Composite membrane of polyguanidine cationic surface for desalination

As global temperatures increase, environment deterioration and the energy crisis become more and more serious (Bodzek et al. 2020). Pollution control and resource utilization are imperative in various fields including water, soil and air on the Earth (Ahmad et al. 2021). In water treatment areas, separation membrane techniques have displayed special advantages in contaminant treatment and resource collection and enrichment owing to the merits of energy saving, simple operation and moderate space occupation. Among numerous membrane techniques, the nanofiltration (NF) membrane has an important place in soluble separation and hazardous substance removal (Domenech et al. 2020). NF techniques could achieve accurate particle size screening under pore size sieve mechanisms and the Donnan effect (Mohammad et al. 2015; Bi et al. 2020). It has displayed great potential in many fields, including sea water desalination (Padaki et al. 2011), toxic heavy metal ion and dye removal (Qi et al. 2019) and separation of Li⁺/Mg²⁺from their nature mixed solution, such as some salt lakes (Sun et al. 2021).

Among NF films, surface-charged membranes have attracted more and more attention due to excellent ion separation according to their sizes and chemical valence (Akbari et al. 2016; Fang et al. 2018). Particularly, positve-charged membrane produce more electric repulsion to multivalent cations than monovalent ones, which provides an outstanding removal efficiency for multivalence metal ions. Thin layer composite membrane techniques through interfacial polymerization (IP) is a convenient and popular method to prepare positve-charged NF membranes (Wang et al. 2011). Xu et al. (2019) prepared a positve charged composite membrane between polyethyleneimine (PEI) and TMC on a polyethersulfone (PES) ultrafiltration membrane. The membrane displayed great separation capacity for Mg²⁺ and Li⁺. After filtration by the membrane, the Mg²⁺/Li⁺ mass ratio in mixed solution declined from an initial 20 to 1.3. The rejection for Mg²⁺ and Li⁺ was high, up to 76%. Bi et al. (2021) functionalized nanographitic carbon nitride (g-C₃N₄) and used it as a monomer in the aqueous phase. The as-product positve-charged membrane effectively decreased the Mg²⁺/Li⁺ ratio from 73 to 1.85. In addition, the membrane showed excellent antifouling performance.

Commonly, compounds containing amine or sulfhydryl group could introduce positive charges to chemical structures through dissociation to ammonium or sulfonium cations. Polyguanidines, a series of oligomers full of NH groups in a backbone, also exhibited the potential to introduced positive charges to membrane surfaces. Li et al. (2014a) synthesized polyhexamethylene guanidine (PHMG) and used it as aqueous monomer under the pH regulation by triethylamine. The composite membrane of the cationic surface provided salt rejection in the order of MgCl₂ > MgSO₄ > Na₂SO₄ > NaCl. The highest rejection was 86.2% for MgCl_2. The membrane also exhibited wonderful antibacteria and antifouling capacity as PHMG was a polycation biocide (Chindera et al. 2016; Wang et al. 2020). Here, two polyguanidine were investigated as aqueous monomers to prepare an NF composite membrane with TMC n-hexane solution, including PHMG and polyhexamethylene biguanidine (PHMB). Piperazine (PIP) and other bases were investigated as pH regulators to adjust the functional layer structure. Guanidine-PIP films were confirmed as positive-charged membranes, and showed high removal rates for metal cations, especially multivalent cations, such as Mg²⁺, Pd²⁺ and Cd²⁺. The PHMG membrane, G-1, could also enrich Li⁺ from the mixed solution of Mg²⁺ and Li⁺. The membranes also exhibited excellent antibacterial ability towards Escherichia coli (E.C.) and Staphylococcus aureus (S.A.). Polyguanidines possessed attractive potential in the preparation of multifunctional membranes in water treatment.

Polysulfone (Mw = 70,000, industrial purity) was supplied by Solvay Advanced Polymer (Belgium). PHMG and PHMB (M_W around 5,000) were kindly provided by Hebei Jinhong Chemical Co., Ltd. Trimesoyl chloride (TMC), tris(hydroxymethyl)aminomethane (Tris), and triethylamine (TEA) were purchased from Beijing J&K Scientific Co., Ltd. MgCl₂, MgSO₄, NaSO₄, NaCl, Pb(NO₃)₂, Cd(NO₃)₂ and LiCl came from Shanghai Titan Technology Co., Ltd. PSF ultrafiltration supporting membrane was prepared through Yu et al. (2009)'s method.

