w a t e r r e s e a r c h 6 3 ( 2 0 1 4 ) 2 5 2 e2 6 1
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/watres
Chelating polymer modified P84 nanofiltration (NF) hollow fiber membranes for high efficient heavy metal removal Jie Gao, Shi-Peng Sun, Wen-Ping Zhu, Tai-Shung Chung* Department of Chemical & Biomolecular Engineering, National University of Singapore, 10 Kent Ridge, Crescent, Singapore 119260, Singapore
article info
abstract
Article history:
High performance nanofiltration (NF) membranes for heavy metal removal have been
Received 9 April 2014
molecularly designed by adsorption of chelating polymers containing negatively charged
Received in revised form
functional groups such as poly (acrylic acid-co-maleic acid) (PAM), poly (acrylic acid) (PAA)
30 May 2014
and poly (dimethylamine-co-epichlorohydrin-co-ethylenediamine) (PDMED) on the posi-
Accepted 4 June 2014
tively charged polyethyleneimine (PEI) cross-linked P84 hollow fiber substrates. Not only do
Available online 21 June 2014
these chelating polymers change the membrane surface charge and pore size, but also provide an extra mean to remove heavy metal ions through adsorption in addition to
Keywords:
traditional steric effect and Donnan exclusion. The adsorbed membranes have comparable
P84 polyimide
water permeability and superior rejections to heavy metals, for instance, Pb(NO3)2, CuSO4,
Hyperbranched polyethyleneimine
NiCl2, CdCl2, ZnCl2, Na2Cr2O7 and Na2HAsO4, with rejections higher than 98%. The mem-
Cheating polymer
branes also display excellent rejections to mixed ions with rejections more than 99%. The
Nanofiltration
newly developed membranes show reasonably stability during 60-h tests as well as mul-
Hollow fiber membrane
tiple washes.
Heavy metal ions
1.
Introduction
Water scarcity is a severe worldwide problem to be solved urgently (Vorosmarty et al., 2010). Therefore, many efforts have been made to safely discharge and reuse the reclaimed wastewater. One of the major contaminates in industrial wastewater is heavy metals. Heavy metals have high toxicity. Even extremely low concentrations of heavy metals in human body can disrupt the body's normal physiological activities. They can also accumulate in certain organs in human body, resulting in pathological changes ranging from odd diseases
* Corresponding author. Tel.: þ65 6516 6645; fax: þ65 6779 1936. E-mail address: [email protected] (T.-S. Chung). http://dx.doi.org/10.1016/j.watres.2014.06.006 0043-1354/© 2014 Elsevier Ltd. All rights reserved.
© 2014 Elsevier Ltd. All rights reserved.
(such as water Minamata disease, bone disease, etc.) to even death (Athar and Vohora, 2001). Thus, heavy metal pollution has drawn increasing attentions throughout the world. Many countries have set-up more and more stringent standards to control heavy metal concentrations in discharged water (Xu and Zhao, 2005), which opens huge opportunities for water treatment technologies. Compared to conventional technologies, membrane separation has many advantages such as cost effective, energy saving, no phase change, environmental friendly and high removal efficiency (Deon et al., 2013; Escobar and Van der Bruggen, 2011). Since the majority of heavy metal ions are
w a t e r r e s e a r c h 6 3 ( 2 0 1 4 ) 2 5 2 e2 6 1
multivalent, one of the promising membrane filtration methods to treat heavy metals is nanofiltration (NF), owing to its unique rejection mechanisms - steric effect and Donnan exclusion (Chiang et al., 2009; Zhou et al., 2009). Compared to reverse osmosis (RO), NF requires a lower pressure while giving a higher flux without much compromise in rejection (Abu Qdais and Moussa, 2004). Recently, there is a growing interest on the positively charged polyethyleneimine (PEI) cross-linked polyimide NF membrane due to its good thermal, mechanical and chemical stabilities as well as excellent rejections to multivalent cations (Cheng et al., 2012; Economy et al., 2009). However, the rejection of the highly positively charged membrane to multivalent anions is not so promising, which limits its application in the field of heavy metal removal. Thus, the purposes of the study are to (1) modify the surface charge of the NF membrane to neutral or slightly charged so that the membrane is capable of removing a wide range of heavy metals and (2) introduce the adsorption mechanism other than size exclusion and charge repulsion to improve the rejection efficiency of the membrane. To our best knowledge, the concept of adsorbing polyelectrolyte chelating polymers onto NF membranes for heavy metal removal has not been proposed before. These polymers, such as poly (acrylic acid) (PAA), poly (acrylic acid-co-maleic acid) (PAM) and poly (dimethylamine-co-epichlorohydrin-coethylenediamine) (PDMED) (structures shown in Fig. S1 of the supporting material), are chosen because (1) the negatively charged functional groups on the polymers can enhance the adsorption of the polymers onto the oppositely charged membrane (Butt et al., 2003); (2) the induced negatively charged functional groups from these polymers can change the membrane surface charge, making it more negatively charged at high pH (Childress and Elimelech, 1996); (3) the adsorption of polyelectrolyte, such as PAA, has been reported to have the antifouling function in the previous literature (Ba et al., 2010); (4) chelating polymers are able to absorb heavy metal ions, which provides additional means to remove heavy metal ions (Van der Bruggen et al., 2004); (5) the polyelectrolyte coating may decrease the membrane pore size and enhance the rejection (Ba et al., 2010); and (6) the adsorbed coating is simple and can be customized according to different applications. To effectively coat the polyelectrolytes and provide mechanically strong support under high pressure operations, a substrate with a macrovoid-free structure and finely tuned pore size was formed by P84 (copolyimide of 3,3’,4,4’-benzophenone tetracarboxylic dianhydride with 80% toluenediisocynate and 20% methyl phenylene diisocyanate (BTDA-TDI/ MDI)) because its excellent thermal and chemical stability (Toh et al., 2007). A hyperbranched PEI with a molecular weight of 60 K gmol1 was used for cross-linking with P84 before adsorption to induce more amine groups on the membrane surface and enhance the membrane performance after cross-linking (Albrecht et al., 2003; Sun et al., 2011). After all modifications, heavy metal salts, including Pb(NO3)2, CuSO4, NiCl2, CdCl2, ZnCl2, Na2Cr2O7 and Na2HAsO4 were used to test the performance of the chelating polymer adsorbed membranes. The newly formed membranes may open opportunities of new NF membranes in wastewater treatment.
2.
Experimental
2.1.
Materials
253
P84 powders (molecular weight (MW) of 153 KDa, HP polymer GmbH, Australia) were used to form the hollow fiber substrates. The solvent and non-solvent to prepare the dope solution were n-methyl-2-pyrrolidine (NMP, 99.5%, Merck, Germany) and methanol (99.8%, Fisher Chemical, UK), respectively. The cross-linking agent, PEI (MW of 60 K gmol1, 50%), was purchased from Acro (USA). Isopropanol (99.98%, Fisher Chemical, UK) and deionized water were used to dissolve PEI. Chelating polymers, such as PAA 2K (MW of 2 K gmol1, 50%), PAA 100K (MW of 100K gmol1, 35%), PAA 250K (MW of 250 K gmol1, 35%), PAM (MW of 3 K gmol1, 50%) and PDMED (MW of 75 K gmol1, 50%), were acquired from Aldrich. To test the heavy metal ion rejections of the PEI crosslinked or polyelectrolyte adsorbed membranes, Pb(NO3)2 (99%), CuSO4∙5H2O (>99%), NiCl2 (98%), CdCl2, ZnCl2 (98.5%), Na2Cr2O7∙2H2O (99%) and Na2HAsO4∙7H2O (>98%) were purchased from Acros. They were dissolved in DI water to form heavy metal solutions with 300 ppm or 1000 ppm concentrations. HCl (37%) and NaOH (>99%, Emsure®, Merck, Germany) were diluted and used to adjust the pH of salt solutions as well as to test the membrane stability.
