Asymmetric Membrane Capacitive Deionization Using Anion-Exchange Membranes Based on Quaternized Polymer Blends

Membrane capacitive deionization (MCDI) for water desalination is an innovative technique that could help to solve the global water scarcity problem. However, the development of the MCDI field is hindered by the limited choice of ion-exchange membranes. Desalination by MCDI removes the salt (solute) from the water (solvent); this can drastically reduce energy consumption compared to traditional desalination practices such as distillation. Herein, we outline the fabrication and characterization of quaternized anion-exchange membranes (AEMs) based on polymer blends of polyethylenimine (PEI) and polybenzimidazole (PBI) that provides an efficient membrane for MCDI. Flat sheet polymer membranes were prepared by solution casting, heat treatment, and phase inversion, followed by modification to impart anion-exchange character. Scanning electron microscopy (SEM), atomic force microscopy (AFM), nuclear magnetic resonance (NMR), and Fourier-transform infrared (FTIR) spectroscopy were used to characterize the morphology and chemical composition of the membranes. The as-prepared membranes displayed high ion-exchange capacity (IEC), hydrophilicity, permselectivity and low area resistance. Due to the addition of PEI, the high density of quaternary ammonium groups increased the IEC and permselectivity of the membranes, while reducing the area resistance relative to pristine PBI AEMs. Our PEI/PBI membranes were successfully employed in asymmetric MCDI for brackish water desalination and exhibited an increase in both salt adsorption capacity (>3×) and charge efficiency (>2×) relative to membrane-free CDI. The use of quaternized polymer blend membranes could help to achieve greater realization of industrial scale MCDI.


INTRODUCTION
The availability of potable water is one of the world's most pressing global challenges. Currently, more than 2 billion people live in countries experiencing high water stress, and up to 4 billion people experience severe water scarcity for at least 1 month each year. 1 Desalinationthe removal of salt from saline watercan help to relieve water stress in some areas of the world. Indeed, various countries have installed industrialscale desalination plants to augment existing water resources. 2 However, existing desalination processes such as multistage flash distillation, multieffect distillation, and mechanical vapor compression all suffer from the need for high electrical, mechanical, or thermal energy input leading to high energy consumption and capital cost. Reverse osmosis (RO) is generally considered the benchmark desalination technology in terms of salt removal and energy efficiency; however, it also relies on the use of high pressures to generate pure water. 3 Reducing the energy input and capital costs of desalination has therefore attracted huge research interest in recent years.
Capacitive deionization (CDI) has recently emerged as an environmentally friendly and energy efficient route to desalination of brackish water supplies. 4 Although at present CDI is not competitive with RO as a stand-alone technology, it has shown promising results to improve the efficiency of hybrid desalination processes. 5 Conventional CDI operation passes a feed salt solution between two porous electrodes. The exploitation of the ionic charge of the solute allows the ions to be electrosorbed in electric double layers (EDLs) in the electrode pores, by application of a voltage between the two electrodes with opposite polarity. 6 The electrodes are then regenerated by reversing or zeroing the voltage. Electrode architectures based on activated carbon, 7 carbon aerogels, 8 carbon nanotubes (CNTs), 9 and graphene 10 have been extensively studied for electrosorptive ion removal in CDI, owing to their high specific surface areas for ion adsorption. Electrode materials can also be "faradaic" in nature; such electrodes have the potential to enhance ion removal due to rapid charge-transfer (redox processes) at the electrode's surface. Transition metal dichalcogenides (TMDs) 11 and MXenes 12 have both been employed as CDI electrodes for desalination purposes, while redox-polymer electrodes such as poly(vinyl)ferrocene/carbon nanotubes (PVF-CNT) have shown great success in electrochemically driven separations of arsenic 13 and organic pollutants 14 from aqueous solutions.