PIP powder was slowly added into 3 wt% PHMB solution to adjust the pH. Then the aqueous solution was poured onto the PSF supporting membrane. After 30 min, the aqueous solution was poured out. A TMC solution (0.2 wt%) of n-hexane was added onto the membrane and discarded after 15 min (An et al. 2011). After full volatilization of residual n-hexane, the membranes were stored in deionized (DI) water. As a control, a 3 wt% PHMB aqueous solution was directly used as an aqueous monomer. The operations were all conducted at room temperature.

In base screening experiments, different bases were added into 3 wt% PHMB solution to adjust the pH. IP operation was the same as described above.

For the PHMG composite membrane, the operation was the same as above.

Conductivity and pH of the solution were separately measured using a conductivity meter (FE38-Meter) and a pH meter (FE28-Standard) from Mettler-Toledo Instruments Ltd. The thickness of the PSF membrane was measured using a helical micrometer, and the pore size of PSF was measured using an automatic mercury porosimeter (AutoPore IV 9500, USA). The porosity was calculated using the following equation: ɛ (%) = 100 × (1 − ρ_Hg/ρ_He), where ɛ refers to the porosity, ρ_Hg and ρ_He refer to the density of mercury and helium. The concentrations of Mg²⁺ and Li⁺ were determined by ICP-OES (PerkinElmer 8300). Surface chemistry of the composite membrane was analyzed by attenuated total reflectance infrared (ATR-IR) spectroscopy by Nicolet IS50 infrared spectrometer (Thermo Fisher, USA). X-ray photoelectron spectrometry (XPS) analyses was graphed through a Thermo Fisher K-alpha spectrometer using focused monochromatized Al Kα radiation. An S-4800 field emission scanning electron microscope (FE-SEM, Hitachi, Japan) was used to observe surface and cross-section morphology of the membranes. The samples were sprayed with gold powder before testing. Average roughness (Ra) of the membrane surface was analyzed by atomic force microscopy (AFM, Bruker, Icon, USA) in a tapping mode. Dynamic water contact angle (DWCA) of the membrane surface was measured using a Drop Shape Analysis 100 contact angle meter (Kruss BmbH Co., Germany). Membrane surface charge properties were characterized by streaming zeta potential measurement containing a SurPASS electrokinetic analyzer (Anton Paar, Austria) at pH 2–10. Measurements were carried out with a 0.01 mol·L⁻¹ KCl solution as the feed at 25 ± 1.0 °C; 0.01 mol·L⁻¹ hydrochloric acid and sodium hydroxide solutions were used to adjust the pH of the solution.

Antibacterial properties of the membranes were evaluated by shake flask tests using E. coli (E.C.) as the model Gram-negative bacteria and S. aureus (S.A.) as the Gram-positive model (Hou et al. 2009). The area of tested membrane was 5 × 10⁻⁴ m². The tests were operated according to the literature (Meng et al. 2015).

The home-made base membrane was measured first. Its thickness was 140 μm and the average pore size was 120.2 nm with 56.10% porosity, see supplementary figure (Fig. S1).

