Ultrafiltration by a super-hydrophilic regenerated cellulose membrane

Ultrafiltration (UF) membrane systems represent an attractive alternative for the treatment of ground and surface waters for supplies of potable water since they can provide a consistently high quality of drinking water (Yu et al. 2014). UF has also been widely used in industrial processes including pharmaceutical, biotechnological and food industries. Commercial UF membranes are manufactured from common polymers such as polyethersulfone, polysulfone (PS) and polyvinylidine fluoride (PVDF) by the phase inversion process to form an asymmetric porous structure (Nunes & Peinemann 2006; de Souza Araki et al. 2010). It is well known that the phase inversion process is affected by the casting solution composition and the conditions during casting (Low et al. 2014).

PVDF is one of the most promising UF membrane materials due to its outstanding properties such as thermal stability, chemical resistance, and mechanical strength compared to other commercial polymeric material (Tao et al. 2012; Song & Jiang 2013; Sun et al. 2013). However, due to its strong hydrophobic nature, it leads to low water fluxes and renders the membrane easily fouled when aqueous solution containing natural organic matters are treated (Edwie et al. 2012; Rahbari-Sisakht et al. 2012; Cui et al. 2013). Accordingly, many researches have been carried out concerning chemical grafting and surface modification (Chiang et al. 2009; Rahimpour et al. 2009; An et al. 2011; Sui et al. 2012), physical blending (Yuan & Dan-Li 2008; Wang et al. 2012; Zhang et al. 2013b) in order to improve the hydrophilicity of PVDF membranes. Nano-sized TiO₂ which is very hydrophilic and chemically stable also possesses anti-bacterial property (Mansouri et al. 2010; Su et al. 2012) has been utilized in membrane modification by endowing hydrophilicity and increased water permeation (Xi et al. 2009), concomitantly, decreasing the probability of membrane pore blocking. However, aggregation of TiO₂ particles is an inherent drawback since it leads to a weak binding force between the nanoparticles (NPs) and the PVDF polymer (Lee et al. 2008). Thus, a simple way to disperse the NPs onto the PVDF membrane surface has to be searched for. Other membrane surfaces such as PS, polyethylene (PE), polycarbonate, polytetrafluoro-ethylene (PTFE) and PVDF have been subjected to surface modification by dopamine, which is a hydrophilic polymer (Lee et al. 2007). To this end, McCloskey et al. (2010) managed to modify integrally skinned asymmetric membranes of PS, PE, PVDF and PTFE thin-film composite membranes (Shao et al. 2014).

From what has been mentioned above, it is observed that all membranes used in UF are hydrophobic polymers, which are amenable to surface fouling, biofouling and pore blocking much more than hydrophilic polymers due to natural organic matter and proteins. Accordingly, it is also realized that efforts have been made to cover the surface of the hydrophobic membranes with hydrophilic materials in order to reduce the susceptibility to fouling. However, the method suffered from gradual particle leaching from the surface so that the membrane performance declined by time. Moreover, it is noticed that the methods and materials used are tedious and expensive.

The aim of the present work is to fabricate super-hydrophilic membranes to be applied in UF of ground water, surface waters and protein suspensions. The membranes are to be prepared by phase inversion, from a solution composed of cellulose acetate (CA) in different solvents, with or without special additives. The coagulated membranes will be all subjected to complete deacetylation to render them super-hydrophilic by removing the bulky hydrophobic acetyl groups, and freeing the hydrophilic hydroxyl groups. The evaporation time of the as-cast membrane and the coagulation water bath temperature and the effect of annealing will be studied. The percent salt rejection and flux of an aqueous salt solution will be measured in a lab-scale RO unit to determine the capability of the membranes to be used in UF of yeast and NaCl solutions. Finally, the membranes are to be subjected to scanning electron microscopy (SEM) examination, Fourier transform infrared (FT-IR), contact angle, porosity and pore size determination.

CA (Panreac, Gomhoreya pharmaceutical Co., Egypt), acetone (A) (Femco Co., Egypt), dimethyl phthalate (DMP) and activated carbon (AC) (Knock-light laboratories Ltd. Corlnbrook Buoks, England), dioxane (D) (El Nasr pharmaceutical chemicals Co., Egypt), dimethyl formamide (DMF) (Adwic laboratory chemicals, Egypt), sodium chloride and sodium hydroxide (Chemajet Chemical Company, Egypt), were all used as such without further treatment.

CA powder was dissolved in a mixture of solvents in different weight proportions, which include one or more of A, D, DMF, and DMP (Table 1). Additives were added to the mixture, and CA varied in proportions, according to Table 1. After complete dissolution, the tightly closed bottle was left aside for 24 h until complete removal of air bubbles. The solution was cast into a membrane on a smooth uniform glass plate of a casting assembly pre-set to the required thickness. The as-cast membrane was left to evaporate for different time periods (either 0.5 or 1.5 min) on the glass plate, then steeped in a water coagulation bath for 1 h at temperatures shown in Table 1, after which it was washed efficiently with running water followed by decantation, then it was stored in distilled water. The membrane was subjected to complete de-acetylation by steeping for 24 h in an aqueous alkaline bath consisting of 1% sodium hydroxide and 20% sodium chloride. The membrane was washed efficiently with running water followed by decantation then stored in distilled water for later use. Four membranes were subjected to annealing in distilled water at 80 °C for 15 min as shown in Table 1.