Alternatively, enhanced CDI performance can be achieved through the use of ion-exchange membranes over the electrodes. Typical membrane capacitive deionization (MCDI) utilizes anion-and cation-exchange membranes over the anode and cathode, respectively. Such membranes allow counterions to pass through the electrode/membrane without obstruction, whereas co-ions (ions of the same polarity as the electrode) are blocked. Co-ion expulsion from electrodes can deplete the salt removal and energy efficiency of an (M)CDI system. 15 Ion-selective membranes can reduce co-ion effects, while simultaneously protecting electrodes from potentially damaging parasitic reactions. 16 A study found that MCDI has the potential to be more energy efficient than RO for water salinities of less than 2 g L −1 . 17 However, the lack of suitable and inexpensive commercial membrane materials means that practical use of MCDI is seldom realized.
Ideal ion-exchange membranes for MCDI should possess high ion-exchange capacity (IEC) and high permselectivity, combined with a low area resistance and material cost. The reduction of material and operational costs has prompted the use of asymmetric MCDI, which employs anion-exchange membranes (AEMs) over the anode only. Indeed, protection of the anode has been shown to be pertinent to long-term MCDI operation due to oxidation reactions degrading the anode. 18 Successful brominated poly(phenylene oxide) (b-PPO) AEMs have been fabricated for MCDI, with reported high desalination performance and high cycling stability. 19 Similarly, AEMs for MCDI were fabricated by polymerization of 4-(chloromethyl)styrene (CMS) monomer in a porous polyethylene substrate. These AEMs achieved a high salt adsorption capacity (16.1 mg g −1 ) when utilized in MCDI combined with a commercial (CMX) cation-exchange membrane. 20 Cross-linked quaternized poly(vinyl alcohol) (q-PVA) AEMs with high IEC have been prepared for MCDI, yielding a salt adsorption capacity of 15.6 mg g −1 with a NaCl concentration of 800 mg L −1 . 21 Nanocomposite AEMs for MCDI based on reduced graphene oxide/polyani-line have also been fabricated, displaying a maximum sorption capacity of 1.56 mg g −1 . However, these nanocomposite membranes relied on the use of expensive reduced graphene oxide nanofiller material. 22 In this study, asymmetric MCDI with an AEM over the anode was studied. This allowed us to quantify the degree of improvement, due to the AEMs, on the performance relative to CDI, without the contribution of a CEM.
Here, we describe a simple fabrication procedure ( Figure 1) for polymer blend membranes based on polybenzimidazole (PBI) and branched polyethylenimine (PEI). In this procedure, these polymers are quaternized postcasting to impart anion-exchange characteristics. Both of these polymers possess an abundance of aliphatic and aromatic amine groups that are amenable to quaternization to produce quaternary ammonium cations within the membranes, resulting in high IEC. 23 PEI has been employed as a membrane filler in poly(ether sulfone) (PES) for the removal of organic dyes in water treatment by nanofiltration, 24 as well as being used as an additive to carbon nanotube (CNT) electrodes in CDI. This results in improved performance over conventional CDI and that of commercial anion-and cation-exchange membranes. 25 We hypothesized that the introduction of quaternized PEI embedded in a quaternized PBI matrix would increase IEC, reduce ionic resistance, and produce AEMs with a high permselectivity to transport chloride (Cl − ) ions. Increased IEC is usually accompanied by an increase in swelling which reduces resistance but simultaneously reduces permselectivity as more solution is absorbed by the membrane. This effect can be limited by reducing the film thickness; the ionic resistance is reduced due to the shorter distance across which ions must diffuse, while permselectivity is not sacrificed due to the smaller volume of membrane which undergoes swelling. 26 As a result, the thickness of our membranes (30−40 μm) was reduced by up to 70% compared to commercially available Neosepta AEMs (100−150 μm). 27 We proposed that these factors would help to maintain high ion-exchange and permselective properties of the membranes, facilitating (Cl − ) transport across the membranes into the electrode pores and increasing the salt removal and energy efficiency of the MCDI systems.
Recently, thin-film composite membranes for organic solvent nanofiltration were synthesized by cross-linking PEI onto PBI substrates. 28 In another study, a nanophase separation structure was created in membranes for vanadium flow batteries by grafting PEI onto a PBI backbone. 29 However, to our knowledge, no results have been reported for the use of PBI/PEI composite polymer blend AEMs for desalination applications. Here, we report on the fabrication of ACS Applied Polymer Materials pubs.acs.org/acsapm Article a highly quaternized polymer blend based on PBI/PEI and its application as an AEM in MCDI for water desalination. We describe the full characterization of membrane structure and the determination of their electrochemical properties. We quantified the MCDI performance in terms of salt adsorption capacity and charge efficiency and compared our results with conventional CDI (without membranes) and relevant MCDI studies.