Figure 1 shows the chemical structures of PHMG and PHMB, which were oligomeric products of hexanediamine with ammonia and sodium dicyanamide, respectively. They were polycations with an average of 12–16 biguanide groups spaced by hexamethylene segments. Solution pH of 3 wt% PHMG and PHMB were 8.9 and 7.7, respectively. Both of them could not provide satisfying rejection of MgCl₂, when they were used as aqueous monomers alone in the NF composite membranes preparation (Figure 2(a)). A 3 wt% PHMB solution was firstly chosen to optimize the aqueous phase formulation and then the PHMG membrane was prepared according to the optimized PHMB recipe. Pure PHMB solution prepared a membrane for 21.0% MgCl₂ rejection as the reactivity of its hydrochloride was very low; at pH values far lower than 11.8, compared with the value of the 3 wt% PIP solution. Five alkalines were investigated as pH regulators of aqueous phase. Overall, the rejection of membranes made from organic base displayed better performance than those from inorganic ones. NaOH and ammonium hydroxide provided very poor adjustment. PIP gave the best assistance because it not only regulated pH but also participated in the IP reaction to build compact cross-linking network. As PIP was added (Figure 2(b)), the rejection to MgCl₂ increased. At pH up to 11, MgCl₂ removal rate arrived at the apex, 99.6%. The mass ratio of PHMB to PIP was 6:1 at this point. As PIP increased, the rejection gradually decreased as high concentration of aqueous monomer did not match the 0.2% TMC concentration needed to prepare a compact network. Therefore, PIP was used as a regulator and the aqueous pH value was fixed at 11. According to the recipe of PHMB, pure MgCl₂ removal rate by PHMG–PIP membrane (G-1) was 97.3% while the value was 54.7 using pure PHMG membrane, supplementary table (Table S1).

Adjusted by PIP, both PHMB and PHMG membranes obtained ideal removal rates for MgCl₂. Thereafter, the structures of these two composite membranes were characterized and their performances were further investigated in water treatment applications. Surface chemistry of the membranes was analyzed by ATR-IR and XPS (Figure 3). Through ATR-IR spectra (Figure 3(a)), it was confirmed that there was a composite layer formed on the PSF support. The broad peaks at 3,400 cm⁻¹ were -N-H stretching in G-1, B-1 and M-PIP curves. The peak sizes were positively correlated with the numbers of -N-H groups. Guanidine-containing surfaces obviously possessed larger areas than M-PIPs. Peaks at 2,930⁻¹ of G-1 and B-1 were -CH₂- stretching in hexamethyl groups. The small peaks at 2,168 cm⁻¹ were attributed to the terminal -C ≡ N group of PHMG and PHMB chains. The doublets at 1,623 cm⁻¹ were -C = O stretching in amide groups, which indicated that polyamide layers were successfully prepared on PSF membranes. The order of peak sizes was B-1 > G-1 > M-PIP, which illustrated that there were more amide groups in guanidine membrane surfaces than for M-PIPs. In addition, B-1 contained more amide bonds as biguanidine contained more -N-H groups than mono-guanidine. Therefore, the surface compactness was B-1 > G-1 > M-PIP. It was also confirmed by MgCl₂ filtration test.

Figure 3(b)–3(d) shows C1s high resolution of XPS curves of membrane surfaces. The insert in Figure 3(b) shows the element scanning spectra of membranes. Oxygen element contents in G-1 and B-1 were smaller than M-PIP due to the introduction of polyguanidine. Both G-1 and B-1 possessed peaks at 286.2 eV (Figure 3(c) and 3(d)), which were the terminal -C ≡ N group in polyguanidines (East et al. 1997). Peaks at 288 eV were -C = O bonds in the polyamide network. Their areas were in the order of B-1 (69,086.8) > G-1 (62,522.6) > M-PIP (57,992.3), which corresponded with ATR-IR data.

The hydrophilicity of membrane surface was evaluated by surface DWCA and zeta potential. Figure 4(a) shows that DWCAs of composite membrane were all smaller than for the PSF membrane. It illustrated that more hydrophilic skins were formed on the top of the supporting layer. The initial DWCAs of G-1 and B-1 were 63° and 55° but they declined to 20° with time with a slightly different developing trend. DWCAs of polyguanidine membranes were smaller than M-PIPs, which demonstrated that the introduction of polyguanidine provided higher hydrophilicity for the surfaces. Figure 4(b) shows that B-1 and G-1 displayed properties of positve-charged membranes. Biguanidine introduced more positive charges to the membrane surface. The isoelectric points of B-1, G-1 and M-PIP were pH 8.86, 7.82 and 4.84, respectively. Polyguanidine, as polycations, introduced large amounts of NH groups to the membrane surface, which dissociated into or while the classic M-PIP was a negative charged membrane.

