Cleaning results of new and fouled nanofiltration membrane characterized by contact angle, updated DSPM, flux and salts rejection

Available online at www.sciencedirect.com

Applied Surface Science 254 (2008) 3983–3992 www.elsevier.com/locate/apsusc

Cleaning results of new and fouled nanofiltration membrane characterized by contact angle, updated DSPM, flux and salts rejection Ahmed Al-Amoudi a,b,*, Paul Williams a, A.S. Al-Hobaib c, Robert W. Lovitt a a

Centre for complex fluids processing, Multidisciplinary Nanotechnology Centre, School of Engineering, University of Wales, Swansea SA2 8PP, UK b Saline Water Conversion Corporation (SWCC), Saline Water Desalination Research Institute Staff, Saudi Arabia c Institute of Atomic Energy Research, King Abdulaziz City for Science And Technology, P.O. Box 6086, Riyadh 11442, Saudi Arabia Received 29 August 2007; received in revised form 14 December 2007; accepted 17 December 2007 Available online 9 January 2008

Abstract In membrane process industries, membrane cleaning is one of the most important concerns from both economical and scientific points of view. Though cleaning is important to recover membrane performance, an inappropriate selection of cleaning agents may result into unsatisfactory cleaning or irreparable membrane. In this study the cleaning performance has been studied with measurements of membrane contact angle, Updated Donnan steric partitioning pore model (UDSPM) and salt rejection as well as flux measurement. Thin film nanofiltration (NF) membranes such as DK, HL and DL provided by GE Osmonics are used in this study. Tests were carried out with virgin DK, HL and DL as well as fouled DK membranes. Several cleaning agents were investigated; some of them were analytical grade such as HCl, NaOH and others such as SDS, mix agents were commercial grade agents that are already in use in commercial plants. Contact angle, DSPM and salt rejection as well as flux of virgin and fouled membranes before and after chemical cleaning were measured and compared. The contact angle measurements with and without chemical cleaning of different virgin and fouled membranes revealed very interesting results which may be used to characterise the membrane surface cleanliness. The contact angle results revealed that the cleaning agents are found to modify membrane surface properties (hydrophobicity/hydrophilicity) of the treated and untreated virgin and fouled membranes. The details of these results were also investigated and are reported in the paper. However, UDSPM method did not give any valuable information about pore size of the untreated and treated NF membranes. The salt rejection level of monovalent and divalent ions before and after cleaning by high and low pH cleaning agents is also investigated and is reported in the paper. Crown Copyright # 2007 Published by Elsevier B.V. All rights reserved. Keywords: Cleaning; Contact angle; Fouling; Nanofiltration membranes; Cleaning agents

1. Introduction The separation characteristics of nanofiltration stand between ultrafiltration (UF) and reverse osmosis (RO) and the membrane selectivity has often been attributed to the interchange of both molecular sieving properties of ultrafiltration and diffusion properties of RO. It is generally recognized that membrane hydrophobicity/hydrophilicity, pore size (and its distribution) and surface charge can be important factors in the separation performance and fouling tendency of nanofiltra* Corresponding author at: Centre for complex fluids processing, Multidisciplinary Nanotechnology Centre, School of Engineering, University of Wales, Swansea SA2 8PP, UK. Tel.: +44 7886664882; fax: +44 1792539029. E-mail addresses: [email protected], [email protected] (A. Al-Amoudi).

tion membranes [1,2]. The interaction of organic and inorganic colloidal substances with membrane surfaces in aqueous media is also an important factor which is dependent not only on the membrane surface charge but also on the hydrophobicity/ hydrophilicity of the surface and pore size and pore size distribution. Therefore, study of membrane surface characteristics and environment are critical to understand and controlling membrane fouling and cleaning processes. A large number of chemical cleaning agents are commercially available; five categories of cleaning agent commonly used are: alkalis, acids, metal chelating agents, surfactants and enzymes [3,4]. Commercial cleaning products are usually a mixture of these chemicals but the actual composition is often unknown. The choice of the optimal cleaning agent or mix composition depends on the feed characteristics. For example, acid cleaning is suitable for the removal of precipitated salts,

0169-4332/$ – see front matter. Crown Copyright # 2007 Published by Elsevier B.V. All rights reserved. doi:10.1016/j.apsusc.2007.12.052