2.2. Membrane Preparation. Table 1 provides the compositions of casting solutions used to prepare membranes. Membranes of PEI content up to a maximum of 20 wt % were fabricated, ensuring nitrogen content (PEI amine groups) could be varied while maintaining the same material processing steps. First, a given amount of PEI and PBI dope solution (26 wt % in DMAc) were separately added to a vessel. A total of 3 g of polymer was used to prepare all membranes. The dope solution was further diluted to the required polymer wt % with DMAc solvent. The mixture was sonicated for 30 min to dissolve PEI into the solvent, followed by mixing with an overhead stirrer (6 h, 60 rpm) to homogenize the mixture. The dope solution was then rolled (1 h, 60 rpm) and degassed for 12 h using an incubator/shaker (40°C, 300 rpm). The dope solution was cast at 40°C on a glass plate to a thickness of 200 μm using an Elcometer 4340 Film Applicator. The membrane was placed in an oven (60°C, 24 h) to evaporate solvent and promote the formation of a dense film layer. The membranes were precipitated by immersion in DI water (0.008 L cm −2 ) resulting in the membrane detaching from the glass plate.
2.3. Quaternization of Membranes. Quaternization of the prepared membranes was carried out via methylation after membrane casting, as adopted in our previous work. 31 Membrane samples (area 625 cm 2 ) were rinsed with acetonitrile and placed into a glass pressure tube that had been filled by Ar. The membrane was submerged in acetonitrile (60 mL), and DIPEA (1.4 mL, 8 mM) was added to the solvent, followed by purging with Ar for 5 min. Iodomethane (MeI, 2 mL, 32.1 mM) was added, and the pressure tube was sealed with an O-ring. The solution was heated at 65°C under constant stirring for 24 h. The membranes were rinsed sequentially with water and acetonitrile and stored in water/acetonitrile mixture to avoid bacterial growth. Anion substitution (iodide to chloride) was carried out by immersing the membranes into a brine solution for 24 h (40 g L −1 ) to condition the membrane for desalination applications.
2.4. Electrode Preparation. The porous carbon electrodes used in this study were prepared from a slurry containing activated carbon powder (YEC-8A, 80%, Fuzhou Yihuan Carbon Co), carbon black conductive additive (10%), and polyvinylidine fluoride (PVDF) binder (10%), in N-methylpyrrolidone (NMP). The slurry was coated onto a graphite foil current collector (99% Purity, Gee Graphite Ltd.) and baked in a vacuum oven at 80°C for 12 h to evaporate the solvent ( Figure S1). The prepared electrodes had an approximate area of 10 mm × 10 mm and had active masses of between 0.01 and 0.04 g.
2.5. Membrane Characterization. Surface and cross-sectional morphologies of the membranes and electrodes were provided by scanning electron microscopy (SEM) images. SEM images were obtained using an FEI Quanta 250 ESEM instrument. Prior to imaging all membrane samples were sputtered with platinum coating to add conductivity and image resolution. Membrane thicknesses were determined using ImageJ software, taking the average of three measurements over the membrane cross-section. Atomic force microscopy (AFM) topography images were gathered using a Bruker Multimode 8 microscope in tapping mode. Images were analyzed using the Nanoscope Analysis 1.9 software and surface roughness, R a , was calculated as an average over a 10 μm × 10 μm image. 1 H nuclear magnetic resonance (NMR) spectra were obtained using a 400 MHz Bruker Avance III spectrometer. Chemical shifts are given in ppm relative to tetramethylsilane (0 ppm). All spectra were obtained in DMSO-d 6 and referenced to the residual solvent peak (2.500 ppm). Fourier-transform infrared (FTIR) spectra were recorded using a Bruker Alpha-P ATR-FTIR spectrometer. Elemental analysis was conducted using a dry combustion method of membrane samples, performed with a CHN-elemental analyzer equipped with a halide titrator. Prior to NMR, FTIR and elemental analysis membrane samples were dried under vacuum (∼10 mbar) for 24 h to remove any solvent and unreacted reagent traces.
