Cellulose acetate with 2.5 degrees of acetylation (DA) was purchased from the Chemical Reagent Corporation (Shanghai, China). SL was purchased from Sinopharm Chemical Reagent Beijing Co., Ltd. 1,4-Dioxane (AR) was used as the solvent and deionized water was used as the precipitating solution.

An environmental scanning electron microscope (ESEM; Quanta 200, FEI Co., Holland) equipped with an energy-dispersive X-ray microanalysis (EDAX) system was used to acquire cross-sectional images of the fibers and the content of elements C, N, O and P in different areas of the sample. The cross-section of the sample was prepared by freeze-fracture in liquid nitrogen. A GIKOIB-3 sputter coater with a gold?palladium target was used for coating gold onto the cross-sections.

IR spectra were obtained with a Spectrum FTIR-8400 system from Perkin-Elmer. The wave number was between 4000 cm ⁻¹ and 650 cm ⁻¹. Ultrathin film was prepared in order to improve transparency.

X-ray photoelectron spectroscopy (XPS) was done with an XSAM-800 spectrometer using an Al Kα (1486.7 eV) X-ray source. To remove charging shifts and to cope with Fermi edge coupling problems, binding energies (BE) were scaled against the peak of the C-(C, H) component coming from contaminating carbon (set to 284.6 eV). The pressure inside the analysis chamber was <2 × 10⁻⁸ Pa. The C1s, O1s, N1s and P2p spectra were recorded, and surface atomic compositions were determined.

A cyclic voltammogram (CV) was obtained to assess the electrical conductivity using a 2500 A electrochemical system (Lanlike Co., China). The electrolytic solution was 0.005 mol・L⁻¹ potassium ferricyanide and 0.1 mol・L⁻¹ potassium chloride. A CA or L-CA film sample was coated onto the surface of a glassy carbon electrode (GCE) to make a working electrode.

A Varian 3800 gas chromatograph equipped with electron capture (GC-ECD) was used to test the accumulating performance for POPs in water. Analysis was done as described [7].

A simple and inexpensive solution-blending method was used to dissolve CA and SL [16]. First, the oxygen within 1,4-dioxane was removed by sparging with nitrogen. Then different SL contents (SL versus CA) and other additives were dissolved in 25 ml of 1,4-dioxane at 50˚C. Subsequently, 5 g of CA was added and the mixture was kept at 80˚C for 2 days with stirring to ensure uniform mixing of the viscous syrups with different SL/CA ratios.

Using constant nitrogen pressure as the driving force, the prepared syrups were extruded directly into the precipitating liquid (deionized water) after a transitory stay in air. As the white filaments emerged, they precipitated from the deionized water and solidified.

Figures 1(a)-(d) show ESEM micrographs of the L-CA fibers with different SL/CA ratios in 1,4-dioxane. All fibers presented finger-like apertures, which were slightly enlarged with increased SL loading. Figure 1(e) shows an image of fibers with a 1:10 SL/CA ratio, from which it was calculated that the thickness of the dense surface layer is ~4 μm and the external surface was full of holes of ~0.1 μm diameter. 1,4-Dioxane/water as solvent/non-solvent can cause instantaneous delamination of the CA membrane, which is an important mechanism for forming finger-like apertures [17]. The presence of apertures is disadvantageous for fiber membranes. So, another solvent/non-solvent mixture, tetrahydrofuran (THF)/water, was investigated. The experimental results showed that a sponge-like structure was formed (Figure 1(f)).

In order to investigate the miscibility of SL and CA, the content of the elements C, O and P in different areas was measured by EDAX for the inner micro-areas and by XPS for the surface elements and the results are given in Table 1. The distribution of C and O is basically uniform in the center and in the sub-layer of the pure CA fiber. The presence of P can be due to impurity in the polymer or the solvent. The distribution of C and O and the C/O ratio at the surface is different from that in the inner areas. The content of P in the center of the L-CA fiber is very similar to that in the sub-layer, which indicates a uniform distribution of SL in the L-CA fiber. Analysis of the P content at the L-CA surface, in the CA surface and within the L-CA, revealed that the P content attributed to addition of SL is basically the same as that within L-CA; i.e. SL is distributed uniformly at the surface and within the L-CA fiber.

The possibility of an interaction between SL and CA was investigated by examining the FT-IR spectra (Figure 2). We compared the spectrum of native CA with that of L-CA and found that the stretching vibration absorption peak of C=O (CH₃CO) of CA shifts from 1754.94 cm⁻¹ to 1752.42 cm⁻¹, the peak of C-O (carboxylate) shifts from 1238.13 cm⁻¹ to 1234.38 cm⁻¹ and the peak of C=O (COO) shifts from 1635.95 cm⁻¹ to 1234.38 cm⁻¹. The peak positions of CH₂, CH, C-O (bridge) and C-O-C (pyranose ring) are not changed, indicating that interaction occurs in the ester position of CA. Comparing the spectrum of native SL to that of L-CA shows that the transmittance peak at 1459.56 cm⁻¹, which is assigned to the symmetric bending of -CH₃ at -N(CH₃)³⁺, disappears. This is attributed to the electrostatic abstraction between N of SL and C=O of CA. The transmittance peak at 1230.62 cm⁻¹ that is assigned to the stretching vibration of P=O shifts from 1230.62 cm⁻¹ to 1234.38 cm⁻¹. The transmittance peak at 1065.49 cm⁻¹, which is assigned to the stretching vibration of R-O-P-O-R', shifts from 1065.49 cm⁻¹ to 1051.36 cm⁻¹. These results indicate that negatively charged O connected to P interacts with the hydroxyl group of CA by hydrogen bonding.