3984

A. Al-Amoudi et al. / Applied Surface Science 254 (2008) 3983–3992

such as CaCO3, while alkaline cleaning is used to remove adsorbed organics [5]. Flux measurement is a direct assessment of fouling and cleaning process. It is usually accepted that flux decline in aqueous solutions containing organic and inorganic molecules is mainly caused by adsorption or crystallization, possibly enhanced by pore blocking and/or cake formation [6,7]. Effective chemical cleaning would therefore be necessary to detach different classes of foulants from the membrane and restore its permeate flux characteristics [8]. Furthermore, the selection of appropriate chemical cleaning agents might be critical because incompatible combinations of cleaning agent and membrane material could lead to irreversible flux loss, unnecessary costs through excessive chemical use and reduction in membrane life [3,8]. A contact angle measurement provides information on the hydrophobicity and hydrophilicity of the membrane surface. The hydrophobicity/hydrophilicity characteristics of the membrane surface are expected to govern the surface wettability by liquids, especially water, and subsequently govern the membrane performance in various applications. A number of researchers have tried to find a way to express hydrophobicity in a quantitative way using contact angle measurement. The effect of surface characteristics has generally been evident in efforts to select optimal pretreatment schemes and operating conditions for various membrane separation processes. However, the characterization of NF membrane surface by contact angle measurement after cleaning is seldom available in literature. In order to predict the membrane separation performance, it is often necessary to know the mean pore size and pore size distribution at the membrane surface. Bowen et al. [9,10], developed a method called Donnan steric partitioning pore model (DSPM), which is based on the extended Nernst–Planck equation, modified to include steric effects, and a modified Donnan equation accounting for the sieving effect as well. Theoretical background description of the DSPM has been given in detail elsewhere [9,10]. The DSPM has been successfully used to describe retention of neutral solutes and ions on NF and Ultrafiltration (UF) membranes and on titania membrane with nanofiltration membrane properties use [11,12]. In this work, an updated version of the DSPM model has been used [13,14] using this model has the benefit that the membrane thickness is removed from the rejection equation because the nanofiltration driving force is redefined in term of an effective pressure as opposed to the volumetric flux. As a direct result the rejection equation for uncharged solute is now only dependent on the pore radius. Therefore, fitting experimental rejection at various applied effective pressures for a solute of known size will directly characterize the membrane pore radius. It is well established that these chemical cleaning agents such as SDS, EDTA, NaOH and HCl are widely used in the applied industrial plants. The major message of this work and previous works is to open the discussion of cleaning strategies to those who have applied these cleaning agents in their plant in order to restore the production. The cleaning processes do

regenerate the quantity but simultaneously deteriorate the quality. Therefore, the efficiency of cleaning agents, which have been used widely in the industrial plants, has to be investigated thoroughly in the laboratory scale before being applied in industrial plants. This study is intended only to evaluate the validity of these techniques, whether it is possible to obtain useful information about the membrane surface after cleaning or not, and also to find out whether a correlation exists between the flux and/or salt rejection and the data obtained from these techniques. If it does, then subsequently the variation of cleaning concentration and the cleaning process will be worthy of study in order to find out what is the optimum cleaning system with the aim of restoring the plant performance without deteriorating the water quality. Thereby the cost of the membrane replacement can be saved and membrane life can be extended. Kim and Fane [15] studied cleaned and fouled UF membranes using flux methods combined with microscopical methods. The results were also analyzed by determining the pore size distributions before and after cleaning. Warczok et al. [16] recently reported that it is possible to find out whether the cleaning procedure was correctly designed or not by analysis of AFM (Atomic Force Microscope) images. In order to evaluate the cleaning results it is important to compare the characterisation of the fouled membranes to those of the virgin and cleaned membranes. Fouled membranes have been investigated in many different ways using various characterisation methods. In most cases, the virgin membranes were cleaned before use and the characteristics of the membrane after precleaning could have changed considerably from that of the virgin membranes. In this study, different kinds of analytical grade and commercial grade cleaning agents have been used to clean both virgin and fouled NF membranes. Both virgin and fouled membranes have been characterized before and after cleaning by contact angle, use of the UDSPM, flux measurements and rejection of monovalent and divalent ions. The overall goal of this study was to investigate the following: (a) To investigate different techniques (using contact angle measurements, UDSPM, rejection and flux) to find out the most suitable option for restoring the membrane flux without harming the quality of the product water. (b) To compare between untreated and treated NF membranes on the basis of the contact angle measurements, UDSPM, rejection and flux. 2. Experimental 2.1. Materials 2.1.1. Membrane Three types of commercial NF membranes, GE OsmonicsDK, DL and HL flat sheet membrane were used in this study. DK, DL and HL thin-film membranes (TFM) are negatively charged having a proprietary active nanopolymer layer based

A. Al-Amoudi et al. / Applied Surface Science 254 (2008) 3983–3992

on polypyperazinamide. The top active layer of DK and DL consists of three sub-layers; while the HL top active layer is composed of two sub-layers (these TFM were provided by GE Osmonics, FA, USA). However, the molecular cut off for all NF membrane is 150–300 Da [17]. Fouled NF Osmonics-DK membranes, which were used in this study, were obtained from Umm-Lujj SWRO plant of Saline Water Conversion Corporation (SWCC) where the autopsy was carried out on the lead element of DK membranes (taken from a commercial plant using seawater feed) after about 9000 h of continuous operation of NF commercial plant. Visual inspection of the NF lead element revealed no physical damage to the membranes. Thick reddish brown deposits were found on the entire surface of the lead elements, indicating that the lead element was fouled. The deposits were easily scraped off from the surface. The amount of foulant per unit area of the lead element was about 0.52 mg/cm2. The deposit analyses results revealed that natural organic matter was the main constituent of the foulant material. More details of the plant operation and DK membrane autopsy results can be found elsewhere [18]. Prior to the experiments, all virgin NF and fouled membranes were soaked over night with or without cleaning agents in high purity water obtained from Millipore Elix 3 unit. 2.1.2. Chemicals Solutions of the polyethylene glycol (PEG) with molecular weight of 200 Da at a concentration of 1 g/l were filtered. PEG was used as a representative uncharged solute in order to determine the membrane pore radius. The concentrations of the product, feed and retentate were determined by high performance liquid chromatography (HPLC)–refractive index (RI) measurements. The analytical grade salts used in these experiments are monovalent sodium chloride (Sigma–Aldrich) and divalent magnesium sulphate-7-hydrite (Sigma–Aldrich). The salt concentration ions of the feed and product were measured via conductivity. High purity water obtained from a Millipore Elix 3 unit (at conductivity less than 1 mS cm1) was used in these experiments at pH of 5.8. 2.1.3. Cleaning agents Two analytical grade chemicals (HCl and NaOH) and two commercial grade chemicals SDS (0.1%) and mixture of trisodium phosphate (0.1%), sodium tripolyphosphate (0.1%) and EDTA (0.1%) (TSP + EDTA + STP) were used as cleaning agents in the experiments. Table 1 shows the name and