2.2. Fabrication of the outer selective hollow fiber substrate The spinning procedure is similar to our previous work (Gao et al., 2014). In short, the dope solution was prepared with vacuum dried P84 powders and was allowed to degas for 12 h after it was well-mixed. The bore fluid and the dope solution were purged by two different pumps at certain flow rates and met at the outlet of the spinneret. The mixture then passed through a certain air gap before entering the coagulant bath for phase separation and getting stretched by a take-up unit. The fibers were subsequently cut and transferred to a water basin to complete the phase inversion. After that, the fibers were immersed in a 50 wt% glycerol aqueous solution for 2 days and then air-dried. The formulations of the dope solution and the bore fluid, together with the spinning conditions, are summarized in Table 1. Hollow fiber modules consisting of 20 fibers with an effective length of 14 cm were prepared for all the experiments.
2.3.
Modification of the as-spun membrane
A cross-flow set-up as described by Sun et al. (Sun et al., 2011) was used to achieve the cross-linking modification between PEI and the outer surface of P84 hollow fiber substrate. The cross-linking solution was formed by dissolving 1 wt% of PEI in a 50 wt% isopropanol aqueous solution. The solution was circulated in the shell side of the membrane substrate at 70 C and 400 ml/min for 1 h. After that, the module was rinsed with DI water at 1 bar to remove the unreacted residuals. The rejection of the cross-linked membrane to MgCl2 was then tested to ensure the cross-linking was efficient and consistent. The adsorptions of various polyelectrolytes were then
254
w a t e r r e s e a r c h 6 3 ( 2 0 1 4 ) 2 5 2 e2 6 1
Table 1 e The spinning conditions of the P84 hollow fiber membrane substrate. Spinning condition P84 dope solution (wt.%) Dope flow rate (ml/min) Bore fluid composition (wt.%) Bore fluid flow rate (ml/min) Air gap (cm) Take-up speed (m/min) External coagulant () Dope temperature ( C) Bore fluid temperature ( C) External coagulant temperature ( C) Room humidity (%) Dimension of spinneret (mm) Die length of spinneret (mm)
Condition P84/methanol/NMP (28:10:62) 3 NMP/DI water (90:10) 1.5 4.5 39.9 Tap water 26 ± 1 26 ± 1 26 ± 1 65e70 i.d./o.d. (1.05/1.6) 6.5
performed on the cross-linked membranes. During the adsorption process, a 50 mg/L aqueous polyelectrolyte solution was circulated through the shell side of the membrane at 1 bar with a bench top NF set-up overnight. The polyelectrolyte adsorbed membrane was then rinsed from the shell to the lumen side at 1 bar to preliminarily remove the unattached chelating polymer. They were transferred to the pilot-scale NF set-up for further cleaning at 10 bar until the flux was stabilized and the conductivities of permeates were below 1 mS/cm. The cleaned membrane was then ready for testing. The procedure of the chemical modification is summarized in Fig. S2 of the supporting material.
2.4.
Characterization of the membrane
2.4.1.
Morphology of the membrane
The morphologies of the plain P84 hollow fiber membranes, the PEI cross-linked membrane and the PAA adsorbed membrane were observed by a field emission scanning electron microscope (FESEM, JEOL, JSM-6700F). To prepare the cross-section samples, the fibers were freeze dried and then fractured in liquid nitrogen.
2.4.2.
Surface charge of the membrane
The surface charge of the membranes was analyzed by a SurPASS electrokinetic analyzer (Anton Paar GmbH, Austria) through streaming potential measurements. To mimic a hollow fiber membrane, flat-sheet membranes with the same dope formulation and post-treatment were prepared. All the membranes were well washed before tests. The testing procedures are as follows. A 0.01 M NaCl solution was used to test the membrane zeta-potential at neutral pH. Then a 0.1 M HCl was used to auto-titrate the solution pH from its neutral pH to pH 2.6. After that, the pH of the 0.01 M NaCl solution was increased gradually from pH 2.6 to pH 11 by auto-titrating the NaCl solution with 0.1 M NaOH. Zeta-potential of the membrane was recorded accordingly. The membrane surface charge as a function of pH and the isoelectric point were then determined.
2.4.3.
Adsorption of heavy metals
The experiments were also conducted using flat-sheet membranes. Three kinds of membranes were prepared; namely,
plain P84 membrane, PEI cross-linked membrane and polyelectrolyte adsorbed membrane. They were freeze dried prior to tests. During the tests, 100 mg of each membrane was cut and immersed into 200 ppm heavy metal solutions containing different single salts individually. The mixtures were then rotated on a roller continuously. The concentrations of the solutions were tracked even after reaching equilibrium, with a total duration of around 60 h. The heavy metal concentrations in the solutions were determined by an inductively coupled plasma optical emission spectroscopy (ICP-OES, Optima 7300DV, Perkin Elmer, USA). The adsorption amounts q (mg/g) of heavy metal ions on each membrane were then calculated through the following formula (Montazer-Rahmati et al., 2011): q¼
c0 ceq V m
(1)
where c0 and ceq represent the initial and equilibrium concentrations of heavy metal ions in the solution (ppm), correspondingly, V is the solution volume (L), and m is the mass of the dry membrane (g).
2.4.4. Pure water permeability and salt rejection of the modified membrane To test the pure water permeability (PWP), DI water was circulated in the shell side of the hollow fiber modules at 10 bar under a constant flow rate of 0.3 L/min and the permeate water was collected within a certain duration. PWP (Lm2 bar1 h1) was calculated using the following formula: PWP ¼
Q ADP
(2)
where Q is the water permeability at the permeate side (L/h), A is the effective filtration area (m2) and DP is the transmembrane pressure (bar). In addition to test single salt heavy metal solutions that comprising 1000 ppm different heavy metal ions, two kinds of mixed ion solutions were prepared according to the compatibilities of the ions. The first kind consists of 1200 ppm Cu, Ni, Zn and Cr ions (each 300 ppm). The second type comprises 300 ppm Pb and 300 ppm Cd ions. Prior to testing, the fibers were rinsed to ensure that the conductivity of the permeate solution was below 1 mS/cm and the PWP was stabilized. During the tests, feed and permeate solutions were collected after the modules were stabilized at least 1 h with salt solutions circulating at 10 bar in the shell side. Each test was measured three times by consecutively taking three samples with a time interval of 1 h to ensure the modules were stabilized for measurements. The ion rejections and their standard deviations were calculated based on the average and standard deviations of the three measurements to ensure that the deviation was within 2%. The concentrations of both the feed and the permeate solutions were determined by ICP-OES. The salt rejection was calculated by the following formula (3): R¼
cp 100% 1 cf
(3)
where cp and cf are concentrations of the permeate and the feed solutions, respectively. To test the stability of the adsorbed membrane after acid or base wash, the chelating polymer adsorbed membrane was washed with a diluted
w a t e r r e s e a r c h 6 3 ( 2 0 1 4 ) 2 5 2 e2 6 1
HNO3 solution (pH ¼ 2.2) or NaOH solution (pH ¼ 10) at 1 bar from the shell side for two days. The membrane was cleaned before performance tests.
3.
Results and discussions
3.1.
Characteristics of the membranes
In this part, PAA 100K is used as the representative of chelating polymers for all characterizations because the density of its negatively charged functional group is between that of PAM and PDMED. For easy discussion, the PEI crosslinked membrane is denoted as Plain-PEI, while the PAA 100K adsorbed membrane is signified as Plain-PEI&PAA 100K.
