Optimization of PES-based Hollow fiber membranes incorporating MgO-modified activated carbon via response surface methodology for enhanced pure water permeability

Morphological and structural characterization of AC and AC-MgO particles

Figure 3 shows the XRD patterns of activated carbon, pure MgO, and AC-MgO particles. The XRD pattern of pure MgO particles exhibits sharp characteristic peaks at 2θ angles of 36.8°, 42.7°, 62°, 74.2°, and 78.3°, corresponding to the (111), (200), (220), (311), and (222) planes, respectively³⁹. The two broad peaks observed at around 22° and 44° in the XRD patterns of both pure AC and AC-MgO particles are attributed to the (100) and (101) diffraction planes of the carbon structure⁴⁰. By comparing the diffraction patterns, it is seen that the XRD patterns of pure AC and the cubic crystalline structure of MgO are superimposed in the XRD pattern of the AC-MgO nanocomposite, indicating that the synthesis was successful.

Figure 4a shows the SEM image of activated carbon particles used in this study, indicating particles with a clear, crack-free, and smooth surface. The SEM micrograph and the corresponding EDS analysis from the AC-MgO nanocomposite particle are presented in Figs. 4b–d. As could be seen from Fig. 4b, a significant change in the surface morphology of the AC particles occurred after their modification with MgO nanoparticles. Additionally, the SEM image at higher magnification (Fig. 4c) confirms the presence of MgO nanoparticles on the nanocomposite surface, exhibiting a small spherical morphology and partial overlap with each other. From the EDS analysis results presented in Fig. 4d, it can be observed that the Mg and O elements were uniformly distributed, indicating that the MgO nanoparticles were dispersed well on the activated carbon.

The microstructure of the AC-MgO nanocomposites was also revealed by TEM and selected area electron diffraction (SAED) images (Fig. 5a, b). The discontinuous ring pattern in the SAED image confirms the polycrystalline nature of MgO nanoparticles. The SAED image shows concentric rings corresponding to the (111), (200), (220), (311), and (222) reflections of MgO particles. HR-TEM imaging (Fig. 5c) reveals lattice fringes with spacings of 0.21 nm, which are attributed to (200). These results align well with the XRD data.

Surface analysis of the membranes

Characterizing the surface properties of ultrafiltration membranes is essential for improving their performance for separation applications. In this study, the effect of adding AC-MgO particles on the surface properties of PES membranes was analyzed in terms of surface roughness and contact angles. As illustrated by the AFM images in Fig. 6, the surface roughness of the composite membranes is higher compared to the pristine PES membrane, with R_a values of 6.94 ± 0.125 nm and 16.59 ± 0.487 nm for the pristine PES and PES-AC/MgO membranes, respectively. The addition of particles resulted in a rougher surface, which facilitated filtration by increasing both the effective filtration area and pore size²⁷. Figure 7 shows the contact angles of the pure PES membrane and the membrane containing 0.354 wt% AC-MgO. The results confirm an improvement in hydrophilicity after incorporation of AC-MgO particles into the membrane. This improvement resulted from the presence of polar functional groups (–OH and –COOH) on the surface of AC-MgO, which migrated toward the membrane surface during membrane formation and allowed greater interaction with water molecules^41,42. Furthermore, as evident from the AFM results, the membrane roughness increased with the addition of particles, which could make the hydrophilic surface even more water-attracting⁴³.

Response surface methodology

Results of ANOVA

Summary of ANOVA results of the RSM model for PWP of the developed membranes is shown in Table 3. An F-value of 1.13 for lack of fit indicates that it is not significant relative to the pure error. Given that at the 95% confidence level, variables with p-values less than 0.05 are generally considered statistically significant⁴⁴, the model is highly significant due to having a p-value less than 0.0001, while the lack-of-fit with a p-value of 0.47 is not significant. These results confirm that the model adequately fits the experimental data and is appropriate for predicting the response within the studied range of variables.

A predictive second-order polynomial model was established using multiple regression analysis within the RSM framework to quantify the effects of four key factors on the pure water permeability, as expressed by the following equation:

To further evaluate the model’s accuracy, the predicted values are plotted against the experimental data in Fig. 8. As shown in this figure, the data points lie very close to the diagonal line, indicating strong consistency and close agreement between the observed and the predicted values. This alignment indicates the adequacy of the model fit. Additionally, the fit is confirmed through the coefficient of determination ((:{R}^{2})), where a high (:{R}^{2}) value of 0.94 demonstrates a strong correlation between the predicted and observed values. Furthermore, the minimal difference (less than 0.2) between the predicted (:{R}^{2}) ((:{R}_{pre}^{2})) and the adjusted (:{R}^{2}) ((:{R}_{adj}^{2})) confirms the validity of the model. Based on these statistically significant results, it can be confidently concluded that this model is effective in predicting optimal operating conditions.