The structural stability of the polymer blend membranes was tested by immersion of the membranes in the electrolyte utilized during this study (200 mg L −1 NaCl in water). In brief, membrane pieces (1 cm width ×3 cm length) were cut and immersed into the electrolyte for a period of 14 days. The weights of dry membrane pieces were recorded before and after immersion and the percentage weight loss in the electrolyte were calculated by m m m weight loss in electrolyte (%) 100% d,before d,after Ion-exchange capacity (IEC) of the AEMs was determined in triplicate using Mohr's method. 32 Membrane pieces (chloride form) were dried in a vacuum desiccator for 24 h and the dry mass of the pieces was recorded. The pieces were subsequently submerged in KNO 3 aqueous solution (40 mL, 1 M) for 24 h to allow chloride ions to fully exchange into the solution. The membranes pieces were washed with DI water which was mixed with the solution. Five drops of K 2 CrO 4 indicator were added to the analyte and titrated against AgNO 3 (0.01 M). After all chloride ions had precipitated to form a white precipitate of silver chloride (AgCl), excess Ag + ions proceeded to form a dark red silver chromate precipitate (Ag 2 CrO 4 ), signifying the end-point of the titration. 33 The IEC (mmol g −1 ) of the membranes was subsequently determined by where c Ag is the concentration of silver ions in AgNO 3 , V S is the titration volume of AgNO 3 , and m d is the dry mass of the membrane. The quaternization degree (QD) was estimated by comparison of the experimental (IEC Exp ) and maximum theoretical ion-exchange capacity (IEC Theory ) given by where (IEC Theory ) is derived from the mass percentage of counterions in the membrane as determined by elemental analysis. This is given by where M I is the molar mass of iodide ion (126.9 g mol −1 ).
Permselectivity measurements were determined from chronopotentiometric curves. The membranes were conditioned by immersion into 0.1 M NaCl solution in DI water for 24 h. The membrane was fixed in a sandwich setup with a circular area of 0.19 cm 2 ( Figure S2). Platinum mesh electrodes (1 cm × 3 cm) were used as working/ counter electrodes and Ag/AgCl were used as reference/sensing electrodes. Both beakers contained equimolar NaCl solution (0.1 M, 50 mL) and were stirred continuously throughout the experiment.
Chronopotentiometry measurements were carried out with a potentiostat (Autolab, Metrohm) in the galvanostatic mode using an overlimiting current of 10 mA. The transition time, τ was determined by the inflection point of the chronopotentiometry curve and substituted into the modified Sand equation to calculate the permselectivity, P, given by where |z| is the absolute valence of the chloride ion, F is the Faraday constant (96 485 C mol −1 ), D is the diffusion coefficient (1.48 × 10 −5 cm 2 s −1 ), t i is the transport number (0.61) of the chloride ion in 0.1 M NaCl, I is the applied current (10 mA), and A is the exposed area (0.19 cm 2 ). 34 The values were calculated as an average of three membrane pieces. Membrane area resistance values were obtained using the sandwich setup depicted in Figure S2 combined with electrochemical impedance spectroscopy (EIS), with equimolar NaCl (1 M, 50 mL) in each beaker. The measurements were carried out using a potentiostat (Metrohm, Autolab) integrated with frequency response analyzer (FRA32M). Platinum mesh electrodes were again used as the working/counter electrodes, and Ag/AgCl electrodes as the reference/sensing electrodes. Impedance measurements were carried out over 50 scans using an AC signal of 0.1 mA and within a frequency range of 1 mHz to 1 kHz. 35 The resistance of the membrane immersed in 1 M NaCl solution (R MS ) was obtained as the impedance of the system. The value was calculated using equivalent circuit fitting ( Figure S3) of Nyquist plots (real vs. imaginary impedance) over the specified frequency range. These resistance values were corrected by the resistance of the solution only (R S ) using a cell with no membrane separating the two solution compartments. The area resistance (R A ) values of the membranes were subsequently calculated by multiplying by the exposed membrane area, A: The water contact angle was measured at room temperature in air using a Kruss drop shape analyzer (DSA 100 Instrument). Membrane samples were dried in a vacuum desiccator for 24 h prior to