Evaluation of process condition impact on copper and lead ions removal from water using goethite incorporated nanocomposite ultrafiltration adsorptive membranes

In the previous few decades, a remarkable quantity of hazardous wastes has been imposed on the human and wild environment as a result of urbanization and industrialization. Especially, the fast increasing heavy metal ions in rivers, surficial streams and wastewaters must be considered as a major life threatening preoccupation (Cao et al. 2010). Unlike organic pollutants, metal ions are not degradable and produce a hazard of accumulation in humans’ organs. Moreover, they are potential risks for the environment, plants, animals, and marine life, especially at higher dosages (Tofighy & Mohammadi 2011). All aforementioned concerns, along with the increasing shortage of freshwaters and competitive use of existing water resources, have made heavy metal ion removal from water more important than it was before. Chemical precipitation (Charerntanyarak 1999), ion exchange (Lalmi et al. 2018), adsorption (Faur-Brasquet et al. 2002), coagulation (Charerntanyarak 1999) and flocculation (Hankins et al. 2006), flotation (Deliyanni et al. 2017), electrochemical treatment (Tran et al. 2017) and membrane filtration (Borbély & Nagy 2009) are different methods developed to remove metallic ions from effluents, among which membrane technology is one of the most frequently studied methods according to literature surveys (Fu & Wang 2011). Therefore, membrane-based ion removal techniques experienced a highly noticeable growth toward resolving or diminishing problems associated with metal ions present in water streams and wastewaters. Numerous metal ions were targeted for these endeavours, such as copper (Cu²⁺) (Ghaemi 2016; Hossaini Zahed et al. 2019), mercury (Hg²⁺), zinc (Zn²⁺) (Borbély & Nagy 2009), lead (Pb²⁺) (Rahimi et al. 2015), cadmium (Cd²⁺) (Choi & Kim 2003), nickel (Ni²⁺) (Ho et al. 2014), and cobalt (Co²⁺) (Verbeken et al. 2009) ions, which are detected in industrial effluents.

Nanocomposite membranes are a new generation of membranes that are fabricated by incorporating nanomaterials with polymeric membrane materials. Incorporating nanomaterials affects membrane performance in two different ways. First, it alters polymeric membrane characteristics such as hydrophilicity, porosity, electric charge density and mechanical stability, and second, it introduces unique functionalities such as antibacterial, photocatalytic or adsorptive capabilities (Yin & Deng 2015). Usually, nanomaterials are included to overcome the demerits of the typical membranes, such as passing over the trade-off between rejection and water permeability, avoiding membrane fouling, and increasing membrane thermal and chemical stability. In case the added nanomaterials would be hydrophilic, segregation of the added nanoparticles from the bulk of the membrane to the surface of the membrane increased the hydrophilicity and increased the surface pore size, porosity and water permeability due to the faster phase separation rate (Mulder 1997). In a high concentration of nanomaterials, dope solution viscosity increases, which delays mixing and decreases the pore size, porosity, and water permeability of the resultant membrane.

In Hossaini Zahed et al. (2019), goethite nanoparticles (GNPs) were incorporated into a polyacrylonitrile (PAN)-based UF membrane, and the fabricated membranes were investigated to remove Cu(II) ion from synthesized wastewater. An effective membrane-based ion removal efficiency can be obtained by optimizing the membrane fabrication condition and the process parameter optimization, feed pH, initial concentration, and transmembrane pressure (TMP) to remove the targeted pollutant from wastewater. In this paper, only the effects of the membrane fabrication conditions were investigated, and the impacts of the process conditions were not yet evaluated.

Feed pH is known as an effective parameter in many separation processes. Song and Htun (Oo & Song 2009) found that feed pH plays a crucial role in boron ion removal in a reverse osmosis membrane. They concluded that higher feed pH values result in greater boron ion removal. This is due to the dissociation of boric acid into borate form, and there exists a significant repulsion force between the membrane surface and borate charges at higher pH values. In another study, the effect of feed solution pH was investigated in some membrane-based processes such as separation of albumin from immunoglobulins (Saksena & Zydney 1994) and separation of oil via ultrafiltration (UF) of oil/water/nonionic surfactant emulsion (Fazullin & Mavrin 2017). To the best of our knowledge, no study has yet focused on the effect of feed pH on ion removal efficiency in effluent treatment using adsorptive nanocomposite membranes.