3985

concentration of cleaning agents used in the tests. The commercial cleaning agents are similar to that used in the Umm-Lujj commercial NF plant [18]. 2.2. Methods 2.2.1. Cleaning procedure New membranes were used for each set of experiments. In one set of experiments, the virgin membranes were rinsed several times in demineralised water in order to remove the impurities from the virgin membrane. The bench scale experiments were adopted in this study. Several sets of experiments with different cleaning agents for virgin and fouled membranes were soaked in cleaning agents over night (18 h) and then stirred for 90 min. The treated virgin and fouled membranes were soaked few hours and properly rinsed using pure water in order to remove the remaining cleaning agents from membrane surface. 2.2.2. Contact angle measurements The hydrophobicity/hydrophilicity of several samples of virgin and fouled membrane was estimated via contact angle measurement using the Fibro DAT 1100. The treated and untreated membranes were air dried at ambient temperature prior to contact angle measurements by sessile drop technique [19,20]. Reported contact angles are the average of 7–9 measurements using pure water drops. The contact angle measurement was carried out for treated and untreated and virgin or fouled NF membranes and repeated three times using another new set of membranes to check the repeatability of results. The measurements were conducted separately for each cleaning agent with fresh membrane. The standard deviation of the results was around 10%, which is well within the range expected 15% of experimental errors. Details of the contact angle measurement procedure can be found elsewhere [21]. 2.2.3. PEG retention In order to characterize the membranes, effective mean pore size and molecular weight cut off pore sizes of the treated and untreated membranes were determined using model uncharged PEG solution on the theoretical basis of the UDSPM. Permeation experiments with a PEG concentration of 1 g/l were carried out using a 200 ml stirred Amicon cell (Model 8200) and detailed descriptions of the instrument can be found elsewhere [15]. The applied pressure was maintained at 400 kN m2.

Table 1 Chemical cleaning concentration Cleaning agents

Concentration

pH

Grade

1. 2. 3. 4. 5.

1M 2M 0.1% 0.1%

2.5 11.3 11.3 11.3

Analytical Analytical Commercial Commercial

HCl NaOH SDS Mixed agent of EDTA, TSP and STP

Note

Similar to that used in actual NF-SWRO plant at Umm-Lujj Each cleaning takes one day

3986

A. Al-Amoudi et al. / Applied Surface Science 254 (2008) 3983–3992

2.2.4. Flux and rejection measurements In order to characterize the cleaning efficiency, both the flux and the rejection of the monovalent (NaCl at 2482 ppm) and divalent (MgSO47H2O at 2482 ppm) salts were measured. Permeation and rejection experiments with monovalent and divalent salts solutions in demineralised water were carried out using a laboratory-scale cross-flow recirculation test unit. The test unit consisted of a rectangular membrane cell, feed reservoir, variable speed gear pump, flow meter for measuring cross-flow, balance for measuring filtrate flow, pressure transducers, temperature control system and computerised data logging system. The detailed descriptions of the unit and instrumentations can be found elsewhere [22]. The pressure of the mean of the inlet and outlet was maintained at 400 kN m2 and cross-flow rate was maintained at 0.35 l/min by variable speed drive. All experiments were carried out at a temperature of 25 8C. 2.2.5. Protocol of model flux and rejection measurement A new piece of membrane was placed in the cell and water was then passed through the cell to wash out any impurities that might be present. This procedure was repeated twice. The demineralised water was circulated through the membrane until a steady flux was obtained. Then, MgSO47H2O model solution at 2464 ppm concentration was passed through the membrane. The permeate flux and rejection were measured after 20 ml of permeate passed through the membrane. Then, the cell was washed thoroughly by circulating fresh water under the same pressure and then similar experiments were repeated using monovalent model solution of NaCl. This procedure was repeated once or twice with other similar fresh virgin membrane. 3. Results Untreated virgin DK, DL and HL TFM and fouled DK TFM, were characterized using contact angle measurement, pore size measurement via the UDSPM and flux as well as rejection measurements. Treated virgin NF TFM and fouled DK TFM were also evaluated and characterized by the contact angle, flux, UDSPM measurement and salt rejection. 3.1. Flux measurements 3.1.1. Flux measurements on untreated virgin NF TFM The pure water fluxes of treated and untreated NF membranes at 25 8C are shown in Table 2. The virgin untreated HL TFM exhibited highest flux for pure water (65.2 kg/m2 h) compared to other untreated NF membranes, while the untreated DK membrane gave lowest flux of pure water (10.5 kg/m2 h). The virgin DL membrane flux was 20 kg/m2 h and the fouled DK TFM did not show any significant flux as a result of external blocking of cake formation. The flux trends of the salt solutions for these NF membranes were found to be similar to that of the pure water for all untreated NF TFM (see Tables 3 and 4).

