Introduction

Since the development of asymmetric membrane by Loeb and Sourirajan (Loeb and Sourirajan 1963), membrane science and technology have attracted much attentions on the topic of separation science. Especially in industrial processes, it is receiving increasing interest in water and wastewater treatments, as well as in water recycling (Parameshwaran et al. 2001; Khayet et al. 2004). Unfortunately, the separation performance is limited by membrane fouling, which causes a decline in permeate flux, leading to increasing pressure with time and increases in the costs of water treatment (Zhao et al. 2000; Clech et al. 2006). Membrane fouling is a very complex process, which is not only influenced by the filtration mode (Mosqueda-Jimenez et al. 2004), but by the hydrodynamic conditions (Hong and Elimelech 1997), such as crossflow velocity and applied pressure. Moreover, the feed water chemistry (pH, ionic strength and divalent cation concentration) and membrane characteristics (pore size, membrane charge, and roughness) play important roles in the permeating performance (Burns and Zydney 1999; Capar et al. 2006; Pujar and Zydney 1994; Mehta and Zydney 2006; Teixeira et al. 2005). The permeability depends on the interactions between the foulant and the membrane and between the foulant and the fouling layer. Effects of the individual component such as bacteria, yeast, proteins and colloids have been examined to identify the mechanisms involved, as well as the influence of operating parameters in microfiltration and ultrafiltration (Belfort et al. 1994; Chan and Chen 2004; Marshall et al. 1993; Fane and Fell 1987). Furthermore, established models have been used to fitting the flux decline during filtration (Duclos-Orsello et al. 2006). However, there are significant gaps in our knowledge about mixed components in the filtration processes, because most of these models are only based on the effects of feed water components on the pore size of the membrane. Little attention has been directly put on the influence of the structural parameters of the membranes (pore size and pore density) on their permeating performance.

Cellulose is one of the most abundant and renewable natural polymers with relatively low cost, good compatibility with biological compounds, remarkable hydrophilic properties as well as its solvent resistance (Samir et al. 2004; Kim and Yun 2006). Moreover, regenerated cellulose membranes have also been widely used in membrane separation technique such as dialysis, ultrafiltration, fractionation and purification of mixtures etc. Dissolution of cellulose in alkali aqueous systems have been attracted much attention (Egal et al. 2008; Cuissinat and Navard 2006; Sescousse and Budtova 2009; Gavillon and Budtova 2007). Recently, we have developed an aqueous solvent system for cellulose dissolution at low temperature, namely 7.0 wt% NaOH/12.0 wt% urea, 9.5 wt% NaOH/4.5 wt% thiourea and 5 wt% LiOH/12 wt% urea precooled to −5 ~ −12 °C, and they have been used to prepare cellulose membranes (Cai et al. 2007a, b, 2008; Lue et al. 2007; Ruan et al. 2004a; Liu and Zhang 2009; Qi et al. 2008). It is noted that LiOH/urea/water solution system has a more powerful ability to dissolve cellulose, and the pore size and the porosity of the cellulose membranes can be controlled by changing the concentration of the cellulose solution and the coagulation temperature (Liu and Zhang 2009). It is well known that in ultrafiltration or microfiltration, the transmission rate of solute is determined mainly by the relative size of the solute and the membrane pores. Therefore, the pore size, size distribution and the density of the membranes are the important parameters for the separation technology. To fit the separation process of the cellulose membrane used in wastewater treatment, we attempted to use water-soluble synthetic and natural polymers as organic matter to evaluate the microfiltration performance of the regenerated cellulose membrane in aqueous system. This work can provide some information to improve the separation materials and to extend the applications of cellulose.

