1. Introduction

Membrane separation has a number of advantages in water treatment processes including the reduction of cost. Membrane separation has been reported to reduce up to 90% of separation cost compared to conventional processes. Thus, membrane separation is investigated for its applications in various separation processes including textile wastewater treatment.

To fabricate ceramic nanofiltration membranes, atomic layer deposition (ALD) has been employed[1-5]. ALD is a process in which thin films are deposited onto substrate through gaseous precursor exposure. ALD can be used to modify separation performances in ceramic membranes, tightening microfiltration (MF) membranes to ultrafiltration (UF) membranes. Further tightening to NF membranes was achieved by a modified ALD technique with tubular ceramic UF membranes.

Fouling causes permeability in NF membranes to decrease dramatically, which often occurs due to a low permeable gel layer forming on the top of the membrane[ 6-9]. Conventional anti-fouling methods such as forward flush and backwash is not effective in recovering the permeability. Therefore, membrane modifications have been proposed to remediate this issue, such as modifying with Fenton catalysts or the use of Ca²⁺ ions.

Ceramic NF membranes are usually fabricated with metal oxides including zirconia, titania, or alumina, which leads to high flux of water due to their hydrophilic nature[10-13]. However, due to the hydrophilic nature, in non-polar solvents, solvent flux shows a declining trend. Therefore, membrane modification such as membrane hydrophobization has been proposed. Fig. 1 represent the classification of the review and Table 1 summarize the membrane efficiency.

2. Nanofiltration Membrane

Nanofiltration process based on ceramic membrane are more robust and less prone to be affected by chemical treatment as compared to polymeric membrane. Polyamide thin film membranes were prepared with ceramic tubular membranes having pore size of 100 and 200 mm as substrates through interfacial polymerization[14]. Various monomers were investigated for their effects on the membranes. For this study, two substrates labelled S-100 and S-200 (with mean pore size of 127 and 197 nm) were used and the two polyamide layers fabricated on the substrate was labelled PA-100 and PA-200. In comparing aqueous monomers, polyethyleneimine (PEI) and piperazine (PIP) based polyamide membranes showed smoother surface morphology than m-phenylene diamine (MPD) based polyamide membranes. The solvent, cyclohexane, was used to wash off excess aqueous solution off the ceramic substrate, which helped form a polyamide thin film which was continuous. The elastic moduli of the membranes were higher than commercial polyamide nylon, with PA-100 membrane showing 5.11 GPa and PA-200 showing 3.95 GPa. Contrary to these results, the elastic moduli of the polyamide surface were higher in PA-100 membrane, at 0.90 ± 0.5 GPa, while PA-200 showed 0.55 ± 0.5 GPa. As pressure increased from 2 bar to 10 bar then declined to 2 bar, pure water permeability (PWP) linearly increased. The PWP of the substrates were 589 ± 23 and 1069 ± 27 LMH bar^-1 for S-100 and S-200 respectively, while PA-100 and PA-200 membranes showed 16.7 ± 0.8 and 18.4 ± 0.7 LMH bar^-1. The molecular weight cut-off (MWCO) was lower compared to reported PEI-based polyamide membranes in other works due to PIP being used as the aqueous monomer, with PA-100 and PA-200 membranes showing 240-260 Da and reported membranes having 350-500 Da MWCO. At high temperatures above 50°C, PEG-300 and MgSO₄ rejection decreased while sucrose and MgCl₂ rejection remained high.

A layer of MXene was loaded onto Al₂O₃ support layer then modified to be hydrophobic. MXene was loaded in different amounts (0, 0.3, 0.5, and 0.7 mL) and hydrophobic MXene modified ceramic membranes were fabricated through hydrophobization using hexadecyltrimethoxysilane( HDTMS)[15].

The modified membrane shoed a water contact angle of around 144.7°. The α-Al₂O₃ support membranes had high porosity and surface hydrophilicity, which resulted in large water droplet size which was highly polydispersed. The modified membranes have higher fractions of small droplets (~400 nm) compared to unmodified membranes, due to the MXene modified layer and homogenous 2D channels.