Table 2 The pure water flux for treated and untreated virgin membrane and fouled NF membranes Virgin DK (kg/m2 h)

Fouled DK (kg/m2 h)

Virgin HL (kg/m2 h)

Virgin DL (kg/m2 h)

Untreated membrane

10.5

Nil

65.2

20.0

Treated membrane by HCl NaOH Mix SDS

16.3 20 27 30.3

4.5 10.8 14.2 29.6

46.5 46.2 54.9 43.8

14.5 24.5 22 33

3.1.2. Flux of cleaned virgins NF TFM The results of pure water flux of cleaned NF TFM at 25 8C are illustrated in Table 2. A common trend of increase in flux was found for DK membranes when cleaned by high and low pH cleaning agents. For cleaned DK membranes with pure water, the highest flux (30.3 kg/m2 h) was achieved by treating the membranes with SDS cleaning agent and lowest (16.3 kg/ m2 h) by using HCl cleaning compared to untreated membrane flux that was 10.5 kg/m2 h. Salt model solutions were also found to exhibit higher fluxes with cleaned DK membranes (see Tables 3 and 4). Performance of DL membranes was found to be different depending on the cleaning agent used (Tables 2–4). Tests with pure water on DL membrane treated by NaOH and SDS cleaning agents showed higher fluxes (24.5 and 33 kg/m2 h, respectively) than the flux of the untreated DL membrane (19.9 kg/m2 h). While DL membrane treated by HCl gave a lower flux (14.5 kg/m2 h) and mix agents gave slightly high flux (19.5 kg/m2 h) compared to untreated flux. Similar trends of flux were noticed for treated and untreated DL Table 3 The model solution flux of NaCl for treated and untreated virgin membrane and fouled NF membranes Virgin DK (kg/m2 h)

Fouled DK (kg/m2 h)

Virgin HL (kg/m2 h)

Virgin DL (kg/m2 h)

Untreated membrane

6.55

Nil

59.5

18.6

Treated membrane by HCl NaOH Mix SDS

10.3 16.2 20.85 21

5.9 17.1 25.6 26.2

39 43.1 45.6 41

13.5 19.1 18 29.3

Table 4 The model solution flux of MgSO47H2O for treated and untreated virgin membrane and fouled NF membranes Virgin DK (kg/m2 h)

Fouled DK (kg/m2 h)

Virgin HL (kg/m2 h)

Virgin DL (kg/m2 h)

Untreated membrane

5.6

0

57.2

14.2

Treated membrane by HCl NaOH Mix SDS

9.2 14.5 17.6 16.1

3.3 13.3 19.1 16.9

32.7 30.7 37.5 32.7

13.6 10.6 11.7 21

A. Al-Amoudi et al. / Applied Surface Science 254 (2008) 3983–3992

3987

Table 5 The effective pore size treated and untreated virgin membrane and fouled NF membranes Type of membrane

New membrane (nm)

Cleaned by NaOH (nm)

Cleaned by HCl (nm)

Cleaned by SDS (nm)

Cleaned by MIX agents (nm)

DK HL DL

0.65 0.64 0.7

0.61 0.66 0.73

0.72 0.65 0.68

0.61 0.66 0.70

0.63 0.66 0.68

membranes during tests with monovalent and divalent salt solutions. 3.1.3. Flux on cleaned fouled DK TFM The fluxes of pure water and the solution models (of both monovalent and divalent salts) during tests on fouled DK membranes after cleaning with high pH cleaning agents (NaOH, Mix agents and SDS) had remarkable increment of flux compared to untreated virgin membrane (control). While cleaning with low pH cleaning agent (HCl), the fluxes of pure water and model salts solution were slightly higher than the control [1,24,9] (Tables 2–4). 3.2. UDSPM In order to calculate the effective pore size of the membranes used, the updated DSPM technique was employed [13,14]. The solution of the UDSPM model for neutral solutes is completely analogous in approach to that of the DSPM model [7,8]. However, in this case the resulting equation derived is: 

Ci p R¼1 C iw R¼



Fi K ic ð1  ðFK ic =2ÞÞ þ ðFK ic =Pe0 Þ

(1)

(2)

where the modified Peclet number (Pe0 ) is Pe0 ¼

K ic V Dx K ic rp2 ¼ DPe K id Di1 K id 8hDi1

For a full derivation see refs. [13,14]. This result is important as redefining the nanofiltration driving force in terms of the effective pressure DPe removes the membrane thickness from the rejection equation and as a result this equation is now only dependent on rp. Therefore, fitting experimental rejection data, various effective pressure for an uncharged solute of known size will give a direct characterization of the membrane pore radius. In this work PEG 200 was employed as the uncharged solute fitting the experimental data for the rejection of PEG various pressure resulted in fitted values for the effective mean pore size for the membranes as displayed in Table 5. Table 5 indicates that there is no significant charge in the effective mean pore size due to chemical cleaning. Any differences in pore size are too insignificant to be able to discriminate effectively between them.