Different effluents contain different amounts of metal ion contaminants according to different processes. Thus, initial feed metal ion concentration is another important factor. However, different researchers have studied to find the optimum concentration at which the highest ion removal is achieved. For example, Ahmad et al. studied the effect of contaminant copper and lead ion concentration on the efficiency of the removal process (Sani et al. 2017). Omri and Benzina also conducted a study to evaluate the effect of feed contaminant concentration on removal efficiency and found that at a fixed pH, the lower ion concentration typically results in better removal efficiency (Omri & Benzina 2012). However, to the best of our knowledge, no experimental evaluation has been performed at a variety of pH ranges.

TMP is a crucial factor affecting the separation performance of the membrane-based separation for the removal of organic and inorganic species. Mnif et al. (2015) studied the effect of TMP on phenol removal from water in reverse osmosis filtration and concluded that higher TMP leads to higher phenol removal. Arsenic semi-metallic ion removal from wastewaters was studied at several TMPs using four commercial nanofiltration (NF) membranes by Zhao et al. (Yu et al. 2013). Similarly, they concluded that more ions are removed as a result of higher TMP. However, few studies have focused on the effect of TMP on ion removal performance of low-pressure filtration like UF. Although Huang et al. (2012) assessed the effect of TMP on membrane fouling in separating Cd²⁺ ion from contaminated synthetic wastewater, the authors mainly focused on the membrane fouling, and decrease in membrane water permeability and the effect of TMP on the ion removal efficiency were not evaluated (Huang et al. 2012).

In the previous published paper (Hossaini Zahed et al. 2019), only the effect of membrane fabrication condition on the divalent metal ion was evaluated, and the impact of process conditions was not assessed. Although some studies have performed a preliminary evaluation of the process conditions on the metal ion removal from the water, these phenomena were not evaluated well. Therefore, this study focused on the effect of the change in feed solution and process operating conditions on removal efficiency. In other words, for a UF nanocomposite membrane embedded with GNPs, feed initial ion concentration and pH were studied, and their mutual effects were considered for the first time. Moreover, operational optimum TMP was chosen based on the preliminary screening tests. To the best of our knowledge, the effect of the process conditions on the nanocomposite UF membrane has been rarely studied. Thus, the effect of process parameters was focused on in this study.

Iron nitrate nonahydrate ((FeNO₃)₃.9H₂O, 99% purity), as a precursor for nanoparticles synthesis, cupric nitrate (Cu(NO₃)₂, 98% purity), hydrochloric acid (HCl, 37% purity), sodium and potassium hydroxide (NaOH, >98% purity, and KOH, >85% purity), poly(vinylpyrrolidone) (PVP, 10 kg mol⁻¹), as pore former, and dimethylacetamide (DMAc, 99% purity), as a solvent, were purchased from Merck™. Lead nitrate (Pb(NO₃)₂, 99% purity) was supplied from Fluka™. Polyacrylonitrile (PAN, 350 kg mol⁻¹), as membrane polymer matrix, and all other chemicals were purchased from domestic chemical suppliers. All the polymers and the synthesised GNPs were dried (1 h at 70 °C) in a vacuum oven prior to being used.

Goethite, as a kind of ferrous hydroxide (α-FeO(OH)) with the hydroxyl-covered surface, was added to the polymeric matrix to improve its hydrophilicity and consequently the obtained membrane performance in water-based separations (Rahimi et al. 2015). The synthesis of goethite briefly involves hydrolysis of the ferric salts in their aqueous solutions, refinement, and finally, separation of the products, as first introduced in 2003 by Bakoyannakis et al. (2003) as a simple and cost-effective method. In order to synthesize GNPs, some minor changes were applied accordingly. Initially, 4.85 g of Fe(NO₃)₃.9H₂O was dissolved in 50 mL of distilled deionized water, and the solution was then precipitated by the gradual addition of 4.8 M KOH solution while it was subjected to highly fast stirring conditions until the solution basic pH of 12 was attained. Then the suspension was sonicated with the power of 305 W at ambient temperature for 30 min and then placed in an oven for 70 min at 100 °C before centrifugation for 10 min at 10,000 rpm. Finally, the achieved solid nanoparticles were washed with acetone and distilled deionized water, and last of all, dried at ambient temperature (Rahimi et al. 2015).