To investigate fouling in ceramic nanofiltration (NF) membranes in the application of high municipal sewage, several fouling control protocols were tested[16]. Four large membranes with 450 Da molecular MWCO were used to assess hydraulic backwash and forward flush. In the absence of hydraulic backwash, membrane permeability showed a decline from 22 to 4.5 L m² h^-1 bar^-1 after 2h filtration while the presence of hydraulic backwash at backwash pressure of 6 bar increased permeability by 43, 37, 31, and 25% respectively. Due to the ceramic NF membranes being damaged after this test, a second trial with lower pressure (1 and 2 bar) was conducted. However, the permeability was not enhanced with hydraulic backwash at low pressures. Cross-flow velocities (0.4 to 4.3 m s^-1) did not significantly affect fouling removal via forward flush, in contrast to polymeric UF membranes, which show an increase in fouling removal following an increase in cross-flow velocities. Two reactions, Fenton (like) reactions and calcium carbonate reaction with acid were tested for precoated membranes. Both trials showed a similar fouling curve in comparing an untreated membrane and a precoated membrane. Uncoated membranes showed an approximately 8% increase in permeability after one cycle of cleaning via hydrogen peroxide, which was significantly enhanced by precoating of the membranes, at 75% increase in permeability. The optimal recovery and efficiency were seen in pH 7 feed solutions and 420 mg L^-1 FeCl₃ concentration. The performance of precoated membranes declined with two trials, due to the unevenly distributed precoat layer on the membrane surface. Uncoated membranes treated with HCl and citric acid showed a 11 and 27% increase in permeability respectively. Precoated membranes showed similar results for HCl cleaning, at 12% permeability increase but showed a dramatic spike for citric acid cleaning, at 76% permeability increase. To compare the three methods, net water production was calculated. A high net water production was achieved when hydraulic flux remained at a high value during filtration and the production downtime was relatively short. The three methods resulted in a similar net water production curve compared to the absence of a fouling method, which had 51 L m^-2 net water production after 130 min including downtime. Calcium carbonate reaction based precoat had the highest net water production at 80 L m^-2 but required a relatively longer production downtime (16.1 min h^-1).

2.1. TiO₂

Sol-gel process is one of the most exciting methods to prepare ceramic NF membranes with very small porosity. But recently atomic layer deposition method is getting more popular. Molecular layer deposition (MLD), a modified atomic layer deposition (ALD), was performed on tubular ceramic ultrafiltration (UF) membranes as the substrate to deposit titanium alkoxide (titanicone) followed by calcination to produce inorganic NF membranes[17]. The substrate was composed of two layers, a supporting and separating layer. The predominant deposition in the near-surface region was limited through short precursor pulse and exposure. As the number of ALD cycles increase, a compaque layer of titanicone is deposited on the separation layer at the top. After calcination, the titanicone was transformed to TiO₂, confirmed by XRD analysis. After 100 deposition cycles, the PWP value of the membranes were approximately 80 L m^-2 h^-1 bar^-1, which was lower than the bare membranes, which showed a PWF of approximately 100 L m^-2 h^-1 bar^-1 due to the TiO₂ layer blocking the ceramic membrane pores. PWF dropped even further to 45 L m^-2 h^-1 bar^-1 after 200 cycles and maintained a steady value of around 30 L m^-2 h^-1 bar^-1 after 300 cycles, indicating the formation of a microporous TiO₂ separation layer. Porosity decreased as the number of ALD cycles increased, from approximately 45% at 50 cycles then stabilizing to approximately 30% at 300 cycles. MWCOs were calculated using to determine the retention performance of the membranes. MWCO showed similar trends to permeability, as it decreased as the number of cycles increased. MWCO was calculated at 7200 Da for bare membranes, which decreased to 5500 Da, 2600Da, and stabilized at 680 Da for 100, 200, and 300 cycles respectively. Membranes fabricated with 300 or more cycles can be considered NF membranes, which showed high water permeability compared to previous works. This method may be modified with different precursors to further tune the membranes.