Experimental sections

Materials

Cellulose (cotton linter pulp) used was provided by Hubei Chemical Fiber Group Ltd (Xiangfan, China). The cellulose sample was washed with distilled water for five times and dried in a vacuum oven at 70 °C before use. Its viscosity-average molecular weight (M η) was determined by using an Ubbelohde viscometer in LiOH/urea aqueous solution at 25 ± 0.05 °C and calculated from the equation [η] = 3.72 × 10−2 M 0.77 w to be 13.6 × 104 (Cai et al. 2008). Analytical grade Polyethylene glycol (PEG) and dextran with different molecular weight were, respectively provided by Xilong Chemical Ltd (China) and Sigma (USA). Other chemical regents were purchased from China with analytical grade and used without further purification.

Preparation of regenerated cellulose membrane

About 200 g aqueous LiOH/urea solution (5.0 wt%/12.0 wt%) was precooled to −12 °C in a refrigerator, and then a desired amount of cellulose was added into the solution immediately, with stirring for 5 min to dissolve. The resultant cellulose solution was subjected to centrifugation at 8,000 rpm for 10 min at 10 °C to eliminate bubbles in the viscous solution. The viscous bubble-free solution was casted on a glass plate and the thickness of the membranes was controlled to be 0.25 mm, and then the glass plate was immersed into the coagulating bath containing 5 wt% H2SO4 to coagulate and regenerate for 5 min at 30 °C. The regenerated cellulose (RC) membranes were firstly washed with running water, then with deionized water, and finally air-dried at ambient temperature. The RC membranes prepared from the cellulose solution with concentration of 4, 4.5, 5 and 5.5 wt% were coded as RC-4, RC-4.5, RC-5 and RC-5.5.

Membranes characterization

Wide angle X-ray diffraction (WAXD) measurement was carried out on a WAXD diffractometer (D8-Advance, Bruker, USA) in symmetric reflection mode. The patterns with Cu Kα radiation (the weighted average λ is 0.15418 nm) at 40 kV and 30 mA were recorded in the 2θ range of 4 to 40o. Native and regenerated cellulose were ground into powders and dried in vacuum oven at 60 °C for 48 h before testing. The crystallinity χc (%) of the samples was estimated by (Rabek 1980):

$$ \chi_{\text{c}} = {{\int\limits_{0}^{\infty } {S^{2} I_{c} \left( s \right)ds} } \mathord{\left/ {\vphantom {{\int\limits_{0}^{\infty } {S^{2} I_{c} \left( s \right)ds} } {\int\limits_{0}^{\infty } {S^{2} I\left( s \right)ds} }}} \right. \kern-\nulldelimiterspace} {\int\limits_{0}^{\infty } {S^{2} I\left( s \right)ds} }} $$
(1)

where s is the magnitude of the reciprocal-lattice vector and is given by s = (2sinθ)/λ, θ is one-half the scattering angle. λ is the X-ray wavelength, I(s) is the intensity of the coherent X-ray scattering from a specimen (both crystalline and amorphous), I(c) is the intensity of coherent X-ray scattering from the crystalline region.

Fourier-transform infrared (FTIR) spectra of the natural and RC membranes were recorded with FT-IR spectrometer (model 1600, Perkin-Elmer Co. USA). The tested samples were prepared by the KBr-disk method. The morphologies of the RC membranes were characterized by using scanning electron micrographs (SEM, FEI QUANTA 200, PHILIP). The wet membranes were frozen in liquid nitrogen snapped immediately, and then freeze-dried. The free surface (the side in contact with the coagulant) and the back surface (the side in contact with the glass plate) of the membranes were sputtered with gold before observation.

The equilibrium-swelling ratio (Q) of the wet membranes was estimated as follows. A piece of the membrane was immersed in distilled water for 5 days until it reached the swelling-equilibrium state, and then removed from the storage solution. The water on its surface was wiped off, and the wet membrane was weighed. The weighed membrane was dried in a vacuum oven at 70 °C for about 48 h and reweighed. The equilibrium-swelling ratio (Q) can be obtained (Bell and Peppas 1996):

$$ Q = {\frac{{W_{h} - W_{d} }}{{W_{d} }}} $$
(2)

where W h is the equilibrium weight of the swollen sample and W d is the weight of the dried sample. Five replicates were performed to obtain an average value of Q.