255
Hollow fiber membranes comprising a macrovoid-free cross-sectional structure are preferred in NF because they may have (1) high mechanical strength to sustain high operation pressures, (2) less defective surface for high rejection and (3) more uniform pore size distribution (Peng et al., 2008; Sun et al., 2010). The ways to fabricate macrovoid-free hollow fiber membranes have been summarized in literature (Peng et al., 2012). A macrovoid-free hollow fiber membrane was therefore achieved in this study by manipulating the spinning conditions as well as the dope solution. As listed in Table 1, a dope solution consisting of a high polymer concentration of 28 wt % and a non-solvent of methanol was prepared. The non-solvent additive was chosen because the resultant dope solution can be spun at a high take-up speed to achieve a macrovoid-free structure (Wang et al., 1999). The
Fig. 1 e FESEM images of the plain P84 membrane (a) cross-section (£200), (b) cross-section (£10,000), (c) outer surface (£50,000) and (d) inner surface (£10,000); and membrane outer surface with post-treatment (e) PEI cross-linked (Plain-PEI, £50,000) and (f) PEI cross-linked and PAA adsorbed (Plain-PEI&PAA 100K, £50,000).
w a t e r r e s e a r c h 6 3 ( 2 0 1 4 ) 2 5 2 e2 6 1
FESEM images of the plain P84 membranes are presented in Fig. 1. At a lower magnification (Fig. 1a), the membrane crosssection looks quite dense. However, a highly porous structure can be observed at a higher magnification (Fig. 1b). No obvious pore structures can be seen on the outer surface (Fig. 1c), indicating the membrane selective layer is apparently defectfree. Besides, the inner layer of the substrate may pose little resistance to water transport as big holes can be found on the inner surface (Fig. 1d). After PEI cross-linking, some small nodules are observed on the membrane surface (Fig. 1e). The membrane becomes a little rougher after the adsorption of PAA 100K (Fig. 1f). Fig. 2 shows the ATR-FTIR spectra of the pristine P84, PlainPEI and Plain-PEI&PAA 100K membranes. Instead of representative peaks of imide at 1770 cm-1 (C]O stretching), 1716 cm1 (C]O stretching) and 1357 cm1 (CeN stretching), new peaks at 1651 cm1 (C]O stretching) and 1543 cm1 (CeN stretching) appear after PEI cross-linking, implying the formation of the amide group (Ba et al., 2009; Gao et al., 2014). After the PAA 100K adsorption, an additional peak was found at 1735e1708 cm1 except all the peaks in Plain-PEI, indicating the successfully adsorption of PAA (Ba et al., 2010). The surface charge characteristics as a function of pH of the plain P84, PEI cross-linked and PAA 100K adsorbed membranes were shown in Fig. 3. The plain P84 membrane has an isoelectric point of 3.2, which is similar to our previous characterizations of polyimide membranes (Gao et al., 2014). After the PEI cross-linking modification, due to the protonation of the amine groups of PEI and the amide groups formed between PEI and P84, the PEI cross-linked membrane is highly positively charged below its isoelectric point of pH 9.4. After the PAA 100K adsorption, the membrane becomes more negatively charged compared to the Plain-PEI because the isoelectric point shifts to pH 6.2 possibly due to the protonation of amine or amide groups (below pH 9.4) and deprotonation of carboxyl groups. The nearly neutral surface charge between pH 6 and 8 may help decrease membrane fouling (Ba and Economy, 2010; Peng et al., 2004).
Fig. 2 e Comparison of ATR-FTRI spectra among the plain P84 membrane, PEI cross-linked membrane (Plain-PEI) and PAA adsorbed membrane (Plain-PEI&PAA 100K).
30
Plain-PEI Plain-PEI&PAA Plain P84 membrane
20
Zeta-potential, (mV)
256
10
0 0
2
4
6
8
10
12
-10 -20 -30 -40
pH, (-)
Fig. 3 e Comparison of zeta-potential trends as a function of pH and the isoelectric points among the plain P84 membrane (isoelectric point: pH 3.2), PEI cross-linked membrane (Plain-PEI, isoelectric point: pH 9.4) and PAA adsorbed membrane (Plain-PEI&PAA 100K).
3.2. Influence of chelating polymer types and molecular weights on NF performance Three types of heavy metal ions, Pb2þ, Cd2þ complex ions (because CdCl2 forms complex metal ions of Cd2þ, CdClþ, 2 CdCl-3 and CdCl2 4 in solutions (Zhu et al., 2014)) and Cr2O7 , were chosen for comparison purposes in this study to represent positive charged ion, complex ion and negatively charged ion, respectively. The influences of different chelating polymers such as PAA 100K, PAM and PDMED on rejection and water permeability of chelating polymer adsorbed membranes are shown in Fig. 4a. The plain-PEI membrane works as the control. The Plain-PEI membrane has a good rejection to Pb2þ (99.15%) and a slightly lower rejection to Cd2þ complex ions (98%). However, (less than 80%) due to the it has a poor rejection to Cr2O2 7 positively charged membrane surface. After adsorption of chelating polymers, all membranes show much higher re2þ and Cd2þ jections to Cr2O2 7 . Besides, their rejections to Pb complex ions also improve. Clearly, the adsorption of chelating polymers with negatively charged functional groups on membranes helps improve the ion rejections, especially the negatively charged ions. When comparing the influences of different chelating polymers to metal ion rejections, the following trend can be found: Plain-PEI&PAM > Plain-PEI&PAA 100K > Plain-PEI&PDMED. It is easy to interpret this phenomenon as PAM has the highest density of negatively charged functional groups among the three polymers. More PAM may be adsorbed on the membrane, making the membrane surface more negatively charged. For PDMED, only the hydroxyl group as the negatively charged functional group is found on the molecular structure. Compared to PAA 100K and PAM, PDMED has more positively charged groups. Though the Plain-PEI&PDMED membrane has a little lower rejection than those of Plain-PEI&PAA 100K and Plain-PEI&PAM membranes, it has performance better than the Plain-PEI membrane. Water permeability for chelating polymer adsorbed
257
w a t e r r e s e a r c h 6 3 ( 2 0 1 4 ) 2 5 2 e2 6 1
Fig. 4 e Influence of (a) different chelating polymers (b) PAA molecular weight on ion rejection (top) and water permeability (bottom) of the membranes.
3.3.
Heavy metal adsorption tests of the membranes
The adsorption of heavy metal ions on the plain P84, Plain-PEI and Plain-PEI&PAA 100K membranes were investigated with the data shown in Fig. 5. The saturation time of the membranes to heavy metals was around 16 h. However, to parallel the long term stability tests as shown in the latter section, the data was collected after 60 h. Among the three membranes, PlainPEI&PAA 100K has the highest adsorption of Pb2þ and Cd2þ complex ion, while Plain-PEI has the highest adsorption of Cr2O2 7 . The plain P84 membrane contributes the least toward heavy metal adsorption. The results confirm that the chelating
polymers on Plain-PEI (PEI is also a chelating polymer (Navarro et al., 1999)) and Plain-PEI&PAA 100K membranes promote the adsorption of heavy metals on the membrane surface. Thus, besides size exclusion and charge repulsion, the newly modified NF membrane can also remove heavy metal ions with the aid of adsorption.
3.4. Rejection tests of the Plain-PEI&PAA 100K membrane Fig. 6 shows ion rejections and water permeability of the PAA adsorbed membrane (Plain-PEI&PAA 100K) to different single
25
adsorption, q(mg/g)
membranes follows the trend of: Plain-PEI&PDMED > PlainPEI&PAA 100K > Plain-PEI&PAM, which is the reverse order of rejections. However, it is lower than that of the Plain-PEI membrane, indicating the successful adsorption of chelating polymer (Jermann et al., 2007). To investigate the influence of molecular weight of the chelating polymer on charge properties and separation performance of the PEI cross-linked membrane, PAA was chosen for experiments because: (1) the rejections and permeability of the Plain-PEI&PAA membrane are in between those of PlainPEI&PDMED and Plain-PEI&PAM membranes, (2) the PAA adsorbed membrane has certain antifouling properties according to Ba et al. (Ba et al., 2010), and (3) several molecular weights of PAA such as PAA 2K, PAA 100K and PAA 250K are available in the market. Fig. 4b shows the influence of PAA molecular weight on the membrane performance. No significant effects of PAA molecular weight on ion rejection and water permeability can be found. Thus, PAA 100K is chosen for the following experiments.