This study selected the phase separation method for membrane fabrication due to its versatility and simplicity. Hence, GNPs were first dispersed in sufficient DMAc utilizing the maximum power of an ultrasound probe for 15 min. PAN and PVP were gradually added to the suspension within 6 h, at 50 °C while it was stirring at 200 rpm. PAN and PVP powders were already kept in an oven at 70 °C overnight in order to remove any potential adsorbed humidity. After all, the mixture was kept stirred for 30 h at 100 rpm. At this point, the mixture of PAN/PVP/GNPs/DMAc was completely homogeneous and set for casting. A casting knife (thickness = 200 μm) was then used to cast the mixture at ambient temperature, and then it was submerged in distilled water for a whole day prior to being tested as the membrane. Membrane composition in this study was 15 wt. % of PAN, 1.3 wt. % of PVP, 0.1 wt. % of GNPs, and the rest up to 100% was DMAc (83.6 wt. %). More detailed information about membrane preparation is available in our previously published paper (Hossaini Zahed et al. 2019). Based on the previous study (Hossaini Zahed et al. 2019), this is the composition at which UF nanocomposite adsorptive membrane showed the optimum water flux of 258 liters per square meter per hour (LMH)/bar and Cu(II) removal of 49.1%. Membrane without nanoparticle inclusion was also prepared to assess the effect of nanoparticle incorporation.

A scanning electron microscope (SEM) was employed to study the morphology of the membranes prepared in this study. Cross-sectional specimens were made by immediately breaking the membrane samples right a few seconds after their immersion in liquid nitrogen. Then the specimens were fixed on fixtures and gold-coated to the thickness of 10 nm, using a C7620 mini Sputter Coater ™ (United Kingdom) prior to being evaluated using a TESCAN™ SEM VEGA.

Transmission electron microscopy (TEM) images were recorded from ultrathin specimens with a thickness of 80 nm, using a M/s Leica diamond knife (Germany) equipped Leica EM UC7 ultra-microtome. TEM images were recorded at an accelerating voltage of 100 kV on a Zeiss-EM10C instrument (Germany).

According to some preliminary screening tests, the feed solution concentration and pH and the operating TMP were recognized as the most important affecting parameters. Accordingly, the feed solution pH (ranged from 2 to 5), Cu(II) concentration (C₀) (ranged from 20 to 200 ppm), and TMP (varied from 1 to 3 bar) were selected as the affecting independent parameters according to the results of the preliminary test and then some new experiments were designed for further evaluation using experimental design method. Table 1 lists the selected affective parameters/levels chosen for the response surface methodology (RSM). Membrane Cu(II) removal results were analysed by means of RSM – Box-Behnken design (BBD) mode (Hossaini Zahed et al. 2019; Peng et al. 2020; Singh & Bhateria 2020).

The membrane feed solution was formed by the addition of predetermined amounts of cupric nitrate (Cu(NO₃)₂) to a sufficient amount of distilled water at room temperature under the medium speed of agitation. The feed solution pH with predetermined Cu(II) concentration was adjusted by dropwise addition of 0.1 M HCl solution. A sufficient amount of distilled water was then added to the solution to make it meet the designed concentration.

Removal of ions was assessed by means of a dead-end membrane setup using membrane samples with an operative surface of 10.2 cm². The cast membranes were located in the cell and subjected to pre-compaction via running pure water for 10 min at a pressure of 3 bar prior to being tested during the real test to remove the compaction effect. Figure 1 depicts a schematic representation of the membrane setup used in this study.

Similarly to the results obtained in the previous study (Hossaini Zahed et al. 2019), the formation of the GNP clusters and aggregates is completely obvious from the SEM image shown in Figure 2, originating from the magnetic properties of the GNP nanoparticles. Henceforth, it was somewhat challenging to record any SEM image wherein GNP nanoparticles can be observed separately. Nonetheless, their particle size was observed up to 60 nm and could be identified confidently in the nanometric range. Their surface area, average pore volume, and pore size were 200.8 m² g⁻¹, 0.40 m³ g⁻¹, and 10.6 nm, respectively, using the nitrogen adsorption-desorption method (Micromeritics ASAP 2020, USA) (Hossaini Zahed et al. 2019).