3.3. Contact angle (CA) 3.3.1. Contact angle measurements on untreated virgin and fouled NF TFM The contact angle is an important parameter in measuring the surface hydrophobicity. In general, the higher the contact angle the higher is the hydrophobicity of the membrane surface [23,24]. It can be seen from Table 6, the contact angle measurement of the smoothest NF HL TFM gave a value of 56.7 compared to the untreated virgin NF DK and DL TFM membranes of 45.1 and 51, respectively. However, the contact angle of the untreated fouled NF DK TFM was higher than the untreated virgin DK TFM by 20.2. The high contact angle indicates that the fouled membranes are likely to be fouled with hydrophobic/hydrophilic natural organic matter (NOM) [25]. 3.3.2. Effect of cleaning virgin NF membrane on contact angle The changes in hydrophobicity/hydrophilicity for virgin DL, HL and DK TFM after cleaning with various cleaning agents are shown in Table 6. In the case of DL TFM the contact angle measurements for both NaOH and HCl cleaning were found to give slightly higher contact angles and thus higher hydrophobicity than untreated membrane. Whereas the contact angle values for the DL TFM cleaned by Mix or SDS cleaning agents were found to exhibit more hydrophilic behavior with CA values of 37.1 and 27.6, respectively, compared to untreated membrane with contact angle value of 51. In the case of HL TFM the contact angle measurements for all NaOH, HCl, Mix and SDS cleaning agents show that the membrane becomes more hydrophilic than the untreated virgin HL TFM. However, HL TFM cleaned with SDS had the lowest contact angle of 26.6 compared to the untreated HL TFM (56.7) or treated HL TFM of NaOH (40.8), HCl (45.9) and Mix (42.4) cleaning agents. Table 6 Contact angle measurements for treated and untreated virgin membrane and fouled NF membranes DK, u

Fouled DK, u

HL, u

DL, u

Untreated membrane

45.1

65.3

56.7

51

Treated membrane by HCl NaOH Mix SDS

45.6 39.45 34.85 28

53.7 47.8 26.9 25.2

45.9 40.8 42.4 26.6

55 55.1 37.1 27.6

3988

A. Al-Amoudi et al. / Applied Surface Science 254 (2008) 3983–3992

The cleaning of virgin DK TFM with cleaning agents such as, mixed agent and SDS, resulted in lower contact angle values than those measured for the untreated membrane. Whilst using HCl chemical cleaning for the DK TFM results is almost the same contact angle to the untreated virgin DK TFM (45.1) (Table 6). Conversely, the cleaning of the virgin DK TFM with NaOH cleaning agent gave a more hydrophilic membrane with a contact angle value of 39.45 compared to the untreated virgin DK TFM. In case of DK TFM, SDS cleaning gave the lowest contact angle. The hydrophobicity/hydrophilicity of the NF TFM virgin membrane varied due to cleaning by different cleaning agents. The maximum difference of contact angle values between treated and untreated virgin NF TFM was observed with SDS chemical cleaning. Table 6 shows that the values of the contact angle for all NF virgin TFM cleaned with SDS changed to give more hydrophilic membrane when compared to untreated virgin TFMs. This occurred with the mix cleaning agents as well. The surface hydrophobicity/hydrophilicity of the virgin and fouled membranes have changed after cleaning processes. In the case of virgin HL TFM cleaned by SDS and mixed cleaning agents, the membrane surface was found to exhibit more hydrophilic behavior than the untreated TFM. 3.3.3. Effect of membrane fouling on contact angle Table 6 shows that contact angle values for the fouled DK TFM, cleaned by NaOH and HCl exhibit a strong similarity to those obtained for the untreated virgin DK TFM compared to SDS, mixed cleaning agents. The graph also illustrates that where the fouled DK TFM is cleaned with SDS, the contact angle decreased drastically from 65.3 to 25.2 (a value even lower that the virgin treated membranes). In the case of the fouled DK TFM membrane cleaned by mix agents, lower contact angle values were found to give a more

hydrophilic surface when compared to untreated virgin DK TFM and/or those cleaned by HCl and NaOH. However, the fouled DK TFM cleaned by mix agents were found to exhibit less hydrophilic surface behavior than those cleaned by SDS agent. 3.4. Rejection measurements Generally, when dealing with loose NF membranes, size exclusion is more important in understanding the rejection of uncharged solute molecules, and electrostatic repulsion is more important for the rejection of ionic species. 3.4.1. Rejection on untreated NF TFM Fig. 1 shows that the rejection of MgSO47H2O for the DL TFM was 97.6% compared to 79% for the HL TFM. The rejection of MgSO47H2O for the DK TFM was 96.5% compared to nearly null for DK TFM fouled membrane. The rejection of NaCl for the virgin DK TFM exhibited the highest rejection of 40% compared to HL and DL TFM (see Fig. 2), whereas the smoothest virgin HL TFM has the lowest rejection of both monovalent (NaCl) and divalent ions (MgSO47H2O). 3.4.2. Rejection on cleaned virgins NF TFM The rejections of monovalent ions (NaCl) for cleaned DK TFM, with all of the chemical cleaning agents exhibited slightly higher values than the untreated membrane. In the case of DL NF TFM cleaned by the various cleaning agents (HCl, NaOH, Mix agents and SDS) the rejection of monovalent and divalent ions was found to exhibit almost the same value to that of the untreated virgin membrane (Figs. 1 and 2). In the case of NF HL membranes the rejection of monovalent (NaCl) for the cleaned membrane with these cleaning agents exhibited a strong similarity to the untreated membrane and the rejection of divalent ions cleaned by HCl

Fig. 1. The rejection of MgSO47H2O for treated and untreated virgin membrane and fouled NF membranes.