20
Cr
Cd
Pb
15 10
5 0 Plain P84 membrane
Plain-PEI
Plain-PEI&PAA
Membrane type Fig. 5 e Adsorption of heavy metal ions on the plain P84, PEI cross-linked (Plain-PEI) and PAA adsorbed (PlainPEI&PAA 100K) membranes.
258
w a t e r r e s e a r c h 6 3 ( 2 0 1 4 ) 2 5 2 e2 6 1
Fig. 6 e Ion rejections (left) and water permeability (right) of the PAA adsorbed membrane (Plain-PEI&PAA 100K) to different single heavy metal ions.
heavy metal ions. As aforementioned, size exclusion (steric effect) and charge repulsion (Donnan exclusion) are two main mechanisms that contribute to the rejections of NF membranes. In this study, the steric effect is primarily related to the hydrated radii of heavy metal ions, while the Donnan exclusion effect mostly depends on the charge of heavy metal ions as well as the surface charge of the membrane. Hydrated radii (Nightingale, 1959; Volkov et al., 1997) and diffusivities (Lv et al., 2008; Mortimer, 2008; Newman, 1991; Sato et al., 1996; Wang et al., 2007) of all the heavy metal ions tested are summarized in Table S1 of the supporting material. These heavy metal ions can be classified into three categories, which are simple cations (Cu2þ, Ni2þ and Pb2þ), oxoanions (Cr2O2 7 2þ and HAsO2 complex ions and 4 ), and complex metal ions (Zn Cd2þ complex ions). All the simple cations in this study have the same charge, i.e. þ2 charges. Rejections of these cations are all significantly high and follow the weakly trend of Pb2þ (99.48%) < Ni2þ (99.73%) < Cu2þ (99.95%), which is the order of hydrated radii of these three ions. For the oxoanions, the rejection of HAsO42 (99.52%) is much higher than that of Cr2O2 7 (93.55%). It can be explained in two ways. Firstly, the pH of Na2HAsO4 solution (pH 7.7) is much higher than that of Na2Cr2O7 solution (pH 4.9). The membrane is more negatively charged in the Na2HAsO4 solution. Secondly, the hydrated radius of an ion is reverse related to diffusivity and crystal radius (Tansel et al., 2006; Zaikov et al., 1988). To compare the 2 2 hydrated radii of HAsO2 4 and Cr2O7 , CrO4 is introduced as a 2 reference. The hydrated radius of Cr2O7 is lower than that of 2 CrO2 4 because Cr2O7 has a larger crystal radius (Zhu et al., 2014). However, the hydrated radius of CrO2 4 is smaller than 2 that of HAsO2 4 because HAsO4 has a smaller diffusivity as shown in Table S1 of the supporting material. Thus, HAsO2 4 has a larger hydrated radius than Cr2O2 7 . Based on the above two reasons, the rejection to HAsO2 4 is higher than that of 2 . However, the rejection of Cr can be improved Cr2O2 7 2O7 significantly by increasing the solution pH as the membrane would become more negatively charged. The rejection of Cr2O2 7 increases from 93.55% to 98.93% when the pH of the solution is adjusted from pH 4.9 to pH 10. For complex metal ions, the rejection of Zn2þ complex ions is higher than that of
Cd2þ complex ions probably due to the larger hydrated radius of Zn2þ. The water permeability of almost all the salt solutions is above 1 Lm2 bar1 h1. The performance of the membrane is comparable to literature for the removal of heavy metal ions by hollow fiber membranes as shown in Table 2 (Gao et al., 2014; Lv et al., 2008; Wang and Chung, 2006; Wang et al., 2007; Wei et al., 2013; Zhu et al., 2014). The rejections of the Plain-PEI&PAA 100K to mixed heavy metal ions are also tested, as shown in Fig. 7. As described in Section 2.4.4, two kinds of mixed ions solutions were prepared. For both solutions, all the ion rejections are above 99.8% except for Cr2O2 7 , which is 97.6%. Compared to Fig. 6, the rejections of most ions in mixed ion tests are slightly higher than their counterparts in single ion tests because a much lower ion concentration was used in the mixed ion tests (each ion is 300 ppm). The water permeability of the first mixed solution made of four ions (Cu, Ni, Zn and Cr ions) decreases a bit, while the second mixed solution made of two ions (Pb and Cd ions) increases a bit compared to the single ion solutions. It may be related to the total salt concentration in the solution because a high salt concentration normally contribute to a low flux due to the effects of concentration polarization (Kilduff et al., 2004).
3.5.
The stability of the Plain-PEI&PAA 100K membrane
Fig. 8 (left) shows the long term stability of the Plain-PEI&PAA 100K membrane for a continuous rejection test of 60 h. CdCl2 was chosen for the experiment because it forms both cation and anions in water and has high toxicity to human beings (Zhu et al., 2014). The rejection remains stable and high throughout the entire experiment. The water permeability also exhibits good stability but it slightly decreases as a function of time probably due to scaling which is normal in NF tests. However, the water permeability can be recovered by normal cleaning with DI water. Various heavy metal ions were also used to test the membrane stability after different pure wash circles. Fig. 8 (right) shows the results where the separation performance of Cd2þ complex ions performs as a reference at certain wash
259
w a t e r r e s e a r c h 6 3 ( 2 0 1 4 ) 2 5 2 e2 6 1
Table 2 e Selected references on removal of heavy metal ions by hollow fiber NF membranes. Ion As (V) (NaH2AsO4) Cu2þ(CuSO4) Cu2þ(CuCl2) Cr (VI) (Na2Cr2O7) Cr (VI) (K2Cr2O7) Ni2þ Cr6þ Cu2þ Pb2þ(Pb(NO3)2) Cd2þ (CdCl2) Pb2þ(Pb(NO3)2) Cr (VI) (Na2Cr2O7) Cd2þ, Zn2þ Pb2þ,Ni2þ,Cu2þ As (V),Cr (VI) Cr (VI)
Testing condition
Rejection, (%)
Feed: 75 ppm (pH ¼ 12) Feed: 64 ppm Feed: 64 ppm P ¼ 15 bar Feed: 3771 ppm pH ¼ 13.25, P ¼ 25 bar Feed: 520 ppm pH ¼ 12, P ¼ 20 bar Feed: 142.23 ppm Feed: 121.23 ppm Feed: 56.55 ppm pH: 2.31, P ¼ 4 bar Feed: 1000 ppm P ¼ 13 bar Feed: 1000 ppm Feed: 1000 ppm Feed: 1000 ppm (pH ¼ 12) P ¼ 1 bar Feed: 1000 ppm Feed: 1000 ppm Feed: 1000 ppm Feed: 1000 ppm (pH ¼ 12)
Water permeability, (Lm2 bar1 h1)
Reference
97 92.5 ~99
1.86 (pure water)
Lv et al., 2008
97.5
0.025
Wang et al., 2007
95.7
1.86 (pure water)
Wang and Chung, 2006
94.99 95.76 95.33
~8.2
Wei et al., 2013
91.05
0.98 (pure water)
Gao et al., 2014
95 93 98
0.826 (pure water)
Zhu et al., 2014
>98 >99 >98, >93 >98
>1 (most ions)
This study
circles to check the membrane sustainability. It can be found that the membrane has high rejections for all ions even after 10 wash circles. By examining the separation performance of Cd2þ complex ions after 5 circles and 10 circles, a slight increase in permeability (~0.4 Lm2 bar1 h1) was noticed with a very minor fluctuation in rejection of
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/watres
Chelating polymer modified P84 nanofiltration (NF) hollow fiber membranes for high efficient heavy metal removal Jie Gao, Shi-Peng Sun, Wen-Ping Zhu, Tai-Shung Chung* Department of Chemical & Biomolecular Engineering, National University of Singapore, 10 Kent Ridge, Crescent, Singapore 119260, Singapore
article info
abstract
Article history:
High performance nanofiltration (NF) membranes for heavy metal removal have been
Received 9 April 2014
molecularly designed by adsorption of chelating polymers containing negatively charged
Received in revised form
functional groups such as poly (acrylic acid-co-maleic acid) (PAM), poly (acrylic acid) (PAA)
30 May 2014
and poly (dimethylamine-co-epichlorohydrin-co-ethylenediamine) (PDMED) on the posi-
Accepted 4 June 2014
tively charged polyethyleneimine (PEI) cross-linked P84 hollow fiber substrates. Not only do
Available online 21 June 2014
these chelating polymers change the membrane surface charge and pore size, but also provide an extra mean to remove heavy metal ions through adsorption in addition to
Keywords:
traditional steric effect and Donnan exclusion. The adsorbed membranes have comparable
P84 polyimide
water permeability and superior rejections to heavy metals, for instance, Pb(NO3)2, CuSO4,
Hyperbranched polyethyleneimine
NiCl2, CdCl2, ZnCl2, Na2Cr2O7 and Na2HAsO4, with rejections higher than 98%. The mem-
Cheating polymer
branes also display excellent rejections to mixed ions with rejections more than 99%. The
Nanofiltration
newly developed membranes show reasonably stability during 60-h tests as well as mul-
Hollow fiber membrane
tiple washes.