Figure 3 shows SEM images of the surface and the cross-sectional views of the control and the 0.1 wt. % GNPs loaded nanocomposite membranes. In Figure 3(a), the surface image of the control membrane demonstrates a smooth surface, as usual for control polymeric membranes. However, as shown in Figure 3(b), the observed GNPs on the nanocomposite membrane surface give it a somewhat grainy look. Knowing the hydrophilic property of the GNPs due to their surface hydroxyl groups, they have the tendency to migrate to the membrane upper surface during phase separation (Chatterjee & De 2014). Although not shown here, in the previous study (Hossaini Zahed et al. 2019), the water contact angle of the membrane surface decreased from 67 to 46° by addition of just 0.1 wt. % GNPs, which proves the hypothesis of the GNPs' migration to the membrane surface. The cross-sectional images of both control and nanocomposite membranes show the asymmetric structure, i.e., the porous surface on the top and finger-like structure underneath. This morphology directly results from quick phase separation on the top and rather gentle underneath, as discussed by Bakeri et al. in detail (Delavar et al. 2017).

Figure 3(c) displays a TEM graph recorded from the cast nanocomposite membrane (0.1 wt.% GNPs). As observed, distinct nano-rods attributed to GNPs were found in the microtomic specimen, demonstrating a proper dispersion of GNPs in the membrane.

Although not shown here, similarly to the previous published paper (Hossaini Zahed et al. 2019), Fourier transform infrared (FTIR) revealed proper incorporation of GNPs in the PAN polymer matrix by considering the surface hydroxyl groups in GNPs. The FTIR analysis results can be found in detail in the previous study (Hossaini Zahed et al. 2019).

Table 3 tabulates the variance results for the impacts of pH, C₀ and TMP as the independent factors and their interactions on Cu(II) ion removal.

The R² value for the fitted equation equals 0.9935, from which it can be found that the proposed model predicts the experimental data accurately. Table 3 also demonstrates that the pH factor's effect is less significant than the C₀ and TMP factors.

Actual experimental data for Cu(II) ion removal versus those predicted by the suggested model are revealed in Figure 4. From Figure 4, the model predicts the actual results well by the model, since the line drawn between the two sets of real and predicted data displays the angle of 45°.

Figure 5 demonstrates the effects of pH, C₀ and TMP on ion removal performance of the prepared membranes. The predicted results of the proposed model are presented in solid black lines, whereas the discrete blue lines show their confidence interval (CI, by 95%). The 95% CI is an interval estimate obtained from the statistical evaluation of the actual experimental data that might estimate the real values of indefinite factors.

According to Figure 5(a), the ion removal trend increases slowly with the increase of pH from 2 to 5. In other words, Cu(II) ion removal decreases at acidic pH. Lower feed pH value eventuates at higher H⁺ concentration; accordingly, this phenomenon accumulates protons on the membrane surface, which minimizes Cu(II) ion adsorptive removal by the membrane incorporated GNPs (Delavar et al. 2017). However, at pH values higher than 5, this trend becomes flat because of the Cu(OH)₂ formation, which significantly reduces Cu(II) ion removal (Sani et al. 2017).

Figure 5(b) shows that at ion concentration roughly below 140 ppm, ion removal increases as ion concentration decreases. In other words, the membrane shows the highest ion removal capacity at the lowest ion concentration. This is because GNPs' active adsorption sites can easily be adsorbed for Cu(II) ion adsorption in lower Cu(II) concentration conditions. Similar findings were published by Delavar et al. (2017). Beyond ion concentration of 140 ppm, a slight increase in the ion removal was observed; the reason for this phenomenon is not clear. Basically, a plateau graph is expected where more increase in C₀ results in no change in ion removal as observed. That means saturation occurs, and all active absorbent sites are occupied with metal ions (Hafez 2012; Omri & Benzina 2012). Although more detailed studies are needed, and the clear and strong reason for this phenomenon is not available now, it was hypothesized that the slightly increasing trend in ion removal for ion concentration higher than 140 ppm may be related to the experimental error that alters the fitted equation.

Figure 5(c) shows that at TMPs up to 2.25 bar, ion removal decreases sharply. Higher TMP causes higher flux, which minimizes the probability of the copper divalent ion being taken by the membrane material. Therefore, its ion removal decreases. Beyond 2.25 bar, TMP hardly affects the Cu(II) ion removal. It is expected that by increasing the TMP the ion removal decreases in all the TMP range. This phenomenon was described as follows: Increasing the feed solution TMP increases the kinetic energy of the feed solution molecules, which can go through the pathway inside the membrane with a small pore size, which was not possible in low TMP. In this case, some part of the solution is treated with the NPs that were not in contact with any solution at low TMP, and fresh NPs inside the tiny pores well purify this part of the solution. Thus, the decreasing trend of the ion removal was damped by the above-mentioned phenomenon, and the removal plateaued above 2.25 bar.

