Fabrication of High-Performance Nanofiltration Membrane Using Polydopamine and Carbon Nitride as the Interlayer

Abstract

In order to recover lithium from brine with a high Mg²⁺/Li⁺ ratio, a positively charged nanofiltration (NF) membrane was prepared by depositing polydopamine (PDA)-coated graphitic carbon nitride (g-C₃N₄) as the interlayer (PDA-g-C₃N₄) and the interfacial polymerization (IP) of polyethyleneimine (PEI) and trimesoyl chloride (TMC) was carried out. Under optimal conditions, the water contact angle of the composite membrane is only 55.5° and the isoelectric point (IEP) is 6.01. The final positively charged NF membrane (M5) exhibits high permeance (10.19 L·m⁻²·h⁻¹·bar⁻¹) and high rejection of Mg²⁺ (98.20%) but low rejection of Li⁺ (13.33%). The separation factor (SF) is up to 48.08, and the Mg²⁺/Li⁺ ratio of the permeate is 0.036 in the simulated brine. In conclusion, the M5 membrane shows a good separation performance for salt lake brine (SF = 12.79 and Mg²⁺/Li⁺ ratio of the permeate = 1.43) and good fouling resistance. Therefore, the positively charged M5 membrane with PDA-g-C₃N₄ as the interlayer has the potential to be used for the recovery of lithium from brine.

1. Introduction

Lithium is known as “an energy metal in the 21st century” and its demand in the energy field, e.g., as batteries and nuclear energy, has been increasing in recent years [1,2,3]. Statistics reveal that over 60% of lithium reserves are present in continental brine [4], which highlights the importance of developing more efficient technology for the recovery of lithium from brine [5,6,7]. Very often, brine is heavily loaded with Mg²⁺ in comparison to Li⁺ because of their similar ionic properties, which undoubtedly makes it more difficult for recovery of lithium from the brine. Membrane separation is considered one of the most competitive separation technologies owing to its simplicity, low cost, high energy efficiency, and eco-friendliness [8,9]. Especially nanofiltration (NF), as a pressure-driven membrane separation process, has been extensively utilized in wastewater treatment and the recovery of highly-valued products [10]. Compared to conventional separation methods, such as precipitation, chromatography, ion exchange, and liquid–liquid extraction, NF is more effective in separating lithium from brine with a high Mg²⁺/Li⁺ ratio, due to the special mono-/divalent ion separation [11]. For example, Somrani et al. studied the recovery of lithium from salt lake brine by NF and low-pressure RO membrane [12]. Sun et al. studied the separation of Li⁺ and Mg²⁺ from brine using a desalination NF membrane [13]. Their results indicated that the structural and surface properties of NF membrane play a crucial role in the separation performance [14]. There is also evidence that a positively charged NF membrane is more desirable for the removal of multivalent cations [15,16].

Nowadays, thin-film composite (TFC) membranes consisting of a dense thin film layer and a porous support layer are the primary configuration for NF membranes. These TFC membranes are usually prepared via interfacial polymerization (IP) that is highly controllable and practicable [17,18]. A positively charged polyamide (PA) TFC membrane has been successfully prepared from the reaction of polyethyleneimine (PEI), an amine-containing monomer, with trimesoyl chloride (TMC) by IP process [19]. Previous studies suggest that the physical and chemical properties of substrates, which normally are commercial ultrafiltration (UF) membranes, have important effects on the formation of the PA layer and the performance of resultant membranes. Nevertheless, the surfaces of most commercial UF membranes (e.g., polyether sulfone (PES) UF membrane) are hydrophobic with relatively low porosity, which may have an impact on the adsorption and penetration of amine monomers on the substrate and thus reduce the adhesion between the PA layer and the substrate. As a result, the resultant NF membranes may have low water permeance and are susceptible to fouling [20,21,22].