The surface and cross-section morphologies of membranes were observed using SEM, (JOEL JSM 6360 LA, Japan). FT-IR of membranes were obtained by the FT-IR spectroscopy (VERTEX 70, Bruker Co., Germany). Water contact angle was employed to evaluate the hydrophilicity of the membrane surface. The static contact angles of water on the membrane surfaces were measured by contact angle goniometer (JC-2000C Contact Angle Meter, Powereach Co., Shanghai, China). The average value of static contact angle on each membrane was calculated with at least five different locations on each membrane.

The FT-IR spectra of the six fabricated UF membranes are illustrated in Figure 1, from which it is clear that the OH functional group is present in all membranes as well as the remaining groups stated in Table 2. There is a great similarity among all the frequencies, in fact they appear to overlap. Besides the typical hydroxyl group which makes the membranes super hydrophilic due to the presence of three OH groups per anhydro-glucose unit, the additional peak 1,640–1,630 m⁻¹ emphasizes the adsorption of water, which both assist in the make and break of hydrogen bonding of water thus causing easy permeation of water across the membrane. In addition, the peak at 1,125–1,170 m⁻¹ which corresponded to C-O-C asymmetric stretching vibration (arabinose side chain), while the peak at 1,040–1,050 m⁻¹ corresponded to C-O stretching in C-O-C glycosidic bonds, which both verified the existence of the ether linkages between the anhydro-glucose units and the asymmetric stretch of the arabinose side chain. It has also been verified that none of the solvents/additives were retained in the membrane matrix and that they were leached out during coagulation and post washing.

It is generally accepted that contact angles indicate whether the surface is hydrophilic or otherwise, so that if the liquid molecules are strongly attracted to the solid molecules, then the liquid drop will completely spread out on the solid surface, corresponding to a contact angle of 0°. In this case, the surface is super hydrophilic, which was the case with the majority of our membranes as presented in Table 3, except the first membrane, which gave 30° contact angle. Accordingly, the results emphasize the super-hydrophilic nature of our membranes even the membrane M1.

The membrane porosity and mean pore size of the fabricated membranes varied according to Table 3 between 66% and 96%, and 14.9 nm and 30.9 nm, respectively. Thus, confirming the pore formation capability and suitability of the membranes in UF. It is noteworthy that porosity is very acceptable.

The membranes were subjected to SEM examination in order to determine the morphology of each membrane, as regards surface topology, number and size of the pores, and surface roughness. Figure 1 shows that all six membranes are asymmetric. M1 which was cast from acetone only, at room temperature, evaporated for 0.5 min, coagulated in distilled water, and had not undergone annealing, gave high flux but relatively high %SR as expected, since the surface pores were relatively wide, due to absence of annealing and short evaporation time, as shown in Figure 2. However, the flux under both pressures varied between 12 and 18 kg/m²h in the early stages, but declined progressively at the higher pressure. Concomitantly, the %SR was in the thereabouts of 8 at first then jumped to 13 and remained constant. On the whole, the membrane performance is more or less acceptable. Also, it exhibited an irregular top layer with surface pores of which many appear to be discontinuous. The membrane cross-section shows that the matrix underlying the skin layer is of medium porosity, which means that the membrane was expected to have a high %SR. In Figure 2(a) two membranes provide higher fluxes than those presented in Figure 3(a). M1 shows a slightly different flux at the two pressures studied. At 1 bar the flux slightly declines then remains almost constant from 20 min thereon. However, at 4.5 bar, the membrane flux declines steadily to about 11 kg/m²h. Thus, the majority of the membranes up till this point suffer from either a moderate or a strong flux decline, which results from membrane compaction under pressure that leads to a continuous drop in flux and presents a serious drawback in such membrane separations requiring high pressure. For this reason, the %SR illustrated in Figure 2(b) shows that at the two pressures applied in case of M1, high %SR which increases strongly with pressure and time, makes the %SR reach a value of 13%SR, which is obviously undesirable.

Membranes M2 and M3 were fabricated from a mixed solvent mixture, in which the acetone was partly replaced by DMF, DMP and D. However, M2 was annealed while M3 was not, but both were evaporated for 1.5 min before coagulation. M2 as shown in Figure 1 reveals a salt rejection layer that contains very ordered, minute and closely packed pores, intact with a porous matrix containing parallel voids. In addition, Figure 3 indicates that the flux and %SR of M2 were lower and higher, in respective order, than those of M3, as expected, due to being annealed, while M3 was not. Moreover, M3 gave a higher flux at both pressures tested (as also shown in Table 3), than M4 and M6, and only 2%SR, which means that it is suitable for UF application, in that salt will be removed with the filtering water, while yeast is rejected.