A. Al-Amoudi et al. / Applied Surface Science 254 (2008) 3983–3992

3989

Fig. 2. The rejection of NaCl for treated and untreated virgin membrane and fouled NF membranes.

exhibited almost the same rejection as that of untreated membranes (Figs. 1 and 2). 3.4.3. Rejection measurements on cleaned fouled DK TFM The rejection of monovalent or divalent ions for fouled DK NF membranes after cleaning with high pH (NaOH, mix agents and SDS) and low pH (HCl) cleaning agents were found to be lower than the control (Figs. 1 and 2). All chemical cleaning agents restored the flux to even higher values than control with salt passage being higher than the control. 4. Discussion Usually the efficiency of a cleaning procedure is evaluated only by flux measurements. However, this gives no information on membrane surface properties. Contact angle measurement, pore size and salt rejection are able to give additional useful information on the surface state of the membrane. The impact of the cleaning procedure on the membrane surface can therefore be evaluated. The surface hydrophobicity/hydrophilicity of the virgin and fouled membranes have changed after cleaning processes. In the case of virgin NF TFM cleaned by SDS and mixed cleaning agents, the membrane surface was found to exhibit more hydrophilic behavior than the untreated virgin TFM. This could be due to adsorption phenomena. It is evident that the cleaning agents can remain to membrane surfaces due to negative functional group of cleaning agents and can also interact with membrane surface charge by hydrophobic interaction and the electrostatic interaction as well. The surface hydrophobicity/hydrophilicity are not only dependent on the use of cleaning agents but also due to adsorption phenomena on the coverage of foulants. However, in the case of fouled DK TFM where the fouled DK TFM is cleaned with SDS, the contact angle decreases drastically from

65.3 to 25.2 (a value even lower that the virgin treated membranes). This suggests that it is possible that the cleaning process altered the membrane surface state due the cleaning solution at pH 11.3 which is above the limit recommended by the membrane manufacturer [1]. A similar trend was observed for UF membranes and can be found elsewhere [26]. The SDS cleaning agent might have removed the hydrophobic NOM fraction DK NF TFM leaving the hydrophilic NOM fraction intact. Moreover, it is understood that SDS was readily adsorbed to the membrane surface; however the extent of contact angle value variation depends on the reaction between the membrane and the cleaning agents’ type. This finding reveals that the SDS cleaning agent plays a significant role of adsorption in the surface state [1]. Different changes in hydrophilicity (both increase and decrease) are related to differences between the membrane materials properties. The material of the top active layer of the HL TFM consists of two sub-layers while DL and DK TFM consist of three sub-layers that could play a role in adsorption phenomena [1]. However, from the data of contact angle obtained in the case of HL TFM treated by high and low pH cleaning agents proved that the membrane became more hydrophilic after treatment compared to the untreated virgin HL TFM. The surface charge of the treated HL TFM also exhibited more negative value compared to the untreated HL TFM [1]. Virgin DL and DK membranes cleaned with high pH cleaning agents (SDS and mix) also exhibited a hydrophilic surface state indicating the adsorption of SDS and mix cleaning agents. This suggests that the third sub-layer of the DL and DK TFM could have slightly resisted adsorption phenomena [1]. In order to calculate the effective pore size of the membranes used, the updated DSPM technique was employed [13,14]. In this case PEG 200 was employed as the uncharged solute and the experimental data for the rejection of PEG 200 at various pressures was fitted to the model. The results are displayed in

3990

A. Al-Amoudi et al. / Applied Surface Science 254 (2008) 3983–3992

Table 5 and they indicate that there is insignificant change in the effective mean pore size due to chemical cleaning. Any differences membranes in pore size are too insignificant to be able to discriminate effectively between them. This limitation could be attributed to either adsorption of the cleaning agents within the pores or the membrane surface charge [1]. A cleaning efficiency of over 100% at high pH, when the recovered flux was higher than the flux of the virgin untreated membrane, is probably due to the presence of a small amount of negatively charged cleaning agents on the membrane surface, making the membrane more hydrophilic, and thus, enhancing the partitioning and passage of water molecules. Similar trends of increase in flux after cleaning were observed for reverse osmosis membrane [27]. The adsorption characteristics of surfactants are governed by the molecular structure of the surfactant molecules (e.g., type of polar head, structure and the length of hydrocarbon chain) and the characteristics of the membrane surface (e.g., charge, hydrophobicity) [28]. At high pH chemical cleaning, the membrane is negatively charged and adsorption occurs due to hydrophobic interaction between the surfactant tail and the membrane surface [29]. During tests with all water solutions, DL membranes cleaned with SDS exhibited a higher flux when compared to the untreated membrane (Tables 2–4). This revealed that the SDS cleaning agent had an effective influence on the membrane surface state. It also concludes that the membrane exhibited a hydrophilic surface state after cleaning with SDS. This finding is supported by contact angle measurements for virgin DL membranes of 56 compared to 27 for DL membrane cleaned by SDS [29]. The HL membranes exhibited higher flux than any other NF membrane during tests with all water solutions, the fluxes of membranes treated with high and low pH cleaning agents were found lower than that of untreated HL membrane. It is expected that due to the presence of carboxyl and amine functional groups in HL membranes, the surface charge of the membrane could affect the flux at both high and low pH. At high pH, the carboxyl groups would be deprotonated (COO–), and at low