In order to solve these problems, an interlayer is proposed to modify the substrate to form a modified PA TFC membrane with a sandwich structure, which can increase the hydrophilicity, permeance, and the separation performance of membranes. Dopamine (DA) deposition is a new and promising surface modification method because of mild conditions and simple processes, and it can be carried out in weakly alkaline solutions at air atmosphere and room temperature [23]. Notably, this process has universal applicability to substrates and does not affect the functionalization of substrates [24]. Polydopamine (PDA) as the interlayer can enhance the stability and compatibility of the separation layer with the base membrane. Nanocomposite technology has attracted much attention in modulating the structural and physicochemical properties of membranes, such as hydrophilicity, fouling resistance, porosity, charge density, chemical, and thermal and mechanical stability [25]. Graphitic carbon nitride (g-C₃N₄) is a new type of two-dimensional nanosheets material with a graphene-like layered structure, large specific surface area, and visible light absorption. Due to its special tri-s-triazine structural unit, it also has excellent mechanical, thermal, and chemical stability and photodegradability [26,27]. More importantly, g-C₃N₄ is easy to prepare and is low-cost. Thus, it is applicable to many fields, such as photocatalysis, gas storage, and membrane separation [28]. The addition of g-C₃N₄ nanosheets contributes to improving the permeance and antifouling ability of the membranes [29]. In this study, PDA-coated g-C₃N₄ (PDA-g-C₃N₄) was used as the interlayer to modify the PES substrate, and then a positively charged PEI/TMC NF membrane was prepared by the IP method.

2. Experiments

2.1. Chemicals

PES UF membranes were provided by Microdyn-Nadir Co., Ltd. (Xiamen, China). PEI (Mw70000, 50%, AR), TMC (99%, AR), lithium chloride (LiCl, AR), magnesium chloride hexahydrate (MgCl₂⋅6H₂O, AR), dopamine hydrochloride (99%), N, N-dimethylformamide (DMF, AR), and tris (hydroxymethyl) aminomethane (Tris, 99.8%) were purchased from Aladdin Co., Ltd. (Shanghai, China). HCl (AR), KOH (AR), ethanol (AR), polyethylene glycol (PEG, 200, 400, 600, 800, 1000, 1500 Da, AR, Adamas-beta^®), n-hexane (98%, AR, Adamas-beta^®), bovine serum albumin (BSA, AR, Adamas-beta^®), and lysozyme (AR, Adamas-beta^®) were provided by Titan Polytron Technologies Inc. (Shanghai, China). Hydrazine monohydrate (AR) was purchased from Siyan Biotech-nology Co., Ltd. (Shanghai, China).

2.2. Membrane Preparation

2.2.1. Preparation of Stripped g-C₃N₄

The preparation of g-C₃N₄ was described in the previous work from our lab [30]: g-C₃N₄ was stripped with hydrazine monohydrate. Briefly, 1.2 g of g-C₃N₄ was dispersed in 35 mL of hydrazine monohydrate at room temperature and sonicated for 10 min, and then 15 mL of DMF was added and sonicated for another 10 min. Stripped g-C₃N₄ (g-C₃N₄-96) was obtained after stirring at 40 °C for 96 h and then washed thoroughly with pure water and freeze-dried for 24 h. Similarly, g-C₃N₄ with different stripping times (24 h, 48 h, and 192 h) were also prepared.

2.2.2. Preparation of NF Membrane

The schematic of the preparation of NF membranes is shown in Figure 1, and the preparation conditions are summarized in

Table 1. Preparation conditions of NF membranes.

No.	Membrane	Deposition of PDA (h)	Content of g-C3N4-96 (wt%)
M0	PES	-	-
M1	PES/PEI/TMC	-	-
M2	PES/PEI-g-C3N4-96/TMC	-	0.02
M3	PES/DA/PEI/TMC	2 h	-
M4	PES/DA/PEI-g-C3N4-96/TMC	2 h	0.02
M5	PES/DA-g-C3N4-96/PEI/TMC	2 h	0.02

. First, PES UF membranes (M0) that were used as the base membranes for surface modification were soaked in ethanol for 2 h to remove air bubbles and impurities in and on the membranes, and then rinsed with pure water. After that, DA and g-C₃N₄-modified NF membranes were prepared as follows: (1) in Figure 1a, DA was immersed and then g-C₃N₄ was added to the PEI aqueous phase to prepare NF membrane (M4); and (2) in Figure 1b, g-C₃N₄ was dispersed in DA aqueous solution (2.0 g·L⁻¹, pH = 8.5, DA hydrochloride in 50 mM·L⁻¹ Tris buffer solution) using a probe sonicator (Biosafer 650-92, China) for 10 min to prepare DA-g-C₃N₄ aqueous solutions of different concentrations (0.01 wt%, 0.02 wt%, 0.03 wt%, and 0.04 wt%). PES membranes were immersed in DA-g-C₃N₄ aqueous solution at 25 °C for 2 h, and subsequently unreacted DA-g-C₃N₄ aqueous solution or low molecular weight PDA aggregates were removed to obtain PES membranes with PDA-g-C₃N₄ as the interlayer. Then, PEI aqueous solution (0.6 wt%) was poured on the interlayer for 2 min. Excess aqueous solution was removed, and then the membrane was covered with the TMC organic solution (0.1 wt%) in n-hexane for 2 min to allow the IP process to occur. Excess TMC was removed, and the membrane was cured at 70 °C for 15 min and then stored in pure water before use.