Subsequently, XPS was measured support the identification of interaction sites by FT-IR analysis. The binding energies of C1s, O1s, P2p, N1s in CA, SL and L-CA fibers are given in Table 2. The C1s peak was resolved by using a least-squares peak-fitting program and the results revealed four chemical shifts: C connected to another C or H (C-C*H₂-C); C is connected to a non-carbonyl O (-C*H₂-O); C is connected to two non-carbonyl

O atoms (O-C*-O); C connected to a carbonyl O and to a non-carbonyl O (). The binding energy of is increased by 0.19 eV, whereas the binding energies of other species are not changed. The fitting for O1s revealed two chemical shifts; i.e. an O is connected to a C by a single bond (-C-O*-) and an O is connected to a C by a double bond and the binding energies are increased by 0.4 eV and 0.367 eV, respectively.

The binding energies of P2p and N1s are increased by 0.405 eV and 0.298 eV, respectively. According to the results of the analysis described above, SL and CA can effectively interact by electrostatic abstraction between the choline on SL and the carbonyl group on CA and by a hydrogen bond between the phosphoric group on SL and a hydroxyl group on CA. This confirmed that our hypothesis about the interaction is correct; SL can be immobilized stably on CA by this interaction.

Water is a complex matrix that can contain a large number of ionized substances, substances with redox capacity or microorganisms capable of degradating L-CA fiber. It is possible that these substances interact with L-CA fiber by ion transport or electron transport, accordingly affect the chemical structure of L-CA fiber. It is salutary to study the electrical conductivity and electron transport ability of composite fibers in order to understand the interaction between various matters in water and composite fibers. Figure 3(a) shows a CV of an aqueous solution containing 0.005 mol・L⁻¹ K₃[Fe(CN)₆] with 0.1 mol・L⁻¹ KCl as the supporting electrolyte, showing a chemically reversible voltammetric process:

Due to the redox couple. When the GCE surface was coated with a CA film, the Faradaic current response observed for the voltammetrically induced transformation decrease (Figure 3(b)). Nevertheless, the presence of the Faradaic current indicated that electron transfer reactions were still occurring between and the CA-coated GCE surface. The de-

crease in Faradaic and capacitive current between the bare electrode and the GCE coated with the CA is due to either the presence of numerous finger-like holes in the CA that allow the aqueous solution to reach the electrode surface (resulting in a diminished number of electron transfer reactions to occur) or electron tunneling from specific sites on the electrode surface that have a reduced coverage of the CA film.

When the GCE was coated with SL-incorporated CA film, the current observed during voltammetric scanning of the -containing solution was less than that observed with the CA film alone

(Figure 3(c)), indicating that the incorporation of SL into CA further reduces electron transfer between the GCE and the solution phase species. Although the distribution of SL in CA is a non-bilayer structure, its reduction of the current is similar to that of SL bilayers or multilayers [18]. The SL bilayer membrane is nearly completely insulative [19].

For a CA-coated GCE, we found that a chemically reversible voltammetric process took place at E_obs = 0.1207 V, DE = 0.4310 V due to the redox couple, indicating widening of the potential separation between the cathodic and anodic peaks, which could be due to the acidic nature of CA [20]. The negatively charged or electrostatically attracted to the acidic site of CA caused

an increase of activation overpotential that can force the charge transfer.

For an SL/CA-coated GCE, besides the chemically reversible voltammetric process at E_obs = 0.0669V, DE = 0.1754 V due to the redox couple, there were oxidation peaks at E_ox = 0.5176 V

and E_ox = 0.6272 V; the corresponding reduction peaks were at E_red = ?0.1856 V and E_red = 0.3208 V. The peak widening of the first redox potential is rather large. Keeping in mind the fact that the SL molecules are charged because they contain both a positive and a negative charge at the polar head [21], it is reasonable to assume that the positive side of the SL polar head is possibly the site involved in the chemical interactions with [Fe (CN)₆]^4-.

The second redox potential is more positive than that of on native GCE. Because

the SL can be oxidized in the presence of oxygen [22], it is possible that the presence of a current leads to a self-redox reaction of SL.

We used heptachlor, endrin, lindane, mirex, deldrin and perfluorooctane sulfonate (PFOS) as model pollutants of POPs and investigated the accumulation capacity of L-CA fibers with 0.1 g of finger-like and sponge-like structures using 200 ml of aqueous solution containing 10 μg of the target pollutant. As shown in Figure 4, the structure of the fiber affected the accumulation capacity and the sponge-like fiber appeared to have greater capacity than that of the finger-like fiber for all test pollutants. For heptachlor, endrin, lindane and deldrin, the accumulation equilibrium was achieved within 10 h with a high accumulation capacity. Accumulated equilibrium for PFOS was also achieved within 10 h but with low accumulation capacity, possibly due to the long-chain structure or to higher solubility in water than that for POPs mentioned above. For mirex, accumulation continued slowly during the whole accumulation period, showing an entirely different style of accumulation due to the large molecular mass and the high level of steric hindrance. This study showed that the kinetics of accumulation of POPs on the L-CA fiber is different for different types of POPs. Further study of the accumulation process will be needed if L-CA fiber is to be used for the removal of POPs during water treatment.