Figure 9a presents the normal probability plot of residuals for the PWP responses. Residuals represent the difference between the measured and the predicted values. From this figure, the distribution of points closely following a straight line indicates that the residuals are approximately normally distributed. Figure 9b displays the residuals plotted versus the predicted PWP values. In this figure, the studentized residues are randomly scattered, indicating that the residuals are independent of the predicted response. The absence of any discernible pattern further supports the assumption of residual independence. Additionally, the slight fluctuation of residuals around the x-axis confirms the assumption of constant variance. These results collectively demonstrate the predictive accuracy of the model.

From the above results, the model predictions for water permeability are reasonably accurate and reliable. The optimal processing parameters predicted by the model are: AC-MgO concentration of 0.354 wt%, air gap distance of 18.933 cm, dope solution flow rate of 2.981 ml/min, and bore fluid flow rate of 7.775 ml/min. The maximum theoretical water permeability of the hollow fiber membrane under the optimal processing conditions was estimated to be approximately 56.715 L/(m².h.bar), and the experimentally measured permeability at this condition was found to be 52.49 L/(m².h.bar). This result suggests that the employed model has good predictive capability of the variables for achieving the optimal response.

Impact of particle concentration and process parameters on PWP

Figures 10 and 11a, d, and f illustrate the correlation between the concentration of AC-MgO particle in dope solution and the pure water permeability. As the content of AC-MgO increases, the PWP of membranes increases; however, a slight decrease is observed at low concentrations of particles. From these results, the higher concentration of particles in the fluid has a more pronounced positive effect on PWP. Such improvement is mainly ascribed to the synergistic effects of increased porosity and hydrophilicity due to the added particles⁴⁵. The relationship between PWP and air gap distance, as shown in Figs. 10 and 11a, c, and e, reveals that permeability increases with the rising air gap, which is attributed to the formation of longer finger-like structures⁴⁶. Figure 10 and 11b, c, and d also show that PWP increases with increasing dope solution flow rate, which is attributed to the significant formation of large macrovoids⁴⁷. Additionally, as depicted in Fig. 10 and 11b, e, and f, the PWP increases by increasing the bore fluid flow rate, likely due to the greater thickness of the layer with the finger-like structures⁴⁸. Further details will be described in the following sections.

Microstructure analysis of manufactured HFM cross section

To investigate the effect of the studied parameters on HFM morphology, SEM micrographs were taken from the transverse cross-section of the manufactured membranes. These images reveal that the cross-section of the membranes generally consisted of three distinct layers, with pore morphology varying depending on the process parameters, as discussed below.

Effect of particle concentration

The SEM images of the manufactured HFM cross-section with various AC-MgO particle content are shown in Fig. 12. As observed from this figure, the membrane cross-section typically comprised three distinct layers: an outer layer, which is generally dense with minimal porosity; a middle layer characterized by a macro-porous structure that influences permeability⁴⁹; and an inner layer composed of micro- and meso-porous structures. Additionally, Fig. 12 illustrates that increasing the AC-MgO content led to the formation of larger finger-like pores. These particles acted as nucleation sites for pore formation by providing surfaces where the non-solvent preferentially interacted with the dope solution. The improved hydrophilicity of the dope solution, resulting from the presence of AC-MgO, accelerated the solvent–water exchange rate and facilitated the development of larger pores and an increase in the thickness of the finger-like structured layer^12,50. These morphological changes, along with increased hydrophilicity due to the presence of AC-MgO particles, which allowed water molecules to more easily wet the membrane surface and penetrate the pores⁵¹, significantly enhanced the PWP of the membrane. These findings are in excellent agreement with the data presented in Figs. 10 and 11.