2.3. Characterization and Measurement

The chemical functional groups of g-C₃N₄ and the membrane surfaces were determined by Fourier transform infrared spectroscopy (FTIR, Thermo Fisher 6700 spectrometer, Waltham, MA, USA). The morphologies of inorganic materials (g-C₃N₄-96) were measured by transmission electron microscopy (TEM, a JEM-2100F, Tokyo, Japan). The element composition was determined by X-ray photoelectron spectroscopy (XPS, ESCALAB 250Xi, USA). The surface and cross-sectional morphologies of membranes were determined by scanning electron microscopy (SEM, SU8010, S4800, Japan). The surface zeta potential of membranes was characterized with a Sur-PASS electrokinetic analyzer (Anton Paar, Graz, Austria). Specifically, 1 mmol·L⁻¹ KCl aqueous solution was used as the electrolyte, and 0.1 mol·L⁻¹ HCl and 0.1 mol·L⁻¹ KOH aqueous solution was used to control the pH value (3–10). The gap height of the measuring cell was fixed at 100 μm and the membranes were soaked in electrolyte for at least 4 h before the measurement. The water contact angle (WCA) was measured on a contact angle goniometer (JC2000D2, China) at room temperature with 1.5 μL of pure water droplet. The cation concentrations were measured using an atomic absorption spectrometer (A3F-13, China). The total organic carbon (TOC) was determined by a TOC-VCPN analyzer (Japan).

2.4. Permeance and Separation Performance of NF Membranes

The permeance and separation performance for mixed salt solutions (MgCl₂·6H₂O and LiCl, Mg²⁺/Li⁺ = 20, 48, 84, pH = 6.64, 6.59, 6.58) and the salt lake brine obtained from CITIC Guoan were tested using a lab-made cross-flow filtration system. The temperature of the feed tank jacket was kept at 20 ± 0.5 °C. Each membrane was pressurized with pure water at 0.4 MPa and 20 ± 0.5 °C for 0.5 h prior to testing to reach a steady state. The solution was replaced with 2.0 g·L⁻¹ salt solutions or salt lake brine diluted 200 times and continuously fed for 0.5 h at 20 ± 0.5 °C and 0.4 MPa. The salt lake brine was diluted to the concentration of the mixed salt solution in order to compare with the results in the literature and reduce the osmotic pressure during the operation. The salt lake brine was used in this experiment to investigate the separation effect of NF membrane for complex solutions and fresh water was used for dilution. The composition of salt lake brine is very complex, and the pH and composition of the salt lake brine and diluted salt lake brine are shown in

Table 2. The pH and composition of the salt lake brine (1) and diluted salt lake brine (2) (g·L⁻¹).

	pH	Li+	Na+	K+	Ca2+	Mg2+	Ba2+	Al3+	B3+	Mg2+/Li+
1	4.23	4.846	4.930	0.3850	0.2698	132.7	5.454	0.7422	5.556	27.38
2	8.56	0.0242	0.0247	0.0019	0.0013	0.6634	0.0272	0.0037	0.2778	27.38

.

The proportion of strong acid–weak base salts or strong base–weak acid salts varies with the decrease or increase of neutral components, such as sodium chloride, sodium sulfate, and water, and the hydrolysis of these salts lowers or raises the pH of the salt lake brine [31]. The rejection rate (R, %), permeance (L·m⁻²·h⁻¹·bar⁻¹), and selectivity factor (SF) of Mg²⁺/Li⁺ were calculated by Equations (1)–(3), respectively. C_p (g·L⁻¹) is the concentration of the permeation, C_f (g·L⁻¹) is the concentration of the feed solution, V (L) is the volume of the permeation, A (m²) is the active surface area of the membrane, Δt (h) is the permeation time, Δp (bar) is the test pressure, C_Li (g·L⁻¹) and C_Mg (g·L⁻¹) are the Li⁺ concentration and Mg²⁺ concentration in permeate and feed solutions.