Heavy metal ions
1.
Introduction
Water scarcity is a severe worldwide problem to be solved urgently (Vorosmarty et al., 2010). Therefore, many efforts have been made to safely discharge and reuse the reclaimed wastewater. One of the major contaminates in industrial wastewater is heavy metals. Heavy metals have high toxicity. Even extremely low concentrations of heavy metals in human body can disrupt the body's normal physiological activities. They can also accumulate in certain organs in human body, resulting in pathological changes ranging from odd diseases
* Corresponding author. Tel.: þ65 6516 6645; fax: þ65 6779 1936. E-mail address: [email protected] (T.-S. Chung). http://dx.doi.org/10.1016/j.watres.2014.06.006 0043-1354/© 2014 Elsevier Ltd. All rights reserved.
© 2014 Elsevier Ltd. All rights reserved.
(such as water Minamata disease, bone disease, etc.) to even death (Athar and Vohora, 2001). Thus, heavy metal pollution has drawn increasing attentions throughout the world. Many countries have set-up more and more stringent standards to control heavy metal concentrations in discharged water (Xu and Zhao, 2005), which opens huge opportunities for water treatment technologies. Compared to conventional technologies, membrane separation has many advantages such as cost effective, energy saving, no phase change, environmental friendly and high removal efficiency (Deon et al., 2013; Escobar and Van der Bruggen, 2011). Since the majority of heavy metal ions are
w a t e r r e s e a r c h 6 3 ( 2 0 1 4 ) 2 5 2 e2 6 1
multivalent, one of the promising membrane filtration methods to treat heavy metals is nanofiltration (NF), owing to its unique rejection mechanisms - steric effect and Donnan exclusion (Chiang et al., 2009; Zhou et al., 2009). Compared to reverse osmosis (RO), NF requires a lower pressure while giving a higher flux without much compromise in rejection (Abu Qdais and Moussa, 2004). Recently, there is a growing interest on the positively charged polyethyleneimine (PEI) cross-linked polyimide NF membrane due to its good thermal, mechanical and chemical stabilities as well as excellent rejections to multivalent cations (Cheng et al., 2012; Economy et al., 2009). However, the rejection of the highly positively charged membrane to multivalent anions is not so promising, which limits its application in the field of heavy metal removal. Thus, the purposes of the study are to (1) modify the surface charge of the NF membrane to neutral or slightly charged so that the membrane is capable of removing a wide range of heavy metals and (2) introduce the adsorption mechanism other than size exclusion and charge repulsion to improve the rejection efficiency of the membrane. To our best knowledge, the concept of adsorbing polyelectrolyte chelating polymers onto NF membranes for heavy metal removal has not been proposed before. These polymers, such as poly (acrylic acid) (PAA), poly (acrylic acid-co-maleic acid) (PAM) and poly (dimethylamine-co-epichlorohydrin-coethylenediamine) (PDMED) (structures shown in Fig. S1 of the supporting material), are chosen because (1) the negatively charged functional groups on the polymers can enhance the adsorption of the polymers onto the oppositely charged membrane (Butt et al., 2003); (2) the induced negatively charged functional groups from these polymers can change the membrane surface charge, making it more negatively charged at high pH (Childress and Elimelech, 1996); (3) the adsorption of polyelectrolyte, such as PAA, has been reported to have the antifouling function in the previous literature (Ba et al., 2010); (4) chelating polymers are able to absorb heavy metal ions, which provides additional means to remove heavy metal ions (Van der Bruggen et al., 2004); (5) the polyelectrolyte coating may decrease the membrane pore size and enhance the rejection (Ba et al., 2010); and (6) the adsorbed coating is simple and can be customized according to different applications. To effectively coat the polyelectrolytes and provide mechanically strong support under high pressure operations, a substrate with a macrovoid-free structure and finely tuned pore size was formed by P84 (copolyimide of 3,3’,4,4’-benzophenone tetracarboxylic dianhydride with 80% toluenediisocynate and 20% methyl phenylene diisocyanate (BTDA-TDI/ MDI)) because its excellent thermal and chemical stability (Toh et al., 2007). A hyperbranched PEI with a molecular weight of 60 K gmol1 was used for cross-linking with P84 before adsorption to induce more amine groups on the membrane surface and enhance the membrane performance after cross-linking (Albrecht et al., 2003; Sun et al., 2011). After all modifications, heavy metal salts, including Pb(NO3)2, CuSO4, NiCl2, CdCl2, ZnCl2, Na2Cr2O7 and Na2HAsO4 were used to test the performance of the chelating polymer adsorbed membranes. The newly formed membranes may open opportunities of new NF membranes in wastewater treatment.
2.
Experimental
2.1.
Materials
253
P84 powders (molecular weight (MW) of 153 KDa, HP polymer GmbH, Australia) were used to form the hollow fiber substrates. The solvent and non-solvent to prepare the dope solution were n-methyl-2-pyrrolidine (NMP, 99.5%, Merck, Germany) and methanol (99.8%, Fisher Chemical, UK), respectively. The cross-linking agent, PEI (MW of 60 K gmol1, 50%), was purchased from Acro (USA). Isopropanol (99.98%, Fisher Chemical, UK) and deionized water were used to dissolve PEI. Chelating polymers, such as PAA 2K (MW of 2 K gmol1, 50%), PAA 100K (MW of 100K gmol1, 35%), PAA 250K (MW of 250 K gmol1, 35%), PAM (MW of 3 K gmol1, 50%) and PDMED (MW of 75 K gmol1, 50%), were acquired from Aldrich. To test the heavy metal ion rejections of the PEI crosslinked or polyelectrolyte adsorbed membranes, Pb(NO3)2 (99%), CuSO4∙5H2O (>99%), NiCl2 (98%), CdCl2, ZnCl2 (98.5%), Na2Cr2O7∙2H2O (99%) and Na2HAsO4∙7H2O (>98%) were purchased from Acros. They were dissolved in DI water to form heavy metal solutions with 300 ppm or 1000 ppm concentrations. HCl (37%) and NaOH (>99%, Emsure®, Merck, Germany) were diluted and used to adjust the pH of salt solutions as well as to test the membrane stability.