As to M4 and M6, which appear together with M2 and M3 in Figure 3, it is clear that M4 offers a reasonable flux at 5 bar, which is twice that at 2.5 bar, at the same time it allows complete salt passage (zero %SR). Moreover, the flux declines during the first 50 min only, after which it remains constant. As regards the micrograph of M4, it is clear that the surface contains many micro-pores, and that the voids within the matrix are mostly wide but vary in size. Accordingly, it could be said that M4 might make a promising UF membrane; in addition, yeast was totally rejected on the membrane as shown in Table 3. On comparing M6 to M3 and M4 as shown in Figure 3, it is realized that M6 suffers from progressive flux decline and that it shows the highest %SR, though acceptable, but is not promising due to the first reason. It seems that leaching of the salt added to the casting mixture, resulted in membrane compaction by time, during the UF experiment. This observation is revealed on inspecting Figure 1, which shows that the membrane was the thinnest of all the membranes in the figure (about one half). Moreover, it is noteworthy that the mass ratio of CA to solvents plus additives was larger in this case than for the rest of the membranes, which may have partly contributed to this result. It is worth mentioning that the presence of DMP which is a plasticizer for CA (Elewa, 2006), in admixture with the other solvents/additives could have caused the membrane to be softer than the others, and thus made it compacted during the UF experiment.

The morphology of M6 of which the composition is shown in Table 1 shows that 0.5 g of NaCl and 0.5 g of AC were added to the formulation, and which was evaporated for 1.5 min, coagulated at 25 °C, and annealed at 80 °C. Figure 3(a) proves that the membrane does not give a high flux, and at the same time it suffers from strong flux decline at 4.6 bar, which makes it unsuitable for both reasons as an UF membrane. It is clear from the figure that the pores are moderate in size and that the skin layer is rather thick probably due to the evaporation time being 1.5 min. Also, the expected leaching of the NaCl during the coagulation failed in increasing the porosity of the surface and in enlarging the voids in the matrix. It is noteworthy that strong flux decline resulting from membrane compaction is revealed, by the extreme thinness of the membrane (compare to the others in Figure 1). The flux of M6, which was tested only at 4.6 bar, was of moderate value, which declined strongly during the 140 min of pressurizing, as expected. In addition, from Figure 2(b), %SR is somewhat higher than required, though constant at 4.25.

Inspecting M5 in Figure 1 shows that numerous pores are present on the surface, though seeming to be shallow. However, the cross-section shows an extremely thin rejection layer, that is intact with a lower matrix containing plenty of micro-pores followed by huge voids in the membrane matrix, which, lead to the very high flux provided by the membrane (see also Table 3). Moreover, the thin solid planes between the large voids contain mini-pores embedded inside them, which surely assisted in the high flux obtained. Figure 2(a) also, shows that M5 offered a promising membrane with a high and steady value of flux (27 kg/m²h) at 5 bar. However, at 2.5 bar the flux was still acceptable and almost no flux decline took place. As regards %SR, it is clear from Figure 2(b) that it equals only 2%, which is extremely low and steady. Therefore, M5 provides the best UF membrane prepared in the present work, in that it provides a very high and sustainable flux at 5 bar, with almost no salt rejection over 100 min of operation, in addition that it is a super-hydrophilic membrane. This is expected to combat surface-, pore- and bio- fouling, which are the most serious limitations of commercial hydrophobic membranes that cause the inacceptable water flux through the membrane. This is emphasized by the value of the contact angle shown in Table 3.

In this study, super-hydrophilic membranes with good performance in ultrafiltrating yeast were prepared, that are resistant to fouling. It has been shown that deacetylated CA membranes can provide excellent super-hydrophilic membranes if the proper conditions are taken care of as regards the casting solution composition, additives such as NaCl and AC, evaporation time of the as-cast membrane, coagulation bath temperature and subjection to annealing. It has been found that all variables had an effect on the flux, %SR and morphology of the membranes. It was found that the best membrane was evaporated for 0.5 min and coagulated at 18 °C and subjected to annealing at 80 °C. It presented a perfect asymmetric regenerated cellulose membrane with a very thin rejection layer intact with a matrix that contains very wide pores, which together resulted in only 2%SR and a high flux of 27 kg/m²hr at 5 bar which was constant for a 100 min of operation. Moreover, M5 did not suffer at all from flux decline, and exhibited a zero contact angle interpreted as being super hydrophilic. To this end, all the membranes contained hydroxyl groups and ether linkages, which are known to be very hydrophilic.

The authors wish to acknowledge the kind support of Science and Technology Development Funds (STDF) of Egypt for funding the research project (ID: 4060) on desalination of seawater through application of the sweeping-air pervaporation technique.