowledged for performing the filtration runs and sample analysis as well. This research was financially supported by the Academy of Finland (project number 75922). References [1] Mänttäri, M., Pihlajamäki, A. and Nyström, M., Comparison of nanofiltration and tight ultrafiltration membranes in the filtration of paper mill process water, Desalination, 149 (2002) 131-136.

[2] Ikeda, K., Nakano, T., Ito, H., Kubota, T. and Yamamoto, S., New Composite Charged Reverse Osmosis Membrane, Desalination, 68 (1988) 109-119.

[3] Mänttäri, M., Nyström, M., Critical flux in NF of high molar mass polysaccharides and effluents from paper industry, J. Membrane Sci., 170 (2000) 257-273.

[4] Nyström, M., Pihlajamäki, A. and Ehsani, N., Characterization of ultrafiltration membranes by simultaneous streaming potential and flux measurements, J. Membrane Sci., 87 (1994) 245-256.

[5] Pihlajamäki, A., Electrochemical characterisation of filter media properties and their exploitation in

enhanced filtration, Lappeenranta University of Technology, Research papers 70, Doctoral Dissertation, Lappeenranta, Finland 1998, 92 pages.

[6] Palacio, L., Calvo, J.I., Prádanos, P., Hernándes, A., Väisänen, P. and Nyström, M., Contact angles and external protein adsorption onto UF membranes, J. Membrane Sci., 152 (1999) 189-201.

[7] Freger, V., Pihlajamäki, A., Shabtai, Y. and Gilron J., Distribution of fixed charge functional groups in the polyamide composite membranes, Supplementary book of abstracts, ICOM’02, 07.-12.07.2002, Toulouse, France, p. 144.