The porosity of the wet RC membrane (P r ) with specific area can be estimated by measuring its thickness d (cm) at the wet state and weight W (g) at the dry state and calculated by (Yang and Zhang 1996):

$$ P_{r} = 1 - {\frac{W}{{\rho_{c} \pi r^{2} d}}} $$
(3)

where r is the radius of the wet membrane, and ρ c is the density of bulk cellulose (1.5 g cm−3).

The pure water permeability test of the wet membranes was carried out on a miniature microfiltration equipment at ambient temperature under 0.1 MPa. Circular membrane discs were set with a porous support and tightened by a rubber O-ring. The effective permeation area of each membrane was about 7.07 cm2. Prior to testing, the pure water flux was measured to ensure that the membrane used was stable. The ultrafiltration rate (UFR) of the wet membrane for pure water was calculated by:

$$ UFR = \frac{V}{StP} $$
(4)

where UFR is the pure water permeability (mL h−1 m−2 Hg−1); V is the volume of permeated water, t the time (h); P the pressure (Pa), and S is the effective surface areas of the membrane (7.07 cm2). The pore sizes (2rf) of the cellulose membranes were determined by the Hagen-Poiseuille law:

$$ \upsilon = {\frac{{\Updelta pd^{2} }}{32\mu d}} $$
(5)

where υ is the fluid velocity of fluid, Δp is the pressure drop across a membrane with thickness of d, μ is the viscosity of the fluid.

The tensile strength (σ b) and elongation at break (ε b) of the membrane at the dry and wet states were measured on a universal tensile tester (CMT 6503, Shenzhen SANS Test machine Co. Ltd, China) according to ISO 527-2, 1993 (E) at a speed of 5 mm min−1, The σ b and ε b values were the average results of five measurements.

Structure parameter model for membrane

According to the filtration processes of a membrane, the dissolved matters in feed water are divided into two categories. One is correlated to the dissolved matter with low molecular size in feed water, which may permeate or form an adsorption layer inside the pores of the membrane and thus reducing the membrane pore size. The other refers to matter with larger molecular size that may form clogging on the membrane surface and reduce the pore density. Therefore, the pollutants in feed water may change the pore density and pore size of the membrane during the permeating process, and they are the important parameters to reflect the membrane filtration characteristics. By using the Hagen-Poiseuille equation, solvent flux (J vi ) through the pores having diameter R i can be expressed as (Singh et al. 1998):

$$ J_{vi} = {\frac{{\pi \Updelta pR_{i}^{4} }}{128\eta \delta }}N_{i} $$
(6)

where η is the solvent viscosity, δ is the length of the pores; J vi is flux of the feed water, N i is the membrane pore density having diameter of R i (the membrane pore diameter is a function of time during the permeating process, so i is for a given time step; i = 1, 2, 3…).

According to the hypothesis, during the membrane filtration processes of the feed water, the values R 0 and N 0 of the original membrane may change with accumulated permeate water volume per unit membrane area (q i ) as (Wang and Wang 2006):

$$ R_{i} = (1 - a_{1} q_{i} )R_{0} $$
(7)
$$ N_{i} = (1 - a_{2} q_{i} )N_{0} $$
(8)

where a 1 is the mean pore size reducing factor, and a 2 is pore density reducing factor. So, the following equations can be deduced from Eqs. (6), (7), (8):

$$ J_{vi} = {\frac{{pDP(1 - a_{1} q_{i} )^{4} R_{0}^{4} }}{128hd}}(1 - a_{2} q_{i} )N_{0} $$
(9)

where J vi is the accumulated flux of feed water per unit membrane area when the accumulated permeate volume per unit membrane area is q i . Equation (9) is defined as the membrane structure parameter model, which reflects that membrane flux changes with the accumulated permeate water volume per unit membrane area, according to Eqs. (6), (9), the following formula can be deduced:

$$ {\frac{{J_{vi} }}{{J_{0} }}} = (1 - a_{2} q_{i} )(1 - a_{1} q_{i} )^{4} $$
(10)

A series of experimental values of J vi /J 0 and accumulated permeate per unit membrane area q i are combined with Eq. (10), then the optimal values of a 1 and a 2 can be calculated via computer simulation.