$$\mathrm{R}=(1-\frac{{\mathrm{C}}_{\mathrm{p}}}{{\mathrm{C}}_{\mathrm{f}}})\times 100\%$$

(1)

$$\mathrm{Permeance}=\frac{V}{A\times \Delta t\times \Delta p}$$

(2)

$$\mathrm{SF}=\frac{{{\displaystyle ({{\displaystyle \mathrm{C}}}_{\mathrm{L}\mathrm{i}}/{{\displaystyle \mathrm{C}}}_{\mathrm{M}\mathrm{g}})}}_{\mathrm{permeate}}}{{{\displaystyle ({{\displaystyle \mathrm{C}}}_{\mathrm{L}\mathrm{i}}/{{\displaystyle \mathrm{C}}}_{\mathrm{M}\mathrm{g}})}}_{\mathrm{feed}}}$$

(3)

2.5. Mean Pore Size and Pore Size Distribution

The molecular weight cut off (MWCO) and effective membrane pore size were determined via different molecular weights of 0.2 g⋅L⁻¹ PEGs (200, 400, 600, 800, 1000, and 1500 Da) under 0.4 MPa and 20 ± 0.5 °C. The stokes diameter (d, nm) of PEG could be calculated according to its average molecular weight (Mw) by Equation (4). The pore size distribution of membranes was calculated by Equation (5) [31,32,33,34,35], where d_p (nm) is the pore size in diameter, σ_p is the ratio of solute diameter at a rejection rate of 84.13%, and 50%, μ_p (nm) is the geometric mean diameter of the solute at a rejection rate of 50%.

$$d=33.46\times {10}^{-3}\times M{w}^{0.557}$$

(4)

$$\frac{dF\left(d_p\right)}{d{d}_p}=\frac{1}{d_{p}ln{\sigma }_p\sqrt{2\pi }}exp\left[-\frac{{\left(ln{d}_p-ln{\mu }_p\right)}^2}{2{\left(ln{\sigma }_p\right)}^2}\right]$$

(5)

2.6. Antifouling Property of Membrane

The antifouling property was investigated in the cross-flow filtration module by BSA (IEP = 4.7, 0.1 g⋅L⁻¹, pH = 6.95) and lysozyme (IEP = 10.7, 0.1 g⋅L⁻¹, pH = 6.68) solution respectively. To reach a steady state, the membrane was pressurized with pure water at 0.4 MPa and 20 ± 0.5 °C for 1 h, and the volumetric permeate flux (J_v0) was measured every 20 min. The BSA solution was continuously fed for 2 h and the steady flux (J_v1) was measured every 20 min. After that, the tested membranes were washed with pure water at 0.1 MPa and 20 ± 0.5 °C for 20 min, and the permeate flux of pure water (J_v2) was measured again under the same operating conditions. After repeating the BSA antifouling test twice, the membranes were cleaned again, the BSA solution was replaced with a lysozyme solution and the same procedure was performed as described above. The steady flux (J_v, L·m⁻²·h⁻¹), relative flux (RF), flux recovery ratio (FRR), total fouling ratio (FRt), reversible fouling ratio (FRr), and irreversible fouling ratio (FRir) were calculated by Equations (6)–(11), respectively:

$$J_V=\frac{V}{A\times \Delta t}$$

(6)

$$RF=\frac{J_{v1}}{J_{v0}}\times 100\%$$

(7)

$$FRR=\frac{J_{v2}}{J_{v0}}\times 100\%$$

(8)

$$FRt=\left(1-\frac{J_{v1}}{J_{v0}}\right)\times 100\%$$

(9)

$$FRr=\left(\frac{J_{v2}-J_{v1}}{J_{v0}}\right)\times 100\%$$

(10)

$$FRir=\left(1-\frac{J_{v2}}{J_{v0}}\right)\times 100\%$$

(11)

3. Results and Discussion

3.1. Characterization of g-C₃N₄-96

The FTIR spectra of g-C₃N₄-96 are shown in Figure 2a. In the spectrum of g-C₃N₄, the absorption peak at 802 cm⁻¹ corresponds to the characteristic breathing mode of triazine units, while those peaks in the range of 1234–1637 cm⁻¹ are assigned to the stretching vibration of C-N and C=N heterocycles, which are similar to those reported in previous studies [36,37]. The broad peak in the range of 3000–3500 cm⁻¹ is owed to the stretching vibrations of N-H or N-H₂ originated from uncondensed amino groups [38]. In Figure 2b, a typical ultrathin nanosheets-like architecture with a crinkly structure is observed for g-C₃N₄-96. This is because urea-inorganic ammonium salts were used as additives, and gas was continuously released during the polymerization of melamine to form g-C₃N₄ nanosheets.