Micro-morphology of membranes was observed using FE-SEM and AFM (Figure 5). Through IP processes, PSF films were coated with composite layers. G-1 and B-1 did not have a typical nodular skin like the M-PIP membrane. There were irregular granular particles on the G-1 surface, while B-1 had a very smooth skin. From cross-section images, the thicknesses of PA layers were 200, 200 and 160 nm for M-PIP, B-1 and G-1, respectively. Surface roughness of composite membranes were higher than for PSF layers. The order of Ra values was M-PIP (12.00) > G-1(8.01) > B-1(6.93), which corresponded with the images observed by FE-SEM. Nodular size decreased as surface roughness and hence particles on B-1 was the smallest. Polyguanidines were amphiphilic oligomers, which formed self-associated aggregates of 5 and 7 nm in aqueous phase, supplementary figure (Fig. S2). In these micelles, hydrophilic NH groups tended to distribute over the surface. This decreased migration speed from the aqueous phase to hexane during the IP process, which finally led to smoother surfaces (Li et al. 2014b; Jin et al. 2015). Biguanidine micelles possessed more hydrophilic surfaces as there were more NH groups in its structure.

As positve-charged membranes, B-1 and G-1 displayed improved capacity for metal cations removal. In order to further investigate the membranes, filtration performances of B-1 and G-1 were evaluated through a series of metal salt filtration tests. Figure 6(a) provides water flux in filtration tests under 0.5 MPa. It can be seen that the PWF of M-PIP was lower than B-1's and higher than G-1's. M-PIP SWF of the MgCl₂ solution was 12.97 L m⁻² h⁻¹. The SWF of the polyguanidine membranes were smaller than their PWF as there was osmotic pressure from the opposite direction to resist water permeating through the membrane. Either PWF or SWF of B-1 was larger than G-1. This indicated that B-1 had more cations on its surface, which provided a stronger repulsion to metal cations and led to the decrease in salt flux. Under a fixed pressure, the total flux of salt and water was fixed. As a result, the water flux of the salt solution increased although B-1 had a more compact surface.

The salt rejections by B-1 and G-1 are shown in Figure 6(b). B-1 displayed better salt removal performance than G-1 in all salt filtration. Except for LiCl, the retention rates were higher than 95%. This could be attributed to more positive charges on B-1 that provided greater repulsion to metal cations, especially multivalent cations. G-1 provided larger retention differences between multivalent and monovalent cations. The rejection of G-1 for divalent salts was all higher than 90%, while for sodium salts, Na₂SO₄ and NaCl, the values were 80.6% and 81.3%, respectively. The rejection of LiCl was only 67.0% as Li⁺ was a monovalent cation and its hydration radius was comparatively small. The order of retention of B-1 to salts was: MgSO₄ > MgCl₂ > Na₂SO₄ > Cd(NO₃)₂ > Pb(NO₃)₂ > NaCl > LiCl. The rejection order of G-1 to salts was: R(MgSO₄) > R(MgCl₂) > R(Cd(NO₃)₂) > R(Pb(NO₃)₂) > R(NaCl) > R(Na₂SO₄) > R(LiCl). It has been well known that classic M-PIP showed poor removal capacity for hazardous metal cations, such as Cd²⁺ and Pb²⁺, because of their strong complex with the negative charged surface. However, the complex would be resisted to some degree if positive electrostatic repulsion from the membrane surface was strong enough. Thereafter, the membrane could obtain the capacity to remove hazardous heavy metal cations from polluted water. To prove this, filtration tests of a 1 g·L⁻¹ Pb(NO₃)₂ solution by B-1 and G-1 were performed (Figure 6(c)). Water flux of both membranes faintly decreased from 14.9 L·m⁻²·h⁻¹ to 14.4 L·m⁻²·h⁻¹ for B-1 and from 13.5 L·m⁻²·h⁻¹ to 13.1 L·m⁻²·h⁻¹ for G-1 within a week. Pb(NO₃)₂ retention by B-1 slightly decreased at the beginning and then remained above 92.1%. The interception of G-1 decreased from 93.6% to 91.7%. The antifouling capacity of membranes was investigated through flux recovery tests of alternate Pb(NO₃)₂ filtration and water rinsing. FRR (Figure 6(d)) of B-1 and G-1 were 93.59% and 92.48%, respectively. XPS spectra showed that the chemistry of both membranes had not changed obviously after filtration of the Pb(NO₃)₂ solution, supplementary figure (Fig. S3). This confirmed that their surfaces were not polluted with heavy metal cations.