Results and discussion

Structure and pore size of the RC membranes

Figure 1 shows the WAXD patterns of the native cellulose sample and regenerated cellulose membranes. The crystalline form of native cellulose has typical diffraction peaks at 2θ = 14.9, 16.5 and 22.7°. The diffraction peaks at 2θ = 12.2, 20.2 and 21.6° of the RC membranes observed were assigned to the (1ī0), (110) and (200) planes of cellulose II crystal (Isogai et al. 1989). This indicated clearly that a conversion from cellulose I (native cellulose) to cellulose II (regenerated cellulose) occurred during dissolution and coagulation in the bath containing 5 wt% H2SO4. The values of χc (%) of the samples calculated from WAXD patterns are listed in Table 1. The χc (%) values of the RC membranes were lower than that of native cellulose, suggesting that the crystal structure of the native cellulose has been destroyed in the LiOH/urea solution, and a new crystal structure formed partly, leading to the predominance of non-crystalline regions. It was worth noting that the χc (%) value of the RC membrane (33.2%) prepared from 4 wt% cellulose solution was lower than that of the regenerated cellulose membranes prepared from NaOH/urea and NaOH/thiourea solution system with the same concentration (Zhou et al. 2002; Ruan et al. 2004b). It also suggested that LiOH/urea solution had a stronger ability to dissolve cellulose than NaOH/urea system by destroying the crystalline structure of the native cellulose.

Fig. 1
figure 1

WAXD patterns of the native cellulose and regenerated cellulose membranes

Full size image
Table 1 Experimental results of the regenerated cellulose membranes prepared with different concentrations of cellulose
Full size table

FT-IR spectra of the native cellulose and RC membranes prepared from different concentration was shown in Fig. 2. The bands at 1,431 and 1,319 cm−1, assigned to symmetric bending and wagging of –CH2, respectively, shifted to lower wavenumbers in comparison with that of the native cellulose, indicating that a destruction of intermolecular hydrogen bonds involving the exocyclic hydroxyl groups (O6) occurred. The band at 896 cm−1, characteristic of the amorphous regions of cellulose, moved toward to lower wavenumbers. This further provided an evidence for the transformation of the crystalline state from cellulose Ι to II (Sang et al. 2005). In comparison with native cellulose, the bands at 1,431 and 896 cm−1 for the RC membranes became weaker and stronger, respectively, indicating the decreased crystallinity of the regenerated cellulose. The results from FT-IR spectroscopy and X-ray diffraction demonstrated that a transition from cellulose Ι to cellulose II occurred during the membrane formation process.

Fig. 2
figure 2

FT-IR spectra of the native cellulose and the RC films prepared from different concentrations of cellulose

Full size image

Figure 3 shows the SEM micrographs of the free surface (the side directly in contact with the coagulant) and back surface (the side in contact with glass) of the RC membranes. All of the membranes displayed an asymmetric structure, namely, the pore size on the free surface was different from that on the back surface. With an increase of the cellulose concentration, the surface morphology of the membranes became denser, and the pore size (both the free and the back surfaces) decreased. The detail data of P r , 2r f and UFR of the RC membranes are summarized in Table 1. The RC-4 membrane had a distinct asymmetric structure including the free surface with many macrovoids and back surface with larger pores (Fig. 3a, e). However, when the concentration of cellulose was 4.5 wt%, the macroporous structure on the back surface and interpenetrated network structure on the free surface were formed, and the macrovoids became inconspicuous (Fig. 3b, f). In view of the results in Table 1, with an increase of the cellulose concentration from 4.0 to 5.5 wt%, the pore size (2r f ) of the cellulose membranes decreased from 52 to 21 nm. Therefore, by changing the cellulose concentration we can control the pore size of the RC membranes.