Multi-channel ceramic NF membranes were prepared using a novel way that adjusts the pore size of ceramic UF membranes via in situ chemical deposition[18]. Titanium isopropoxide (TTIP) was used as the precursor in the preparation of the membranes. Two different solvents, isopropanol and ethanol, were investigated in the preparation step using ceramic membranes with 5 kDa MWCO as the substrate. Under 20 wt% precursor concentration and 350°C temperature for thermal treatment, the membranes showed a dramatic decrease in pure water permeability and MWCO following the in situ chemical deposition. Both solvents resulted in similar membrane performances and decrease in pore size. To assess the effect of precursors in membrane fabrication, TTIP and tetrabutyl titanate (TBOT) were selected. Both membranes prepared by the two precursors showed similar values for water permeation and retention MW, both of which had improved from the original membrane. As precursor concentration increased, pore volume, specific surface area, permeability of water, and MWCO decreased. Pure water permeability decreased due to the higher number of pore channels filled as precursor increased. Thermal treatment temperatures of 100~200°C resulted in a rapid decrease in weight of the powder while the powder stabilized at a constant weight at 300°C thermal treatment temperature. A pilot scale-up of the membrane preparation process was designed for industrial applications. Three multi-channel ceramic UF membranes with different pore radius were treated in the system. Ceramic UF substrates with pore radius under 2.5 nm could be prepared to a ceramic NF membrane with MWCO of around 1 kDa and 1.0 nm average pore radius. The NF membrane showed a drastic decrease in pure water permeability, although the values remained high for an NF membrane. In the recovery of sodium dehydroacetate, the NF membranes showed 98% recovery rate.

Polyamide/TiO₂ NF membranes were prepared using modified ceramic hollow fiber (CHF) microfiltration membranes as a substrate and polyethyleneimine (PEI) and trimesoyl chloride (TMC) as the monomers for IP method[19]. The TiO₂-modified ceramic hollow fiber (TiCHF) membranes showed a smaller pure water permeability (PWP) compared to CHF membranes and 89.3% dextran 70 rejection due to the TiO₂ layer. An increase in PEI concentration from 0.3 to 0.6 w/v% resulted in a significant PWP decrease and an increase in magnesium chloride rejection increased from 76.7 to 90.8%. At these concentrations, cross-linking degree was low. 0.9 w/v% was determined to be the optimal PEI concentration improve cross-linking density and salt rejection, with 78.6 L m^-2 h^-1 PWF and 92.1% rejection. TMC concentration had similar effects to PWF, with PWF decreasing as TMC concentration increased. A TMC concentration of 0.3 w/v% was determined to be optimal, considering the relatively high PWF of 105.5 L m^-2 h^-1 and MgCl₂ rejection (95.5%). While the formation of a dense active layer was successful in a short reaction time (5s), the IP procedure needed 20s for a successful progress of the reaction. The TiO₂-modified ceramic hollow fiber nanofiltration (TiCHFNF) membranes showed a dense and uniform skin layer and had superhydrophilic properties. Salt rejection of NaCl, Na₂SO₄, MgCl₂, and MgSO₄ were 95,5%, 84.0%, 69.2%, and 44.5% respectively. TiCHFNF membranes were more effective in the rejection compared to polymeric NF membranes. The antifouling performance and stability of the membranes were investigated using BSA as an adsorbent the washing with 0.5 wt% alkali solution for 0.5 h. The flux recovery ratio (FRR) was 75% before washing, and after 7 cycles of contamination-recovery, rejection of was above 96.1% and flux was maximum 83 L m^-2 h^-1. In similar tests using HA, its rejection was 100% and flux was around 99 L m^-2 h^-1. The membranes showed good long-term stability, with little change in PWF and having MgCl₂ rejection higher than 94.3% after a 180 h testing period.