Owing to the large surface repulsion from the positive charge membrane, cation concentration over the surface area was lower than the bulk solution. Since Mg²⁺ suffered larger repulsion, the concentration over surface area (⁠⁠) should be smaller than for the bulk solution (⁠⁠). Comparably, Li⁺ had low co-ion charge, which made it have smaller repulsion. It was easier for Li⁺ to diffuse to the surface area to keep cation concentration homogeneity and charge equilibrium in the whole solution. As a result, (Li⁺ concentration on the membrane surface) was higher than (Li⁺ concentration in the bulk solution). Hence, the ratio of Mg²⁺ to Li⁺ over the membrane surface was smaller than in the bulk solution. Furthermore, Li⁺ had smaller hydrated radii than Mg²⁺, which made it easier to be ‘pushed’ through the membrane to maintain electrostatic neutrality equilibrium under Donnan exclusion and steric hindrance mechanisms. More Li⁺ permeating through the membrane resulted in a downstream concentration increase.

Long-time Mg²⁺/Li⁺ separation test was carried out to investigate the enrichment performance of Li⁺ by G-1 using 3.6 g·L⁻¹ MgCl₂ and LiCl mixed solution and the result was graphed as shown in Figure 7. In 80 h, the rejection rate of Mg²⁺ and Li⁺ decreased together to different degrees. Mg²⁺ retention slightly decreased from 92.4% to 83.7%, while Li⁺ rejection declined from −3.6% to −21.5%. The separation factor of Mg²⁺/Li⁺ began to fall from 14.1 to 8.7 after 15 h and then remained steady.

PHMB and PHMG are mainly used as bactericides in preservatives or disinfectant. Therefore, the antimicrobial performance of B-1 and G-1 was investigated through colony formation experiments using Gram-negative E.C. and Gram-positive S.A bacteria as models. Figure 8 shows the digital photographs of dishes containing bacteria-inoculated agar. Both B-1 and G-1 exhibited excellent antibacterial capacities. Bactericidal rates of membranes were higher than 99% even though the culture liquid had been diluted 1,000 times, supplementary figure (Fig. S4). SEM images showed that the cell walls of E. C. and S. A. were destroyed after contact with guanidine membranes, supplementary figure (Fig. S5). The result illustrated that polyguanidine chains still displayed excellent antibacterial ability even though they were fixed in a cross-linking network. Guanidine groups could dissociate into positive charges through partially forming amide groups, which ensured the biocide capacity. It was the polyguanidine micelle, that participated in the IP reaction but not single free molecules. There were still many unreacted NH groups left to dissociate into or to play a role in sterilization.

PHMB and PHMG are polycations. Under the assistance of PIP, they were successfully introduced onto the surfaces of the NF composite membrane through a IP reaction and prepared the positve-charged membranes. The membranes displayed excellent performance in multivalent metal salts removal from water. In particular, the long-time rejection of Pb(NO₃)₂ was higher than 90%, which provided the potential for hazardous heavy metal cation removal. Polyguanidine membranes showed moderate rejection of monovalent metal cations and retention of LiCl was only 67% by G-1. In the Mg²⁺/Li⁺ mixture filtration, G-1 decreases the Mg²⁺/Li⁺ ratio from 60 to 3.7 accompanied with a negative Li⁺ rejection of −3.6. It exhibited the possible application in Li⁺ enrichment. In addition, both B-1 and G-1 showed excellent bactericidal performance in E.C. and S.A. tests as plenty of NH was left to dissociate into or after the IP process. Polyguanidine is a species of promising oligomers in water treatment membrane preparation, including mono-/di-valent cation separation and hazardous heavy metal cation removal.

The authors acknowledge the financial support from the National Natural Science Foundation of China (Grant Nos. 22075206, 21875162) and Tianjin Key Projects of New Materials Science and Technology (17ZXCLGX00050).

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