2.2. Fabrication of the outer selective hollow fiber substrate The spinning procedure is similar to our previous work (Gao et al., 2014). In short, the dope solution was prepared with vacuum dried P84 powders and was allowed to degas for 12 h after it was well-mixed. The bore fluid and the dope solution were purged by two different pumps at certain flow rates and met at the outlet of the spinneret. The mixture then passed through a certain air gap before entering the coagulant bath for phase separation and getting stretched by a take-up unit. The fibers were subsequently cut and transferred to a water basin to complete the phase inversion. After that, the fibers were immersed in a 50 wt% glycerol aqueous solution for 2 days and then air-dried. The formulations of the dope solution and the bore fluid, together with the spinning conditions, are summarized in Table 1. Hollow fiber modules consisting of 20 fibers with an effective length of 14 cm were prepared for all the experiments.
2.3.
Modification of the as-spun membrane
A cross-flow set-up as described by Sun et al. (Sun et al., 2011) was used to achieve the cross-linking modification between PEI and the outer surface of P84 hollow fiber substrate. The cross-linking solution was formed by dissolving 1 wt% of PEI in a 50 wt% isopropanol aqueous solution. The solution was circulated in the shell side of the membrane substrate at 70 C and 400 ml/min for 1 h. After that, the module was rinsed with DI water at 1 bar to remove the unreacted residuals. The rejection of the cross-linked membrane to MgCl2 was then tested to ensure the cross-linking was efficient and consistent. The adsorptions of various polyelectrolytes were then
254
w a t e r r e s e a r c h 6 3 ( 2 0 1 4 ) 2 5 2 e2 6 1
Table 1 e The spinning conditions of the P84 hollow fiber membrane substrate. Spinning condition P84 dope solution (wt.%) Dope flow rate (ml/min) Bore fluid composition (wt.%) Bore fluid flow rate (ml/min) Air gap (cm) Take-up speed (m/min) External coagulant () Dope temperature ( C) Bore fluid temperature ( C) External coagulant temperature ( C) Room humidity (%) Dimension of spinneret (mm) Die length of spinneret (mm)
Condition P84/methanol/NMP (28:10:62) 3 NMP/DI water (90:10) 1.5 4.5 39.9 Tap water 26 ± 1 26 ± 1 26 ± 1 65e70 i.d./o.d. (1.05/1.6) 6.5
performed on the cross-linked membranes. During the adsorption process, a 50 mg/L aqueous polyelectrolyte solution was circulated through the shell side of the membrane at 1 bar with a bench top NF set-up overnight. The polyelectrolyte adsorbed membrane was then rinsed from the shell to the lumen side at 1 bar to preliminarily remove the unattached chelating polymer. They were transferred to the pilot-scale NF set-up for further cleaning at 10 bar until the flux was stabilized and the conductivities of permeates were below 1 mS/cm. The cleaned membrane was then ready for testing. The procedure of the chemical modification is summarized in Fig. S2 of the supporting material.
2.4.
Characterization of the membrane
2.4.1.
Morphology of the membrane
The morphologies of the plain P84 hollow fiber membranes, the PEI cross-linked membrane and the PAA adsorbed membrane were observed by a field emission scanning electron microscope (FESEM, JEOL, JSM-6700F). To prepare the cross-section samples, the fibers were freeze dried and then fractured in liquid nitrogen.
2.4.2.
Surface charge of the membrane
The surface charge of the membranes was analyzed by a SurPASS electrokinetic analyzer (Anton Paar GmbH, Austria) through streaming potential measurements. To mimic a hollow fiber membrane, flat-sheet membranes with the same dope formulation and post-treatment were prepared. All the membranes were well washed before tests. The testing procedures are as follows. A 0.01 M NaCl solution was used to test the membrane zeta-potential at neutral pH. Then a 0.1 M HCl was used to auto-titrate the solution pH from its neutral pH to pH 2.6. After that, the pH of the 0.01 M NaCl solution was increased gradually from pH 2.6 to pH 11 by auto-titrating the NaCl solution with 0.1 M NaOH. Zeta-potential of the membrane was recorded accordingly. The membrane surface charge as a function of pH and the isoelectric point were then determined.
2.4.3.
Adsorption of heavy metals
The experiments were also conducted using flat-sheet membranes. Three kinds of membranes were prepared; namely,
plain P84 membrane, PEI cross-linked membrane and polyelectrolyte adsorbed membrane. They were freeze dried prior to tests. During the tests, 100 mg of each membrane was cut and immersed into 200 ppm heavy metal solutions containing different single salts individually. The mixtures were then rotated on a roller continuously. The concentrations of the solutions were tracked even after reaching equilibrium, with a total duration of around 60 h. The heavy metal concentrations in the solutions were determined by an inductively coupled plasma optical emission spectroscopy (ICP-OES, Optima 7300DV, Perkin Elmer, USA). The adsorption amounts q (mg/g) of heavy metal ions on each membrane were then calculated through the following formula (Montazer-Rahmati et al., 2011): q¼
c0 ceq V m
(1)
where c0 and ceq represent the initial and equilibrium concentrations of heavy metal ions in the solution (ppm), correspondingly, V is the solution volume (L), and m is the mass of the dry membrane (g).
2.4.4. Pure water permeability and salt rejection of the modified membrane To test the pure water permeability (PWP), DI water was circulated in the shell side of the hollow fiber modules at 10 bar under a constant flow rate of 0.3 L/min and the permeate water was collected within a certain duration. PWP (Lm2 bar1 h1) was calculated using the following formula: PWP ¼
Q ADP
(2)
where Q is the water permeability at the permeate side (L/h), A is the effective filtration area (m2) and DP is the transmembrane pressure (bar). In addition to test single salt heavy metal solutions that comprising 1000 ppm different heavy metal ions, two kinds of mixed ion solutions were prepared according to the compatibilities of the ions. The first kind consists of 1200 ppm Cu, Ni, Zn and Cr ions (each 300 ppm). The second type comprises 300 ppm Pb and 300 ppm Cd ions. Prior to testing, the fibers were rinsed to ensure that the conductivity of the permeate solution was below 1 mS/cm and the PWP was stabilized. During the tests, feed and permeate solutions were collected after the modules were stabilized at least 1 h with salt solutions circulating at 10 bar in the shell side. Each test was measured three times by consecutively taking three samples with a time interval of 1 h to ensure the modules were stabilized for measurements. The ion rejections and their standard deviations were calculated based on the average and standard deviations of the three measurements to ensure that the deviation was within 2%. The concentrations of both the feed and the permeate solutions were determined by ICP-OES. The salt rejection was calculated by the following formula (3): R¼
cp 100% 1 cf
(3)
where cp and cf are concentrations of the permeate and the feed solutions, respectively. To test the stability of the adsorbed membrane after acid or base wash, the chelating polymer adsorbed membrane was washed with a diluted
w a t e r r e s e a r c h 6 3 ( 2 0 1 4 ) 2 5 2 e2 6 1
HNO3 solution (pH ¼ 2.2) or NaOH solution (pH ¼ 10) at 1 bar from the shell side for two days. The membrane was cleaned before performance tests.
3.
Results and discussions
3.1.
Characteristics of the membranes
In this part, PAA 100K is used as the representative of chelating polymers for all characterizations because the density of its negatively charged functional group is between that of PAM and PDMED. For easy discussion, the PEI crosslinked membrane is denoted as Plain-PEI, while the PAA 100K adsorbed membrane is signified as Plain-PEI&PAA 100K.