Two types of nanofiltration (NF) membranes were investigated for applications in the treatment of acid mine waters (AMW)[20]. The isoelectric point (IEP) of the TiO₂ ceramic membranes were determined to be 5.4 ± 0.5 while the polymeric MPF-34 membranes had a value ranging from 4.5 to 5.5. The rejection of Na₂SO₄ increased as the pH of the solution increased. The ceramic membrane consisted of a TiO₂ layer and a Al₂O₃ layer while the MPF-34 membrane consisted of three layers: polyester, support layer, and the active layer. pH was increased using H₂SO₄, to determine the effects of solution acidity on membrane performance. Both membranes showed H+ rejection below 10%, while HSO4- rejections were below 20% and below 52% for the ceramic and polymeric membranes respectively. The polymeric membrane showed a superior metal rejection at 80 % compared to the ceramic membranes which had values below 30%. The metal rejection depended on the metal ion properties for the ceramic membrane, which had highest rejection for Al(III) at 25-30%, and decreasing rejection for REEs(III), transition metal ions (Cu(II), ZN(II)), and Ca(II). MPF-34 membranes showed high rejection for all metals and an increase in sulphate rejection but had a low permeate flux compared to its ceramic counterpart. As Al(III) concentration in the solution increased, sulphate rejection in both the ceramic membrane and MPF-34 membrane increased, from 4-12% to 13-19% and from 29-41% to 42-47% respectively. Increasing the Fe(III) concentration in the solution resulted in an increase in sulfate rejection of the ceramic membranes but did not affect the sulfate rejection of the polymeric membranes significantly. Both membranes showed the lowest concentration factor values below 1 for H+ due to its negative rejections.

Atmospheric pressure atomic layer deposition (APALD) was used to prepare highly permeable TiO₂ deposited tight ceramic nanofiltration (NF) membranes. In coating silicon wafers, as APALD cycles increased, the thickness of the deposited TiO₂ layer increased by a growth-per-cycle (GPC) rate of 0.39 nm[21]. These values were higher than ALD deposited TiO₂ layers, which had a GPC rate of 0.04-0.06 nm due to the vacuum environment of the ALD process. TiO₂ layers were successfully deposited on the wafers using APALD. The pore size distribution was determined by the rejection of various PEG and CO₂ adsorption method. Most (90%) pores on the coated membranes had a similar pore size to pristine membranes, at 0.5-0.8 nm, indicating the actual growth rate of TiO₂ on the membranes was lower than on silicone wafers. Water permeability was between 11 and 16 L m^-2 h^-1 bar1 for membranes with MWCO of 260-380 Da. Water permeability slightly decreased as the number of cycles increased due to precursors being impregnated and deposited into the separation layer.

A novel ceramic forward osmosis (FO) membrane having a nanocomposite interlayer was prepared with nanocomposite TiO₂/CNT interlayer on a mullite substrate prepared via dry-wet spinning technique and sintering[ 22].

The ceramic substrate had an asymmetric structure consisting of a sponge like structure in between long inner and outer fingerlike macrovoids. The outer layer had a uniform porous structure with porosity of 22 ± 1.2%, which helped reduce internal concentration polarization (ICP). The intermediate sponge like layer helped the substrate gain high mechanical stability but impeded the formation of a thin PA layer on ceramic substrate. Thus, TiO₂ and CNT interlayers were coated onto the substrate to provide a smoother surface with smaller pores. Three different membranes, M-PA, M-T-PA, and M-T/CNT-PA were created with mullite substrate, TiO₂ coated mullite substrate, and TiO₂/CNT coated mullite substrate respectively. M-PA membranes had a thicker PA layer and larger pores, nor did it show a ridge-and-valley structure or nanovoids. In contrast, both M-T-PA and M-T/CAN-PA membranes showed nanovoids and ridge-and-valley structure, due to the smaller pores and smoother surface. M-T/ CNT-PA membranes showed a higher number of nanovoids and had a slightly higher surface roughness compared to M-T-PA membranes. In addition, M-T/CNT-PA membranes also had higher water permeability as well as salt rejection. The higher water selectivity compared to salts was due to less defects in the membrane and a higher cross-linking in the PA layer. All three values were significantly lower than commercial CTA FO membranes, which possess 950 μm S value. In pressure retarded osmosis (PRO) mode, M-T-PA membranes showed an approximately 75% increase in water flux, with 18.6 ± 1.7 L.m^-2 h^-1 compared to M-PA membranes due to the decreased thickness of the PA layer. M-T/CNT-PA membranes showed even further improvement by 168% due to the presence of the CNT-constructed network layer helping form a rougher surface PA layer with higher number of nanovoids. Similar results were shown in FO. Water flux and reverse salt flux increased for all three membranes as sodium chloride concentration increased in both PRO and FO modes, due to the higher osmotic force. The specific salt flux (J_s/J_w) was lowest and water flux was highest in M-T/CNT-PA membranes for both PRO and FO modes. As the thickness of CNT layer increased up to 222 nm, maximum water flux of 28.5 ± 1.8 L.m^-2 h^-1 was achieved, while further increase to 304 nm caused a decrease in water flux to 22.7 ± 1.5 L.m^-2 h^-1.