the measurements. Droplets were formed on the membrane surface using DI water with a droplet volume of 1.5 μL. The contact angles are reported as an average of three droplets. Water uptake and linear swelling ratio measurements were determined in triplicate in the chloride form of the membrane. All membrane pieces were submerged in DI water for 24 h and the wet mass (m w ) and length (L W ) were recorded. The samples were dried in a vacuum desiccator for 24 h and the dry mass (m d ) and length (L d ) were recorded. The water uptake and linear swelling ratio as a percentage were determined by the following equations: 2.6. MCDI Desalination Tests. All prepared membranes were used as AEMs in the MCDI cell. The MCDI cell comprises two Perspex end-plates, two porous carbon electrodes coated onto graphite current collectors and the AEM covering the anode. The two electrodes were separated by an acrylic spacer (width = 2 mm) to open up a flow channel and prevent short-circuiting. A schematic of the asymmetric MCDI cell, including cell dimensions, is depicted in Figure S4. MCDI desalination experiments were carried out on a bench-scale flow-between system as depicted in Figure 2. Membranes were conditioned in electrolyte (NaCl, 200 mg L −1 ) for 24 h prior to desalination tests. All desalination tests were performed in batchmode at T = 25°C whereby the effluent stream was recycled back into the feed reservoir. 60 mL of feed salt solution (200 mg L −1 ) was pumped continuously through the system using a peristaltic pump (VWR) at a flow rate of 5 mL min −1 . Voltages of 1.2 and 0.0 V were used for the adsorption and desorption cycles respectively, with each cycle lasting for 1 h. Systems were then recycled using 2 h adsorption (1.2 V)/desorption (0.0 V) cycles, to confirm stability of the MCDI system over an extended time period (+30 h). Cell voltages were applied using a potentiostat (Metrohm, Autolab) and desalination was measured by a conductivity meter (Seven Excellence, Mettler Toledo) in the feed reservoir. Duplicate measurements were taken for all membrane designations and conversion from conductivity (μS cm −1 ) to salt concentration (mg L −1 ) was done by a calibration plot ( Figure  S5).
The desalination performance of the system was evaluated by the salt adsorption capacity (SAC), charge efficiency (Λ) and salt adsorption rate (SAR). The energy normalized to salt adsorption (ENAS) is an important parameter defining the amount of salt removed per unit energy. These metrics are defined by  ACS Applied Polymer Materials pubs.acs.org/acsapm Article reported; whereby PBI/PIM-1 composites were ion-stabilized by amine protonation via treatment with HCl. 36 Table 2 lists structural properties of fabricated membranes. Thickness and roughness parameters were determined via ImageJ and Nanoscope software, respectively. All polymer blend membranes showed an increase in thickness compared to the Q0 (PBI only) membrane. This is attributed to an increased mass loading of PEI into the original casting solution, given that an identical air gap (200 μm) was used to cast all membranes. AFM images show a gradual increase in surface roughness for membranes Q0 to Q20. The average arithmetic roughness values (R a ) were calculated over a 10 × 10 μm AFM image. The increasing surface roughness has potential to increase the hydrophilicity of membranes according to Wenzel theory; attributed to an increase in the solid/liquid contact area on the membrane surface. The PEI domains embedded within the PBI matrix create a rougher surface with undulations, increasing the surface area compared to the pristine Q0 membrane. This effect, combined with the hydrophilic nature of PEI compared to hydrophobic PBI, can create a surface with enhanced wetting properties. 37 This could improve the initial uptake of chloride ions and subsequent transport through the AEMs. The SEM and AFM images suggested successful embedding of PEI within the PBI backbone. Further, all membranes showed good structural stability after immersion tests in the studied electrolyte (NaCl, 200 mg L −1 ) after a period of 14 days; with all membranes retaining +95% of their original mass after immersion. This inferred that the membranes retained excellent confinement of PEI within the blended matrix and stability of the membrane structure after extended exposure to saline water.

Membrane Quaternization.