255
Hollow fiber membranes comprising a macrovoid-free cross-sectional structure are preferred in NF because they may have (1) high mechanical strength to sustain high operation pressures, (2) less defective surface for high rejection and (3) more uniform pore size distribution (Peng et al., 2008; Sun et al., 2010). The ways to fabricate macrovoid-free hollow fiber membranes have been summarized in literature (Peng et al., 2012). A macrovoid-free hollow fiber membrane was therefore achieved in this study by manipulating the spinning conditions as well as the dope solution. As listed in Table 1, a dope solution consisting of a high polymer concentration of 28 wt % and a non-solvent of methanol was prepared. The non-solvent additive was chosen because the resultant dope solution can be spun at a high take-up speed to achieve a macrovoid-free structure (Wang et al., 1999). The
Fig. 1 e FESEM images of the plain P84 membrane (a) cross-section (£200), (b) cross-section (£10,000), (c) outer surface (£50,000) and (d) inner surface (£10,000); and membrane outer surface with post-treatment (e) PEI cross-linked (Plain-PEI, £50,000) and (f) PEI cross-linked and PAA adsorbed (Plain-PEI&PAA 100K, £50,000).
w a t e r r e s e a r c h 6 3 ( 2 0 1 4 ) 2 5 2 e2 6 1
FESEM images of the plain P84 membranes are presented in Fig. 1. At a lower magnification (Fig. 1a), the membrane crosssection looks quite dense. However, a highly porous structure can be observed at a higher magnification (Fig. 1b). No obvious pore structures can be seen on the outer surface (Fig. 1c), indicating the membrane selective layer is apparently defectfree. Besides, the inner layer of the substrate may pose little resistance to water transport as big holes can be found on the inner surface (Fig. 1d). After PEI cross-linking, some small nodules are observed on the membrane surface (Fig. 1e). The membrane becomes a little rougher after the adsorption of PAA 100K (Fig. 1f). Fig. 2 shows the ATR-FTIR spectra of the pristine P84, PlainPEI and Plain-PEI&PAA 100K membranes. Instead of representative peaks of imide at 1770 cm-1 (C]O stretching), 1716 cm1 (C]O stretching) and 1357 cm1 (CeN stretching), new peaks at 1651 cm1 (C]O stretching) and 1543 cm1 (CeN stretching) appear after PEI cross-linking, implying the formation of the amide group (Ba et al., 2009; Gao et al., 2014). After the PAA 100K adsorption, an additional peak was found at 1735e1708 cm1 except all the peaks in Plain-PEI, indicating the successfully adsorption of PAA (Ba et al., 2010). The surface charge characteristics as a function of pH of the plain P84, PEI cross-linked and PAA 100K adsorbed membranes were shown in Fig. 3. The plain P84 membrane has an isoelectric point of 3.2, which is similar to our previous characterizations of polyimide membranes (Gao et al., 2014). After the PEI cross-linking modification, due to the protonation of the amine groups of PEI and the amide groups formed between PEI and P84, the PEI cross-linked membrane is highly positively charged below its isoelectric point of pH 9.4. After the PAA 100K adsorption, the membrane becomes more negatively charged compared to the Plain-PEI because the isoelectric point shifts to pH 6.2 possibly due to the protonation of amine or amide groups (below pH 9.4) and deprotonation of carboxyl groups. The nearly neutral surface charge between pH 6 and 8 may help decrease membrane fouling (Ba and Economy, 2010; Peng et al., 2004).
Fig. 2 e Comparison of ATR-FTRI spectra among the plain P84 membrane, PEI cross-linked membrane (Plain-PEI) and PAA adsorbed membrane (Plain-PEI&PAA 100K).
30
Plain-PEI Plain-PEI&PAA Plain P84 membrane
20
Zeta-potential, (mV)
256
10
0 0
2
4
6
8
10
12
-10 -20 -30 -40
pH, (-)
Fig. 3 e Comparison of zeta-potential trends as a function of pH and the isoelectric points among the plain P84 membrane (isoelectric point: pH 3.2), PEI cross-linked membrane (Plain-PEI, isoelectric point: pH 9.4) and PAA adsorbed membrane (Plain-PEI&PAA 100K).
3.2. Influence of chelating polymer types and molecular weights on NF performance Three types of heavy metal ions, Pb2þ, Cd2þ complex ions (because CdCl2 forms complex metal ions of Cd2þ, CdClþ, 2 CdCl-3 and CdCl2 4 in solutions (Zhu et al., 2014)) and Cr2O7 , were chosen for comparison purposes in this study to represent positive charged ion, complex ion and negatively charged ion, respectively. The influences of different chelating polymers such as PAA 100K, PAM and PDMED on rejection and water permeability of chelating polymer adsorbed membranes are shown in Fig. 4a. The plain-PEI membrane works as the control. The Plain-PEI membrane has a good rejection to Pb2þ (99.15%) and a slightly lower rejection to Cd2þ complex ions (98%). However, (less than 80%) due to the it has a poor rejection to Cr2O2 7 positively charged membrane surface. After adsorption of chelating polymers, all membranes show much higher re2þ and Cd2þ jections to Cr2O2 7 . Besides, their rejections to Pb complex ions also improve. Clearly, the adsorption of chelating polymers with negatively charged functional groups on membranes helps improve the ion rejections, especially the negatively charged ions. When comparing the influences of different chelating polymers to metal ion rejections, the following trend can be found: Plain-PEI&PAM > Plain-PEI&PAA 100K > Plain-PEI&PDMED. It is easy to interpret this phenomenon as PAM has the highest density of negatively charged functional groups among the three polymers. More PAM may be adsorbed on the membrane, making the membrane surface more negatively charged. For PDMED, only the hydroxyl group as the negatively charged functional group is found on the molecular structure. Compared to PAA 100K and PAM, PDMED has more positively charged groups. Though the Plain-PEI&PDMED membrane has a little lower rejection than those of Plain-PEI&PAA 100K and Plain-PEI&PAM membranes, it has performance better than the Plain-PEI membrane. Water permeability for chelating polymer adsorbed
257
w a t e r r e s e a r c h 6 3 ( 2 0 1 4 ) 2 5 2 e2 6 1
Fig. 4 e Influence of (a) different chelating polymers (b) PAA molecular weight on ion rejection (top) and water permeability (bottom) of the membranes.
3.3.
Heavy metal adsorption tests of the membranes
The adsorption of heavy metal ions on the plain P84, Plain-PEI and Plain-PEI&PAA 100K membranes were investigated with the data shown in Fig. 5. The saturation time of the membranes to heavy metals was around 16 h. However, to parallel the long term stability tests as shown in the latter section, the data was collected after 60 h. Among the three membranes, PlainPEI&PAA 100K has the highest adsorption of Pb2þ and Cd2þ complex ion, while Plain-PEI has the highest adsorption of Cr2O2 7 . The plain P84 membrane contributes the least toward heavy metal adsorption. The results confirm that the chelating
polymers on Plain-PEI (PEI is also a chelating polymer (Navarro et al., 1999)) and Plain-PEI&PAA 100K membranes promote the adsorption of heavy metals on the membrane surface. Thus, besides size exclusion and charge repulsion, the newly modified NF membrane can also remove heavy metal ions with the aid of adsorption.
3.4. Rejection tests of the Plain-PEI&PAA 100K membrane Fig. 6 shows ion rejections and water permeability of the PAA adsorbed membrane (Plain-PEI&PAA 100K) to different single
25
adsorption, q(mg/g)
membranes follows the trend of: Plain-PEI&PDMED > PlainPEI&PAA 100K > Plain-PEI&PAM, which is the reverse order of rejections. However, it is lower than that of the Plain-PEI membrane, indicating the successful adsorption of chelating polymer (Jermann et al., 2007). To investigate the influence of molecular weight of the chelating polymer on charge properties and separation performance of the PEI cross-linked membrane, PAA was chosen for experiments because: (1) the rejections and permeability of the Plain-PEI&PAA membrane are in between those of PlainPEI&PDMED and Plain-PEI&PAM membranes, (2) the PAA adsorbed membrane has certain antifouling properties according to Ba et al. (Ba et al., 2010), and (3) several molecular weights of PAA such as PAA 2K, PAA 100K and PAA 250K are available in the market. Fig. 4b shows the influence of PAA molecular weight on the membrane performance. No significant effects of PAA molecular weight on ion rejection and water permeability can be found. Thus, PAA 100K is chosen for the following experiments.