Fig. 3
figure 3

SEM imagines of the back surface (a, b, c, d) and the free surface (e, f, g, h) of regenerated cellulose membranes prepared from different concentrations: 4% (a, e), 4.5% (b, f), 5% (c, g) and 5.5% (d, h), respectively

Full size image

On the basis of the process of porous membrane formation (Kamide et al. 1993), the diffusion and convection between cellulose solution and coagulant play an important role in controlling the morphology and pore size of the membranes. In our findings, the cellulose was dissolved in LiOH/urea solvent to form clear liquid. When the cast sheet from the cellulose solution was immersed into coagulation bath, the diffusion between non-solvent and solvent in the casting cellulose solution (gel state) occurred. (Mao et al. 2006) With a large amount of non-solvent diffusion into the casting solution, the two-phase separation of cellulose rich-phase in the gel sheet and lean-phase in solution occurred, which was responsible for the formation of a porous structure. There was the mesh structure in the cellulose membranes (see Fig. 3), which was weaved with cellulose bunch in the cellulose rich-phase, as a result of the strong self-association force (Weng et al. 2004). In this process, cellulose backbone with semi-stiff molecular chain contributed to support the pore wall, whereas its –OH groups acted as formation of pore, because of its water absorbency. The penetration rate of the coagulation bath on the free surface is faster than that of the back surface, resulting in an asymmetric structure. Moreover, with an increase in the cellulose concentration, the cellulose poor phase decreased, leading to the formation of a relatively denser structure and smaller pore size in the regenerated cellulose membranes.

Effects of pore size and density on the water flux

Figure 4 shows the equilibrium-swelling degree of the regenerated cellulose membranes as a function of cellulose concentration. With an increase of the cellulose concentration, the equilibrium-swelling ratios of the RC membranes decreased, further confirming that the pore size and density decreased. Therefore, there was a large resistance to matters passing through the membranes, leading to a decrease in the flux. The water permeability (UFR) decreased from 11.2 to 8.4 mL h−1 m−2 mm Hg−1, when the concentration of the cellulose solution increased from 4 to 5.5 wt%. The results from UFR and water content of the membranes supported that the pore size decreased with an increase of the cellulose concentration.

Fig. 4
figure 4

Equilibrium-swelling degree (Q) of the regenerated cellulose membranes as a function of cellulose concentrations

Full size image

The influences of the cellulose concentration on the tensile strength (σb) and elongation at break (εb) of the RC membranes at the dry and the wet state are shown in Fig. 5. The mechanical properties of the RC membranes prepared from the cellulose solution with concentration in the range of 4.0–5.5% changed hardly, and the mechanical properties of the RC membranes prepared from this LiOH/urea system was higher than those prepared from NaOH/urea and NaOH/thiourea solution system, as well as the commercial cellophane. Particularly, the RC membranes at the wet state exhibited good tensile strength, indicating that RC membrane can be used well at the wet state in water, and would have promising applications in the separation technology fields.

Fig. 5
figure 5

The influence of cellulose concentration on the stress and strain of the RC films: a at dry state, b at wet state

Full size image

Effects of fouling on flux curves

The water-soluble PEG and dextran with different molecular weight as fouling source were added in the feed water. The apparent molecular weights of the organic matters were listed in Table 2. Figure 6 shows the flux curves of RC membranes with the accumulated permeate per unit membrane area. There was a sharp reduction in the flux at the beginning of the filtration process, which was an indicative of rather rapid reversible fouling taking place initially. The nearly constant permeate flux provided an indication of the formation of a surface fouling layer, probably resulting in a more compact and/or thicker layer, leading to the reduced flux. The fitted results from the established model were close to the experimental value, suggesting that the microfiltration membrane structure parameter model based on this study could be used to evaluate the microfiltration process of water. From the results in Table 2 and Fig. 6, the organic matters with molecular weight of 4–20 kDa played a key role on the control of the fouling of the RC membranes.