Figure 6 depicts 3D graphs of the selected dual parameters' interaction, in which two parameters are changed within their predetermined ranges selected in the BBD experimental design method, and the third one is kept constant at its middle value.

It can be seen in Figure 6(a) that the lowest membrane ion removal is reached when both TMP and C₀ gain their maximum levels; in other words, a simultaneous increase in TMP and C₀ reduces the membrane ion removal. Figure 6(b) shows that at any arbitrary C₀ in the selected range, the higher feed pH causes the higher ion removal, and at any arbitrary feed pH, ion removal is reduced as C₀ increases from 20 to 200 ppm. Figure 6(c) also shows that lower TMP results in higher ion removal regardless of feed pH; however, the minimum removal value is found at TMP of 2.25 bar and higher, as discussed before.

As expected, the optimum conditions were selected by using the proposed RSM model to obtain the maximum of Cu(II) ion removal. Their values in this regard are 4.17 for pH, 20.291 ppm for C₀, and 1.009 bar for TMP, in which Cu(II) removal is 45.0%. A target pH (4.2), C₀ (20 ppm), and TMP (1 bar) were finally set regarding these proposed conditions in the experiment. The actual membrane performance was compared with that predicted by the proposed model at the same conditions in the experiments. The experimental results revealed 49.1% for removal of Cu(II) ion, which reveals an error of less than 9%. This shows a worthy agreement between the theoretically predicted value(s) and the experimentally measured actual result(s).

At the end, for the optimum condition of C₀ (20 ppm), pH (4.2) and TMP (1 bar) achieved from the Cu(II) ion removal optimization, the lead (Pb(II)) ion removal from synthetic wastewater was experimentally done. The result showed that 43% of the initial lead ions were removed, which is close to Cu(II) ions in the same filtration experiment conditions. The observed difference in the removal percentage may be due to the different ions of Cu(II) and Pb(II), as the former has a smaller diameter than the latter one. In other words, the larger ion occupies more space on the GNPs' adsorbent surface; hence, a lower number of ions are adsorbed, and a lower removal performance capacity is achievable. Besides the size of the ion, another effective parameter is hydration diameter, when ions are in aqueous solution and form aqua complexes in which the maximum number of water molecules are clustered around them. According to Persson (2010), both Cu(II) and Pb(II) form octahedrons with six water molecules forming the clusters around the metallic ion, with the larger hydrated ion radius for Pb(II) compared to Cu(II). Thus, the hydration diameter of Pb(II) is larger than that of Cu(II).

Previously prepared nanocomposite membranes based on PAN (15 wt. %)/PVP (1.3 wt. %) and the synthesized goethite nanoparticles (GNPs by 0.1 wt. %) via the phase separation method were used in the current study. The effects of pH, Cu(II) ion concentration, and transmembrane pressure on the Cu(II) ion removal were investigated. The obtained optimal conditions using BBD based on RSM was used to decide optimum conditions. Those at which the highest copper (Cu(II)) ion removal was achieved were 20 ppm for the feed ions concentration, 4.1 as the feed pH, and the operating TMP of 1 bar. According to the Box-Behnken experimental design, the predicted value for the membrane removal performance was 45%, which was reasonably close to 49.1%, the value obtained from the experiment. In addition, removal of lead (Pb(II)) metal ions was evaluated by the same membrane with the aforementioned initial feed conditions and operational TMP, and 43.0% ion removal was observed. Relatively low removal percent for Pb(II) ion can be attributed to the fact that the size-exclusion mechanism was not working due to the large pore size of the utilized ultrafiltration membrane, and adsorptive removal mechanism was employed mainly due to the incorporated GNPs' active site. Nasir et al. (2019) prepared PAN nanocomposite membranes incorporated with carbon nanotube and aluminium fumarate as adsorbent. Although they reached more than 90% Pb(II) removal, water permeability of their membrane was less than 8 LMH/bar (much less than that obtained in this research), which means their prepared membrane is located in the NF range. In another study, Pakdel Mojdehi et al. (2019), development of a polyethersulfone (PES) membrane modified by polyaniline TiO₂ adsorptive membrane was used for Cu(II) removal. In this study again, although near 90% Cu(II) ion removal was reported, water permeability was around 50 LMH/bar, much lower than that obtained for the membrane prepared in this study.

The authors of this paper would like to acknowledge Iran National Science Foundation (INSF) for the financial support dedicated to the present research (Grant # 96008182).

All relevant data are included in the paper or its Supplementary Information.