3.2. Characterization of Membrane

3.2.1. Chemical Structure and Morphology

The chemical structures of the surface of M0, M1, M2, M3, M4, and M5 membranes were determined by FTIR (Figure 3a). There are several peaks in M1, M2, M3, M4, and M5 membranes: O-H and uncondensed terminal amino groups (3375 cm⁻¹), amide I, C=O band (1665 cm⁻¹), amide II and C-N stretching from IP (1577 cm⁻¹), and the stretching vibration of -COO (1071 and 1012cm⁻¹) originated from the hydrolysis of -COCl in TMC [39]. However, the vibration strength of M1, M2, M3, M4, and M5 membranes at 1665 and 1577 cm⁻¹ is higher than that of M0 membrane. These results indicate that IP reaction occurred on the surface of the NF membrane. The element composition and contents in different layers of the M5 membrane were determined by in situ XPS (Figure 3b). Three main emission peaks are observed, which are ascribed to O1s (531 eV), N1s (399 eV) and C1s (285 eV), respectively. There are two small peaks at 232 eV (Cl2p) and168 eV (S2p) in the PES UF membrane (PES layer), and the Cl content is negligible. The C content is higher, but the N content is lower on the PES layer (

Table 3. Relative elemental content in different layers of M5 membrane (from XPS).

Sample	Elemental Relative Content (at %)
	C	N	O	Cl	S
PEI+TMC layer	70.19	11.08	17.37	1.35	-
PDA-g-C3N4 layer	73.68	6.95	16.03	0.89	2.44
PES layer	79.88	2.96	14.08	-	3.08

). Compared to the PES layer, the contents of N and O elements are higher in the PDA-g-C₃N₄ interlayer, which is due to the high N content in PDA and g-C₃N₄ [40], and the N content is increased from 2.96% to 6.95%. The N content in the PEI + TMC layer is increased to 11.08%, which is mainly derived from N-C=O and NH₂ (NHR) groups. The amide groups are attributed to the introduction of O=C-N groups by IP of PEI and TMC, and a dense PA layer is formed on the surfaces of membranes. The primary and secondary amines are derived from unreacted PEI. Thus, the cross-linking degree of the NF membrane can be assessed by the percentage of amide groups.

The in situ high resolution C1s and N1s XPS spectra of M5 membrane are deconvoluted and fitted to analyze the surface chemical properties (Figure 4). The C1s XPS spectrum of PES layer is deconvoluted into four peaks (Figure 4a), which are attributed to the C-C bond (284.8 eV), the C-N (285.4 eV) bond derived from the preparation of PES, the C-O bond (286.3 eV), and the C-S bond (291.5 eV) of PES, respectively [41]. The PDA-g-C₃N₄ layer shows four peaks corresponding to C-C (284.8 eV), C-N (285.8 eV), C-O (286.3 eV), and N-C=N (288.0eV) (Figure 4b). The N-C=N bond is attributed to C sp2 hybridization that represents the unique tri-s-triazine rings of g-C₃N₄ [20]. For the PEI + TMC layer, four peaks are observed at 287.1 (O=C-N), 286.3 (C-O), 285.8 (C-N), and 284.8 (C-C) eV (Figure 4c). The peak of O=C-N (287.1 eV) indicates that IP reaction has occurred [42]. The high-resolution N1s spectra are shown in Figure 4d–f. There is only one peak at 400 eV (C-N) for the PES layer, which corresponds to C-N introduced during the preparation of PES membrane. Several new peaks are observed in PDA-g-C₃N₄ and PEI + TMC layers. For the PDA-g-C₃N₄ layer, the peaks at 402.4, 401.2, 399.7 and 390.0 eV correspond to protonated N, conjugated N and C-N bonds, respectively, while for the PEI + TMC layer, the peaks at 402.8, 401.4, 399.8, and 389.9 eV correspond to the protonated N, conjugated N, O=C-N bond, and C-N bond of amides, respectively [43].