Table 2 The apparent molecular weight (M) and the content of organic matter in experimental feed water samples (mg/L)
Full size table
Fig. 6
figure 6

Effects of the components of feed water on the fitted permeating performance of the RC membranes with different pore size and density

Full size image

We calculated the molecular size and compared it to the pore size of the RC membrane according to the reported works (Tao et al. 2007), which was close to the minimum pore size of the cellulose membrane. Based on the results obtained in this work, the organic matter in different water-soluble samples could be divided into two categories: one with molecular weight larger than 20 kDa was coded as sample 1, and the other less than 20 kDa was coded as sample 2. Therefore, sample 2 in water could form adsorption layers inside the membrane pores to reduce the pore size, and sample 1 could form clogging in the membrane surface to decrease the pore density. The corresponding calculated values of a 1 (the membrane pore size change coefficient) and a 2 (the pore density change coefficient) in microfiltration process with water-soluble samples have been summarized in Table 3. The membrane pore density reducing factor a 2 decreased with an increase of the content of sample 1, and the trends of their changes for all the membranes were basically the same. This revealed that sample 1 mainly led to the reduction of the membrane pore density during the membrane filtration process. However, the trend of changes in a 1 from sample was different from that of a 2, it could be ascribed to the greater influence of sample 2 on a 1. When the wastewater had components with higher molecular weight, namely relatively larger hydrated radius (larger than the pore size of cellulose membrane). Therefore, it is difficult for them to penetrate through the cellulose membrane, and forms clogging on the surface or wall up the pores of the membrane to influence on a 2. However, a 1 was related to the components with lower molecular weight. The components with lower molecular weight in wastewater sometimes can be absorbed into the pores of the membrane, and they also can move into solution or penetrate through the membrane. The random values of a 1 would be resulted from the interaction of components with cellulose membrane; it was a dynamic process, leading to the complex process of separation process.

Table 3 The relationship between the fitting d values of model parameters (a 1 and a 2) of four RC membranes and five feed water samples with specified molecular weight (M)
Full size table

Modeling of the fouling behaviors

It has been reported that the salt concentration, ionic strength, Zeta potential, the hydrophilic and hydrophobic properties of organic matters, and as well as membrane parameters such as pore size, surface electric charge and the hydrophilic and hydrophobic properties, also affect the separation performance of the membrane (Yuan and Zydney 1999; Burns and Zydney 2000). In view of the above results, a schematic illustration for the microfiltration process of wastewater is proposed in Fig. 7. According to the pore size of the RC membrane, the organic matters in feed water with larger molecular size than the membrane pore size could form clogging on the membrane surface, leading to a decrease in the pore density of the membrane. However, the dissolved organic matters with molecular size smaller than the membrane pore size could penetrate the membrane or form adsorption layer inside the membrane to reduce membrane pore size. Therefore, the model used in our work could describe clearly the separation process. Appropriate pretreatment could then be adopted to achieve the purposes of reducing membrane fouling, which would promote flux efficiency and prolong the useful life of the membrane, as well as to lighten the environmental burden.

Fig. 7
figure 7

Simplified illustration of microfiltration process of the RC membrane used for wastewater containing organics with different molecular size (a), (b), and fitted model curve and experimental result of RC-4 for microfiltration of wastewater sample 1 (c)

Full size image

Conclusions

Regenerated cellulose membranes with different pore size have been successfully prepared from LiOH/urea solution system precooled to −12 °C. During coagulation, the two-phase separation of cellulose rich-phase in the gel sheet and lean-phase in solution occurred, which was responsible for the formation of a porous structure. The pore size of the regenerated cellulose membrane decreased with an increase of the cellulose concentration, resulting in a decline in permeability of the membrane. However, the mechanical properties of the membrane changed hardly by varying the concentration of cellulose. The water-soluble synthetic and natural polymers as organic matter were used to evaluate the microfiltration performance of the regenerated cellulose membrane in aqueous system. For wastewater treatment the organic matter with molecular weight more than 20 kDa effected significantly on the membrane pore density, and reducing factor a 2, whereas the membrane pore size reducing factor a 1 exhibited little or no relationship with organic matter with molecular weight less than 20 kDa.