pH, the amino groups would be protonated (NH3+). In both cases, the electrostatic repulsion between the charged groups could have an effect on the surface charge of the membrane and thereby causes a decrease in flux [30]. The electroviscous effect, which is a physical phenomenon that occurs when an electrolyte solution is pressed through a narrow capillary or pore with charged surfaces, is most pronounced when double-layer effects are significant. At high pore surface charge, the permeate solution appears to exhibit an increased viscosity when its flow rate is compared with flow at low pore surface charge. Nevertheless, NF HL TFM after cleaning with these agents has a higher surface charge due to higher adsorption than DK and DL TFM [1]. The flux would be at a maximum when the capillary is uncharged [29]. Accordingly, the water flux of DL and DK membranes would be higher than HL as the capillary has a lower charge in case of DK and DL membranes when compared to HL membrane. In the case of fouled DK TFM, The fluxes of pure water and the salt solutions after cleaning with high pH cleaning agents (NaOH, Mix agents and SDS) showed remarkable increase in flux when compared to the control. The recovered flux of salt solutions after high pH cleaning agents were used was higher than the flux of the model salt solutions for both treated and untreated virgin DK membranes, probably due to the presence of a small amount of NOM (hydrophilic) and/or high cleaning agents on the membrane surface, making the membrane more hydrophilic, and thus, enhancing the partitioning and passage of water molecules. In addition, the presence of the hydrophilic part of the NOM could have reduced the magnitude of the pore charge resulting in superior flux (electroviscous effect). The flux results of the fouled DK membranes before and after cleaning showed that the high pH cleaning agents performed well in cleaning fouled membrane. Figs. 3 and 4 shows indication of correlation between the flux of both pure water and salts model solutions with contact angle values (hydrophobicity/hydrophilicity) can be obtained for fouled and virgin NF membrane. It can be drawn that the

Fig. 3. The correlation between the flux and contact angle values for virgin membrane cleaned by various cleaning agents.

A. Al-Amoudi et al. / Applied Surface Science 254 (2008) 3983–3992

3991

Fig. 4. The correlation between the flux and contact angle values for fouled membrane cleaned by various cleaning agents.

relation between the flux and contact angle values is inversely proportional. As the contact angle values decreased (hydrophilic) the flux is increased. Therefore, the relation between the flux and contact angle could be quantitative; thereby the cleaning efficiency can be evaluated by contact angel with flux at different chemical agent’s concentration and operation conditions. Similar trend for HL and DL NF membrane can be recognized. Generally, when dealing with loose NF membranes, size exclusion is more important in understanding the rejection of uncharged solute molecules, and electrostatic repulsion is more important for the rejection of ionic species. The higher rejection was obtained from monovalent ions (NaCl) for cleaned virgin and fouled DK TFM, with high pH chemical cleaning. This finding reveals that the membrane surface state was more hydrophilic after high pH cleaning. This could be caused by either remaining of the high cleaning agents of SDS and mixed agents (adsorption phenomena) in/on the membrane or remaining of the hydrophilic fraction of NOM acting, and thus, enhancing the partitioning and passage of water molecules compared to the salt. The salt passage also varied from one cleaning agent to other. For example, the salt rejection (divalent and monovalent) was approximately 50% lower than the control with mix and SDS cleaning agents. This indicates that the chemical reactions between the chemical agents and the foulant take place either by changing the morphology of the foulant or by altering the surface chemistry of fouling layer in order to remove the foulants from the membrane surfaces leading to change in the network cross-linking of the polymer, and thus, increasing salt passage (Figs. 1 and 2) [31].

to modify membrane surface properties (hydrophobicity/ hydrophilicity) for both virgin and fouled membranes; however, the present study suggests that the cleaning does have a major effect on the performance of NF membranes and their surface properties. The contact angle measurements indicate that Mix and SDS cleaning agents had a very adverse effective on the virgin and fouled NF membranes. The membrane surface properties were modified to the extent where the flux was increased over 100% and membrane hydrophilicity was increased significantly (low contact angle measurement) indicating irreversible chemical reaction and adsorption at the membrane surface. However, NaOH cleaning results revealed better performance with considerable cleaning efficiency whereas HCl cleaning resulted in poor cleaning efficiency. The DSPM method did not give any valuable information about the pore size of the untreated and treated membranes cleaned by various cleaning agents. The pore size differences between the treated and untreated NF membranes were too small to be considered significant for any investigation. The rejection method too did not provide significant information in order to assist in further understanding of membrane interactions with foulants and cleaning agents. In general, the data for contact angle measurements was supported by the corresponding water flux values and the correlation between contact angle value and flux can be obtained and combining data in this manner allows a greater understanding of membranes, their interactions with foulants and chemicals and also the relationship between the characterization techniques.

5. Conclusion

Acknowledgements

The overall result from this work illustrates that the contact angle measurement provides valuable information and details about the membrane surface state after cleaning processes, and it is also a reliable technique to assess the surface state of cleaned NF membranes and usually, chemical cleaning is found

We thank SWCC for donation of the DK fouled membrane. We also thank GE Osmonics for the donation of the virgin NF membrane (DK, DL and HL). We thank Dr. Chris Phillips from the paint and coating Dept. for his support in contact angle experiments.