The surface and cross-sectional SEM images of the PES, PES/PDA-g-C₃N₄ interlayer, and M5 membranes are shown in Figure 5. The porous surface of PES UF membrane (Figure 5a) disappears after the co-deposition of PDA and g-C₃N₄ and instead a dense layer is formed with PDA aggregation (Figure 5b). The surface of M5 membrane is denser and smoother after the IP of PEI and TMC (Figure 5c). On the other hand, the PES membrane has an asymmetric morphology with micro-porous finger-like and spongy-like structures (Figure 5d). A clearly visible thin selective layer is formed on the surface of PES membrane (Figure 5e,f), which is considered to be the interlayer formed by the co-deposition of PDA-g-C₃N₄ and the active separation layer formed after IP reaction.

3.2.2. Zeta Potential and Hydrophilicity of Membranes

As shown in Figure 6a, the isoelectric point (IEP) of M0 is 4.75, indicating that the PES membrane is negatively charged at pH > 4.75. The IEPs of M1, M2, M3, M4, and M5 membranes (6.01, 7.83, 7.95, 8.24, and 8.67, respectively) are much higher compared to M0 membrane, owing to a large number of positive unreacted primary and secondary amine groups (-NH³⁺ and -NH²⁺) from PEI [44]. M1, M2, M3, M4, and M5 membranes are positively charged at pH < 6.01. The IEPs of M3, M4, and M5 membranes are lower than that of the M1 membrane, which is attributed to the presence of abundant -OH and C-O on the PDA layer. Although the IEP of M5 membrane is only 6.01, its relative SF is the highest because of the effects of hydrophilicity and pore size on separation performance. NF membranes generally operate at a pH of about 6.0 [45], and theM5 membrane is positively charged during NF operation. The WCA was measured to analyze the hydrophilicity of membrane surfaces (Figure 6b). The WCAs follow the order of M1 > M0 > M5 > M4 > M3 > M2. It should be noted that the WCAs of the M1 membrane are always higher than that of other membranes, indicating that the PA NF membrane is less hydrophilic. The g-C₃N₄ in the PEI aqueous phase forms hydrogen bonds with water molecules, which enhances the affinity of water molecules with the membrane surface and reduces the WCAs of M2 membrane [22]. The hydrophilicity can be improved by adding g-C₃N₄ or depositing DA on the surface of the base membrane. The WCAs of the M5 membrane decrease because of the formation of the PDA-g-C₃N₄ interlayer.

3.2.3. MWCO and Pore Size Distribution of M5 Membrane

The MWCO can be used to evaluate the pore size of the NF membranes, which is obtained by the Mw as the rejection of neutral solutes reaches 90%. Figure 7 shows the rejection of M5 membrane to neutral solutes (PEG 200, 400, 600, 800, 1000, and 1500 Da). The MWCO of M5 membrane is about 755 Da, and the pore radius is about 0.41 nm. It is known that the MWCO value of NF membranes are in the range of 200–1000 Da. It can be concluded that M5 membrane fabricated in this study is a NF membrane, and a dense NF active layer is formed on the surface of M5 membrane.

3.3. Separation Performance and Antifouling Property

3.3.1. Effects of Preparation Conditions on NF Membrane Performance

Figure 8a,b shows the effects of PEI concentration and immersion time on the SF and permeance of NF membranes. As the PEI concentration increases from 0.2 to 0.8 wt%, the permeance increases, the SF first increases and then decreases. The IP process is slower at lower PEI concentration, leading to a loose PA selective layer with lower selective separation performance. The IP rate increases with increasing PEI concentration, forming a denser PA selective layer. However, the continuous increase of PEI concentration would enhance the permeance and reduce the selective separation performance. This is because at high PEI concentrations, there is a limited supply of TMC and a large number of unreacted amine groups are left in the PA active layer, which would reduce the degree of cross-linking [45]. Thus, the optimal PEI concentration is determined to be 0.6 wt%. Increasing the immersion time leads to no significant improvement of the separation performance but a reduction in the permeance. Thus, the optimal immersion time is determined to be 2 min. As shown in Figure 8c,d, as the concentration of TMC in n-hexane increases from 0.05 to 0.4 wt%, the separation performance first increases and then decreases, while the permeance gradually decreases. This is because there is no sufficient TMC to react with PEI at low TMC concentrations and the IP is weak. However, once the optimum IP is achieved, the increasing of the TMC concentration would actually decrease the separation performance, because the hydrolysis of unreacted acid chloride groups in TMC to carboxyl groups reduces the degree of IP between PEI and TMC. Considering that the IP reaction time is short, increasing the time will not help to increase the permeance. Thus, the TMC concentration of 0.1 wt% and the reaction time of 2 min are considered to be the best choice. The effects of g-C₃N₄ concentration and the immersion time of DA are shown in Figure 8e,f, respectively. Adding g-C₃N₄ could increase the permeance of the membrane, but too high g-C₃N₄ content would reduce the separation performance. Thus, the optimum amount of g-C₃N₄ is determined to be 0.02 wt%. Once the immersion time exceeds 2 h, increasing the immersion time will not increase the separation performance but will significantly reduce the permeance. Therefore, 2 h is selected as the most appropriate immersion time, which is shorter than that reported in the previous literature [23,24]. Figure 9 shows the effect of stripping time of g-C₃N₄ (24 h, 48 h, 96 h, and 192 h) on the separation performance and permeance. It an be seen that the membrane fabricated for 96 h has the best separation performance and permeance.