Effect of air gap

The effect of air gap distance on the thickness of the layer with the finger-like structure in the HFM, are presented in Fig. 13. It is shown that by increasing the distance from 2 cm to 12 cm and 42 cm, while keeping other parameters constant, the size of this layer in samples S8, S2, and S1 increased from 30 μm to 43 μm and 60 μm, respectively (Fig. 13a–c). Similarly, the thickness of the finger-like structured layer changed from 38 μm to 60 μm for the samples S15 and S18, respectively (Fig. 13d, e). Moreover, increasing the air gap distance caused greater pore diameters, as the fibers were exposed to air for a longer time before being submerged in the coagulation bath, allowing water vapor to induce phase inversion more extensively^52,53. According to the SEM micrograph presented in Fig. 13, the improved PWP of the composite hollow-fiber membranes can be ascribed to the combined effects of longer finger-like structures in the middle layer, and a shorter dense structure on the outer layer compared to membranes fabricated with a shorter air gap⁶. All these factors contributed to the increase in water permeability through the membrane, indicating a direct relationship between PWP and air gap distance, as is also evident in Figs. 10 and 11.

Effect of dope solution flow rate

It has been demonstrated that the flow rate of the dope solution during the fabrication of hollow fiber membranes significantly influences their structural morphology. In this study, four different flow rates were used to investigate the effects of dope solution flow rate on the PWP and morphology of hollow fiber membranes. As can be seen, increasing the flow rate from 2 to 3 ml/min for samples S28 and S27, resulted in an increase in the outer diameter of hollow fibers from 473 to 534 μm, and a significant increase in their wall thickness from 34 μm to 90 μm (Fig. 14a, b). Similarly, increasing the dope flow rate from 1.5 ml/min to 2.5 ml/min for samples S12 and S2 resulted in the outer diameter increasing from 539 to 609 μm and the wall thickness rising by 54% from 52 to 80 μm, respectively (Fig. 14c, d). These findings indicate that increasing the polymer dope solution flow rate mainly affects the outer dimension of the hollow fiber, leading to a notable thickening of the fiber wall. These observations are consistent with previous studies on PES hollow fiber membranes⁴⁷, which reported that higher flow rates produce fibers with thicker walls. Moreover, the outer surface of membranes fabricated with higher dope solution flow rates exhibited the largest pore sizes. This is attributed to the shorter residence time, which limited solvent evaporation into the air. Consequently, phase inversion of the membrane’s outer surface occurred entirely within the external coagulation bath. In this environment, rapid demixing between the solvent and water led to the formation of larger surface pores compared to membranes exposed to longer evaporation-induced phase inversion⁵⁴. As illustrated in Fig. 14, an increase in the outer diameter of the membrane is also associated with a greater wall thickness and the development of larger finger-like pores. These structural changes contribute to an enhancement in PWP.

Effect of bore fluid flow rate

The effect of flow rate of bore fluid on fiber morphology, while other process parameters were held constant, is also illustrated in Fig. 15. For samples S19 and S20, increasing the flow rate from 5 ml/min to 8 ml/min increased the thickness of the middle layer with the finger-like structure from 30 μm to 35 μm and the inner diameter from 344 to 358 μm, respectively (Fig. 15a, b). Similarly, for samples S2 and S10, increasing the flow rate from 6.5 ml/min to 9.5 ml/min increased the inner diameter from 458 μm to 470 μm and the thickness of the middle layer from 43 μm to 56 μm, respectively (Fig. 15c, d). These data indicate that increasing the bore fluid flow rate through the spinneret led to elongated voids and produced a thinner fiber wall, resulting in enhanced PWP. This observation is attributed to increased internal pressure caused by a higher bore fluid flow rate on the inner surface of the hollow fiber during spinning, which in turn leads to a reduction in membrane wall thickness⁵⁵. Additionally, at high flow rates, the bore fluid rapidly diffused into the doper solution. Consequently, phase inversion occurred faster throughout the dope solution relative to polymer migration towards the bore fluid. This led to the formation of a more homogeneous void distribution⁵⁶.

As explained above, the improved PWP of the developed HFMs can be attributed to their enhanced hydrophilicity and the presence of large finger-like pore structures in the middle layer. Meanwhile, the inner layer, characterized by its small pore size, is primarily responsible for the selectivity and separation performance of the membrane. For potential application in hemodialysis, this membrane design allows small solutes such as urea to pass freely, enabling effective removal of uremic toxins. At the same time, larger molecules, including proteins (e.g., albumin) and blood cells, are efficiently retained by the inner layer, thus preventing their leakage into the dialysate solution ^57,58. Therefore, despite increased permeability, selective filtration is maintained.