3992

A. Al-Amoudi et al. / Applied Surface Science 254 (2008) 3983–3992

References [1] A. Al-Amoudi, et al., Cleaning results of new and fouled nanofiltration membrane characterized by zeta potential and permeability, Sep. Purif. Technol. 54 (2) (2007) 234–240. [2] V. Mavrov, et al., Desalination of surface water to industrial water with lower impact on the environment, Part 2: improved feed water pretreatment, Desalination 110 (1/2) (1997) 65–73. [3] H. Lee, et al., Cleaning strategies for flux recovery of an Ultrafiltration membrane fouled by natural organic matter, Pergamon 35 (2001) 3301– 3308. [4] G. Morel, et al., Surfactant modified ultrafiltration of nitrate ions removal, J. Membr. Sci. 134 (1997) 47–57. [5] J. Vrouwenvelder, et al., Tools for fouling diagnosis of NF and RO membranes and assessment of the fouling potential of feed water, Desalination 157 (2003) 361–365. [6] B. Van der Bruggen, L. Braeken, C. Vandecasteele, Flux decline in nanofiltration due to adsorption of organic compounds, Sep. Purif. Technol. 29 (1) (2002) 23–31. [7] S. Lee, J. Kim, C.-H. Lee, Analysis of CaSO4 scale formation mechanism in various nanofiltration modules, J. Membr. Sci. 163 (1) (1999) 63–74. [8] H. Zhu, M. Nystrom, Cleaning results characterized by flux, streaming potential and FTIR measurements, Colloids Surf. A: Physicochem. Eng. Aspects 138 (2/3) (1998) 309–321. [9] W.R. Bowen, A.W. Mohammad, N. Hilal, Characterisation of nanofiltration membranes for predictive purposes use of salts, uncharged solutes and atomic force microscopy, J. Membr. Sci. 126 (1) (1997) 91–105. [10] W.R. Bowen, H. Mukhtar, Characterisation and prediction of separation performance of nanofiltration membranes, J. Membr. Sci. 112 (2) (1996) 263–274. [11] C. Labbez, et al., Analysis of the salt retention of a titania membrane using the ‘‘DSPM’’ model: effect of pH, salt concentration and nature, J. Membr. Sci. 208 (1/2) (2002) 315–329. [12] W. Richard Bowen, T.A. Doneva, H.B. Yin, Polysulfone—sulfonated poly(ether ether) ketone blend membranes: systematic synthesis and characterisation, J. Membr. Sci. 181 (2) (2001) 253–263. [13] W.R. Bowen, J. Welfoot, Modelling the performance membrane nanofiltration—critical assessment and model development, Chem. Eng. Sci. 157 (2002) 1121. [14] W.R. Bowen, J. Welfoot, P. Williams, A linearised transport model for nanofiltration: development and assessment, AIChE 143 (2002) 760. [15] K.-J. Kim, A.G. Fane, Performance evaluation of surface hydrophilized novel ultrafiltration membranes using aqueous proteins, J. Membr. Sci. 99 (2) (1995) 149–162.

[16] J. Warczok, et al., Concentration of apple and pear juices by nanofiltration at low pressures, J. Food Eng. 63 (2004) 63–70. [17] Manufacturers Technical Manual of Osmonic manf. [18] A.S. Al-Amoudi, A.M. Farooque, Performance, restoration and autopsy of NF membranes used in seawater pretreatment, Desalination 178 (2005) 261–271. [19] R. Du, A. Chakma, X. Feng, Interfacially formed poly (N,N-dimethylaminoethyl methacrylate)/polysulfone composite membranes for CO2/N2 separation, J. Membr. Sci. 290 (2007) 19–28. [20] A.R. Roudman, F.A. DiGiano, Surface energy of experimental and commercial nanofiltration membranes: effects of wetting and natural organic matter fouling, J. Membr. Sci. 175 (1) (2000) 61–73. [21] TAPPI, Surface wettability and absorbance of sheeted materials using an automated contact angle test. T504 om-89, In TAPPI Test Methods 1995 (1995). [22] W.R. Bowen, T.A. Doneva, H.-B. Yin, Separation of humic acid from a model surface water with PSU/SPEEK blend UF/NF membranes, J. Membr. Sci. 206 (1/2) (2002) 417–429. [23] L. Palacio, et al., Contact angles and external protein adsorption onto UF membranes, J. Membr. Sci. 152 (2) (1999) 189–201. [24] J.T.F. Keurentjes, et al., Hydrophobicity measurements of microfiltration and ultrafiltration membranes, J. Membr. Sci. 47 (3) (1989) 333– 344. [25] J. Cho, et al., Characterization of clean and natural organic matter (NOM) fouled NF and UF membranes and foulants characterization, Desalination 118 (1998) 101–108. [26] N.D. Lawrence, et al., The use of streaming potential measurements to study the fouling and cleaning of ultrafiltration membranes, Sep. Purif. Technol. 48 (2) (2006) 106–112. [27] W.S. Ang, S. Lee, M. Elimelech, Chemical and physical aspects of cleaning of organic-fouled reverse osmosis membranes, J. Membr. Sci. 272 (1/2) (2006) 198–210. [28] C.E. Hoeft, R.L. Zollars, Adsorption of single anionic surfactants on hydrophobic surfaces, J. Colloid Interface Sci. 177 (1) (1996) 171– 178. [29] A.E. Childress, M. Elimelech, Relating nanofiltration membrane performance to membrane charge (electrokinetic) characteristics, Environ. Sci. Technol. 34 (17) (2000) 3710–3716. [30] M.J. Hurndall, E.P. Jacobs, R.D. Sanderson, The performance of novel reverse osmosis membranes made from poly-2-vinylimidazoline. I, Desalination 86 (2) (1992) 135–154. [31] A. Wies, M.R. Bird, M. Nystrom, The chemical cleaning of polymeric UF membranes fouled with spent sulphite liquor over multiple operational cycles, J. Membr. Sci. 216 (2003) 67–79.