2.2. Composite

Composite ceramic membrane is prepared to improve the NF properties in which the excellent properties of all the components can be combined. A filtration-coating method was used to prepare Al-Zr ceramic nanofiltration (NF) membranes on ceramic ultrafiltration (UF) membranes to assess its applications in uranium removal[23].

After the filtration-coating process of the UF membrane with a 5 μm thick layer, the pore size of the membrane decreased from a typical UF membrane pore size (303.2 ± 118.3 nm) to a NF pore size (4.3 ± 0.7 nm). In assessing the molecular MWCO, the ceramic NF membranes showed 90% PEG rejection around 1,000 Da and 24 and 42% for 200 and 400 Da respectively while the ceramic UF membrane showed less than 5% for all PEG. The pure water flux (J_0-NF) value of NF membranes was measured at 141 ± 2 LMH, while pure water flux (J_0-UF) of the UF membranes was 364 ± 14 LMH. In terms of uranium rejection, three different pH levels were tested. The maximum uranium rejection (91 ± 2%) was found at pH 7.4 and the minimum rejection (73 ± 2%) was found at pH 5.0. The effect of natural organic matter (NOM) was evaluated through the addition of SR-NOM in the uranium solution. In the absence of SR-NOM, the absorption rate of uranium was at 26.1 ± 2.0% at pH 7.4. After SR-NOM was added, the rate decreased dramatically to 6.7%.

TiO₂/ZrO₂ composite ceramic membranes having various pore sizes were fabricated using the sol-gel process and adjusting the hydrolysis ratio to control pore size[24]. As hydrolysis ratio increases, particle size distribution becomes broader with increase in sol size. The membranes derived from different sol gel sizes were tabled TZ-5, TZ-8, TZ-10 for 5, 8, and 10 nm respectively. As sol gel size increased, higher water flux was recorded due to the smaller pore size blocking the pores in the intermediate layer. The MWCO of the membranes increased from 620 to 860 Da as sol gel size increased from 5 to 8 nm. With pore sizes 1.2, 1.4, and 1.5 nm for TZ-5, TZ-8, and TZ_10 respectively, all three prepared membranes showed adequate MWCO and pore size for nanofiltration. The zeta potential of the membranes was positive in two solutions (CaCl₂ and MgCl₂) while it was negative in one (Na₂SO₄). In NaCl solution, the zeta potential depended on the pH value, with a positive zeta potential in an acidic solution and a negative zeta potential in an alkaline solution. At a pH of 6, the membranes had the highest rejection to MgCl₂ followed by CaCl₂ , NaCl, and Na₂SO₄, due to the Donnan exclusion mechanism.

3. Conclusions

Fouling of membrane in nanofiltration process is a huge problem specially in wastewater treatment. Polymeric membrane is affected both by reversible as well as irreversible fouling and treating with chlorine containing compound to get rid of it. Chlorine treatment reduce the life time of the membrane due to the instability of amide linkage in polyamide thin film membrane. Ceramic membrane are more robust as they are stable to harsh chemical environment, high temperature and membrane fouling can be taken care of by chemical treatment by treating under any harsh environment. This review discusses about various type of ceramic membrane applied for wastewater treatment in nanofiltration process.