20
Cr
Cd
Pb
15 10
5 0 Plain P84 membrane
Plain-PEI
Plain-PEI&PAA
Membrane type Fig. 5 e Adsorption of heavy metal ions on the plain P84, PEI cross-linked (Plain-PEI) and PAA adsorbed (PlainPEI&PAA 100K) membranes.
258
w a t e r r e s e a r c h 6 3 ( 2 0 1 4 ) 2 5 2 e2 6 1
Fig. 6 e Ion rejections (left) and water permeability (right) of the PAA adsorbed membrane (Plain-PEI&PAA 100K) to different single heavy metal ions.
heavy metal ions. As aforementioned, size exclusion (steric effect) and charge repulsion (Donnan exclusion) are two main mechanisms that contribute to the rejections of NF membranes. In this study, the steric effect is primarily related to the hydrated radii of heavy metal ions, while the Donnan exclusion effect mostly depends on the charge of heavy metal ions as well as the surface charge of the membrane. Hydrated radii (Nightingale, 1959; Volkov et al., 1997) and diffusivities (Lv et al., 2008; Mortimer, 2008; Newman, 1991; Sato et al., 1996; Wang et al., 2007) of all the heavy metal ions tested are summarized in Table S1 of the supporting material. These heavy metal ions can be classified into three categories, which are simple cations (Cu2þ, Ni2þ and Pb2þ), oxoanions (Cr2O2 7 2þ and HAsO2 complex ions and 4 ), and complex metal ions (Zn Cd2þ complex ions). All the simple cations in this study have the same charge, i.e. þ2 charges. Rejections of these cations are all significantly high and follow the weakly trend of Pb2þ (99.48%) < Ni2þ (99.73%) < Cu2þ (99.95%), which is the order of hydrated radii of these three ions. For the oxoanions, the rejection of HAsO42 (99.52%) is much higher than that of Cr2O2 7 (93.55%). It can be explained in two ways. Firstly, the pH of Na2HAsO4 solution (pH 7.7) is much higher than that of Na2Cr2O7 solution (pH 4.9). The membrane is more negatively charged in the Na2HAsO4 solution. Secondly, the hydrated radius of an ion is reverse related to diffusivity and crystal radius (Tansel et al., 2006; Zaikov et al., 1988). To compare the 2 2 hydrated radii of HAsO2 4 and Cr2O7 , CrO4 is introduced as a 2 reference. The hydrated radius of Cr2O7 is lower than that of 2 CrO2 4 because Cr2O7 has a larger crystal radius (Zhu et al., 2014). However, the hydrated radius of CrO2 4 is smaller than 2 that of HAsO2 4 because HAsO4 has a smaller diffusivity as shown in Table S1 of the supporting material. Thus, HAsO2 4 has a larger hydrated radius than Cr2O2 7 . Based on the above two reasons, the rejection to HAsO2 4 is higher than that of 2 . However, the rejection of Cr can be improved Cr2O2 7 2O7 significantly by increasing the solution pH as the membrane would become more negatively charged. The rejection of Cr2O2 7 increases from 93.55% to 98.93% when the pH of the solution is adjusted from pH 4.9 to pH 10. For complex metal ions, the rejection of Zn2þ complex ions is higher than that of
Cd2þ complex ions probably due to the larger hydrated radius of Zn2þ. The water permeability of almost all the salt solutions is above 1 Lm2 bar1 h1. The performance of the membrane is comparable to literature for the removal of heavy metal ions by hollow fiber membranes as shown in Table 2 (Gao et al., 2014; Lv et al., 2008; Wang and Chung, 2006; Wang et al., 2007; Wei et al., 2013; Zhu et al., 2014). The rejections of the Plain-PEI&PAA 100K to mixed heavy metal ions are also tested, as shown in Fig. 7. As described in Section 2.4.4, two kinds of mixed ions solutions were prepared. For both solutions, all the ion rejections are above 99.8% except for Cr2O2 7 , which is 97.6%. Compared to Fig. 6, the rejections of most ions in mixed ion tests are slightly higher than their counterparts in single ion tests because a much lower ion concentration was used in the mixed ion tests (each ion is 300 ppm). The water permeability of the first mixed solution made of four ions (Cu, Ni, Zn and Cr ions) decreases a bit, while the second mixed solution made of two ions (Pb and Cd ions) increases a bit compared to the single ion solutions. It may be related to the total salt concentration in the solution because a high salt concentration normally contribute to a low flux due to the effects of concentration polarization (Kilduff et al., 2004).
3.5.
The stability of the Plain-PEI&PAA 100K membrane
Fig. 8 (left) shows the long term stability of the Plain-PEI&PAA 100K membrane for a continuous rejection test of 60 h. CdCl2 was chosen for the experiment because it forms both cation and anions in water and has high toxicity to human beings (Zhu et al., 2014). The rejection remains stable and high throughout the entire experiment. The water permeability also exhibits good stability but it slightly decreases as a function of time probably due to scaling which is normal in NF tests. However, the water permeability can be recovered by normal cleaning with DI water. Various heavy metal ions were also used to test the membrane stability after different pure wash circles. Fig. 8 (right) shows the results where the separation performance of Cd2þ complex ions performs as a reference at certain wash
259
w a t e r r e s e a r c h 6 3 ( 2 0 1 4 ) 2 5 2 e2 6 1
Table 2 e Selected references on removal of heavy metal ions by hollow fiber NF membranes. Ion As (V) (NaH2AsO4) Cu2þ(CuSO4) Cu2þ(CuCl2) Cr (VI) (Na2Cr2O7) Cr (VI) (K2Cr2O7) Ni2þ Cr6þ Cu2þ Pb2þ(Pb(NO3)2) Cd2þ (CdCl2) Pb2þ(Pb(NO3)2) Cr (VI) (Na2Cr2O7) Cd2þ, Zn2þ Pb2þ,Ni2þ,Cu2þ As (V),Cr (VI) Cr (VI)
Testing condition
Rejection, (%)
Feed: 75 ppm (pH ¼ 12) Feed: 64 ppm Feed: 64 ppm P ¼ 15 bar Feed: 3771 ppm pH ¼ 13.25, P ¼ 25 bar Feed: 520 ppm pH ¼ 12, P ¼ 20 bar Feed: 142.23 ppm Feed: 121.23 ppm Feed: 56.55 ppm pH: 2.31, P ¼ 4 bar Feed: 1000 ppm P ¼ 13 bar Feed: 1000 ppm Feed: 1000 ppm Feed: 1000 ppm (pH ¼ 12) P ¼ 1 bar Feed: 1000 ppm Feed: 1000 ppm Feed: 1000 ppm Feed: 1000 ppm (pH ¼ 12)
Water permeability, (Lm2 bar1 h1)
Reference
97 92.5 ~99
1.86 (pure water)
Lv et al., 2008
97.5
0.025
Wang et al., 2007
95.7
1.86 (pure water)
Wang and Chung, 2006
94.99 95.76 95.33
~8.2
Wei et al., 2013
91.05
0.98 (pure water)
Gao et al., 2014
95 93 98
0.826 (pure water)
Zhu et al., 2014
>98 >99 >98, >93 >98
>1 (most ions)
This study
circles to check the membrane sustainability. It can be found that the membrane has high rejections for all ions even after 10 wash circles. By examining the separation performance of Cd2þ complex ions after 5 circles and 10 circles, a slight increase in permeability (~0.4 Lm2 bar1 h1) was noticed with a very minor fluctuation in rejection of