The Mg²⁺ rejection, Li⁺ rejection, SF and the permeance of M1, M2, M3, M4, and M5 membranes in 2.0 g·L⁻¹ LiCl and MgCl₂ mixed solution (Mg²⁺/Li⁺ = 20) are compared (Figure 10). For M4 membrane, the rejection is 94.61% for Mg²⁺ and 27.82% for Li⁺, the SF is 13.39, and the permeance is 9.14 L·m⁻²·h⁻¹·bar⁻¹. M5 membrane exhibits a permeance up to 10.19 L·m⁻²·h⁻¹·bar⁻¹ and a SF up to 48.08, the rejection is high (98.20%) for Mg²⁺ but low (13.33%) for Li⁺, and the Mg²⁺/Li⁺ ratio of the permeate is 0.036. As shown in Figure 1, for M4 membrane, g-C₃N₄ is an additive to the PEI solution and a PA layer is from the the IP reaction with IMC, while for M5 membranes, g-C₃N₄ is co-deposited with DA as the interlayer. It is seen that the separation performance of M4 membrane is not much improved compared to M3 membrane (deposited PDA, no g-C₃N₄), which indicates that the presence of g-C₃N₄ in the IP layer has no significant effect on the membrane properties. The fabrication of a defect-free intermediate layer with an appropriate thickness is an important step for the preparation of multilayer composite membranes with high-permeability [46,47]. The better performance of M5 membrane could be attributed to the use of PDA-g-C₃N₄ as the interlayer material to avoid pore penetration of the selective layer. The addition of g-C₃N₄ nanosheets can form a thin and defect-free interlayer with PDA and completely cover the surface of the PES substrate [48]. Thus, the positively charged NF membrane prepared with PDA-g-C₃N₄ as the interlayer has better separation performance and permeance.

3.3.2. Separation Performance and Permeance of M5 in Different Solutions

Based on the excellent separation ability of Mg²⁺ and Li⁺ in the salt solution with a Mg²⁺/Li⁺ ratio of 20, the separation performance and permeance of M5 membrane in salt solutions with a Mg²⁺/Li⁺ ratio of 48 and 81 are examined. It is found that at a Mg²⁺/Li⁺ ratio of 48, the SF of M5 membrane is 15.14 and the permeance is 8.60 L·m⁻²·h⁻¹·bar⁻¹; while at a Mg²⁺/Li⁺ ratio of 84, the SF is 10.77 and the permeance is 7.51 L·m⁻²·h⁻¹·bar⁻¹ (Figure 11). The SF of the M5 membrane decreases with the increase of the Mg²⁺/Li⁺ ratio, but the rejection of Mg²⁺ is always higher than 92.36%. In order to further investigate the separation performance of M5 membrane for Mg²⁺ and Li⁺ in complex solutions, the salt lake brine collected from CITIC Guoan was used in the separation experiments. The rejection of Mg²⁺ of M5 membrane is dropped to 88.94%, and the Mg²⁺/Li⁺ ratio of the permeate is 1.43, which may be related to the complex composition of salt lake brine. There may be other monovalent (e.g., Na⁺ and K⁺) and divalent cations, in addition to Mg²⁺ and Li⁺. In spite of this, the M5 membrane still exhibits an excellent separation of Mg²⁺ and Li⁺ (SF = 12.79).

Table 4. Comparison of separation performance of NF membranes in the literature and this work.