Anion-exchange characteristics was imparted to the membranes by chemical modification via a nucleophilic methylation (S N 2) reaction. In PBI, methylation occurs by substitution of methyl groups onto the aromatic nitrogen on the imidazole ring to form imidazolium cationic moieties throughout the PBI backbone. 38 Likewise, PEI contains an abundance of primary, secondary and tertiary amine groups that can undergo quaternization. Figure 4a) displays 1 H NMR spectra for all AEMs in DMSO-d 6 solvent. The signals between 7.5 and 8.5 ppm can be assigned to aromatic protons within the imidazole/imidazolium rings. The characteristic signal showing the N−H proton on the imidazole ring for PBI is found at 13.2 ppm. 39 This signal is no longer distinguishable in any of the quaternized membranes Q5−Q20; indicating that most of the N−H protons in PBI were methyl substituted during alkylation or otherwise deprotonated. For all quaternized membranes a new multiplet signal at 4.1 ppm was observed. This signal is well-known and corresponds to the methyl protons substituted onto the nitrogen of benzimidazolium. 40 These assignments indicate successful quaternization of the PBI backbone in the AEMs.
Additional 1 H NMR signals imply successful PEI incorporation and quaternization. Typical signals at 2.3−3.0 ppm can be observed at different intensities for all of Q5, Q10, and Q20, which are not apparent in the PBI and Q0 membranes. These can be assigned to methylene (−CH 2 −) protons in the aliphatic branched PEI chains. 41 Further signals at 0.8−1.2 ppm match well with literature values assigned to quaternized polyethylenimine after alkylation with a methyl group. 42 Further information on the quaternization of membranes was gathered via elemental analysis ( Table 3). The iodide content gives the percentage by mass of counterion within the membrane after methylation. A higher iodide content indicates a higher level of quaternization and a higher theoretical ion-exchange capacity (eq 4) of the membranes. Q0 has the lowest iodide content (23%) of all the prepared membranes. The iodide content subsequently increases with increasing PEI content in the blend AEMs. However, only a slight increase was observed upon doubling the mass loading between Q10 (35%) and Q20 (36%). This implied that nitrogen groups became less accessible to quaternization at higher mass loadings of PEI; as PEI domains become shielded by neighboring PBI−PEI groups to a greater extent. Nevertheless, the NMR and elemental analysis data demonstrate that with higher amounts of PEI in the PBI matrix, this can increase the level of counterions and the theoretical IEC of the membranes. This should result in increased Cl − removal and salt adsorption capacity when quaternized membranes are used as an AEM in an MCDI system. Figure 4b) displays FTIR spectra of all fabricated membranes and pristine PBI and PEI polymers. A broad transmittance peak can be observed (3000−3500) cm −1 for all membranes, corresponding to the N−H stretching on the imidazole rings and primary amine groups in PEI. The peak found at 1620 cm −1 can be assigned to the CN stretch on the imidazole ring. Spectra for Q10 and Q20 membranes display two peaks in the region (2850−2950) cm −1 . These peaks are characteristic of aliphatic C−H stretching vibrations and have been previously reported as due to the addition of a greater number of alkyl groups in a quaternized membrane. 43 Due to the higher loading of PEI, the C−H vibration would become more intense as a result of higher number of methyl substituents after quaternization. The complementary NMR, elemental analysis and FTIR data indicate successful quaternization of PBI and PEI, and these data suggest that quaternized PEI can introduce additional counterions into the membrane.