Membrane	Operation Conditions	Feed Solution	Rejection Rate of Mg2+ and SF	Permeance (L·m−2·h−1·bar−1)
M5 This work	0.4 MPa, 20 °C	2.0 g⋅L−1 (salt solution), 20	98.4%, 48.07	10.19 to 2.0 g⋅L−1 MgCl2 + LiCl
		2.0 g⋅L−1 (salt solution), 48	93.7%, 12.78	8.27 to 2.0 g⋅L−1 MgCl2 + LiCl
		2.0 g⋅L−1 (salt solution), 84	92.4%, 10.86	8.66 to 2.0 g⋅L−1 MgCl2 + LiCl
		3.0 g⋅L−1 (salt lake brine), 27	88.9%, 12.79	3.80 to 3.0g⋅L−1 salt lake brine
PES/PEI/TMC [39]	0.8 MPa, 25 °C	2.0 g⋅L−1 (salt solution), 20	94.8%, 20.00	3.7 to 2.0 g⋅L−1 MgCl2
PES/PEI/TMC [49]	0.4 MPa, 20 °C	2.0 g⋅L−1 (salt solution), 21.4	96.9%, 7.13	14 to pure water
PES/PEI/TMC [50]	0.8 MPa, 25 °C	2.0 g⋅L−1 (salt solution), 30	no report, 12.15	4.17 to pure water
		2.0 g⋅L−1 (salt solution), 60	95.6%, 5.80	3.40 to pure water
(PES-GO)/PEI/TMC [51]	0.3 MPa, 25 °C	2.0 g⋅L−1 (salt solution), 20	95.1%, 16.13	11.15 to 2.0 g⋅L−1 MgCl2 + LiCl

compares the separation performance of several reported NF membranes and M5 membranes prepared in this work. It was found that the rejection of Mg²⁺ differs substantially among NF membranes. Importantly, the M5 membrane shows higher Mg²⁺ rejection and permeance than other NF membranes. According to Figure 12, the permeance (5.01–6.06), Mg²⁺/Li⁺ ratio of the permeate (0.73–1.43), and SF (10.28–16.69) change little during the longtime filtration test. As a result, the M5 membrane has high stability during the filtration process.

3.3.3. Antifouling Property of the M5 Membrane

The variations of water flux with time and the antifouling property of M5 membrane during the cyclic filtration for BSA and lysozyme solutions are shown in Figure 13. The permeate flux is lower than the water flux and maintained at a steady state in each circle, indicating that the adsorption/deposition and back diffusion of BSA and lysozyme reaches an equilibrium [52]. However, it could not be fully recovered because membrane pores could be blocked by contaminants [53]. In general, the higher the hydrophilicity of the membrane is, the greater the resistance to fouling adsorption will be. This is due to the formation of hydrogen bonds with water molecules on the hydrophilic surface and the formation of a hydrated layer on the membrane surface that can prevent organic contaminants from being absorbed into the membrane matrix [54]. A higher FRR indicates superior antifouling performance of membranes. The values of FRR, FRt, FRr, and FRir change slightly with the increasing filtration cycle, indicating that the M5 membrane has an excellent antifouling property for BSA and lysozyme. As a result, the M5 membrane can efficiently reduce membrane fouling and exhibit good stability during the cyclic BSA and lysozyme filtration.

4. Conclusions

A positively charged NF membrane with a sandwich structure was successfully fabricated by depositing PDA-g-C₃N₄ as the interlayer on the surface of PES membrane, followed by the IP process of PEI and TMC on the interlayer. The concentrations of PEI (0.6 wt%), TMC (0.1 wt%), g-C₃N₄ (0.02 wt%, stripping 96 h), and interlayer reaction time (2 h) are optimized. The final NF membrane (M5) has a low WCA (55.5^°) and the IEP is 6.01, which is attributed to the hydrophilic, positively charged active separation layer on the NF membrane surface resulting from the IP process of PEI and TMC. Notably, the permeance of the M5 membrane is up to 10.19 L·m⁻²·h⁻¹·bar⁻¹ and the rejection of Mg²⁺ is 98.2%. This is because the use of PDA-g-C₃N₄ as the interlayer is conducive to the penetration and diffusion of water molecules on the membrane surface. The high permeance will not affect the salt rejection rate, which is mainly due to the pore size of the NF membrane and the strong electrostatic repulsion between the abundant positive charges and the polyvalent cations on the membrane surface. The M5 membrane also has good and stable separation performance for complex salt lake brine and antifouling property. To conclude, the as-prepared positively charged NF membrane has the potential to be used for recovering Li⁺ from brine due to its ready availability, high efficiency, and good adaptability.