3.3. Hydrophilicity of Membranes. Hydrophilicity and wettability properties of the AEMs were determined by analysis of the contact angles and water uptake of membrane pieces (Figure 4c). Good hydration properties of membranes can positively affect ion transport; however, excessive swelling can hinder certain electrochemical properties of the membranes such as permselectivity. 44 A lower contact angle is representative of a membrane's hydrophilic nature due to a reduction in the surface tension between the water/membrane interface. 45 Q0 and Q5 displayed the most hydrophobic properties with contact angles of 62°and 63°, respectively. This hydrophobicity is reinforced by water uptake values of 19% (Q0) and 16% (Q5). Usually, the introduction of

ACS Applied Polymer Materials
pubs.acs.org/acsapm Article hydrophilic PEI domains into hydrophobic PBI is expected to decrease contact angle while increasing water uptake. These results suggest that the addition of PEI in small quantities (5 wt %) had no marked effect on the hydrophilic properties of the blend membrane. Increasing the mass loading of PEI resulted in a discernible drop in the contact angle to 53°( Q10) and 50°(Q20) and a simultaneous increase in water uptake of 26% (Q10) and 34% (Q20). Promisingly, these values of water uptake match closely to the typical values of Neosepta-AMX (25−30%) and Fumasep-FAA (15−30%) ionexchange membranes. 46 Therefore, it is reasonable to deduce that water permeability for these membranes is not sufficiently high to compromise the membrane performance in MCDI. These values show that increasing the mass loading of PEI can act to modify the wetting properties of the membranes, as water molecules become confined in hydrophilic passages (PEI) in an otherwise hydrophobic (PBI) matrix. This increase in hydrophilicity has the potential to increase hydration and the anion conductivity of the membranes.

Electrochemical Properties of Membranes.
The ion-exchange capacity (IEC) of membranes is a crucial property; in AEMs, a high IEC will theoretically increase the number of chloride ions removed from saline water and subsequently transported and stored in the porous activated carbon electrodes. 47 The introduction of PEI into PBI should introduce many cationic groups and corresponding counterions into the membrane after methylation. Figure 5a) displays experimental and theoretical IEC values for prepared and commercial membranes. Q0 had a relatively low IEC of 1.53 mmol g −1 , supporting the aforementioned elemental analysis results with Q0 having the lowest iodide content of all AEMs tested. The IEC of PEI-incorporated membranes Q5, Q10, and Q20 showed strong improvement, giving high values of 2.27, 2.50, and 2.63 mmol g −1 , respectively. The quaternization degree (QD, eq 3) was determined by direct comparison of experimental and theoretical IEC. Calculated QD values were 84%, 92%, 90% and 93% for membranes Q0, Q5, Q10, and Q20, respectively. QD values were over 90% for all blend membranes, indicative of highly quaternized structures with high concentrations of charge carriers and counterions. All prepared PEI-incorporated membranes exhibited higher IEC values than the commercially available and high-performing AMX and FAA AEMs ( Figure 5a); 27,46 giving further credence to their application in MCDI. The experimental IEC values justified the incorporation of PEI groups into the PBI matrix: Q5, Q10, and Q20 all displayed superior ion exchange properties to the pristine Q0 membrane. These superior properties can be attributed to the blend membranes containing polymers which were both quaternized, resulting in abundant anion transport pathways throughout the membranes. Table 4 provides values for the permselectivity and area resistance of the prepared membranes. Permselectivity was used to quanitify the counterion (chloride) selectivity of the membranes, which was determined from chronopotentiometric curves ( Figure S6). The quaternized blended membranes showed higher permselectivity (Table 4) than did the pristine Q0 membrane, likely the result of the higher volume of fixed cationic groups (quaternary ammonium donated by the PEI polymer) acting to block the transport of sodium co-ions through the membrane. While all quaternized blended AEMs showed higher permselectivity than Q0, the values obtained were less than ideal permselectivity (P = 1) due to the increased water uptake of the PEI-containing membranes. Nevertheless, the hydration of the membranes was not so excessive as to sufficiently dilute the high concentration of quaternary ammonium charge carriers and sacrifice membrane permselectivity. 48 Membrane area resistance is another crucial electrochemical property of ion-exchange membranes. Low area resistance is preferential to improve ion transfer across the electrode/ membrane interface, hence increasing the energy efficiency of the MCDI process. 49 All blended membranes showed reduced area resistance compared to Q0 membrane (8.76 Ω cm 2 ). Q10 (1.99 Ω cm 2 ) and Q20 (1.46 Ω cm 2 ) membranes exhibited significant reduction in area resistance, owing to their enhanced hydrophilic properties and IEC. The combination of high IEC, low area resistance, and good permselectivity of the highly quaternized blended AEMs pointed to excellent membrane properties for application in MCDI. Area resistance values obtained for Q10 and Q20 were below those of selected commercial IEMs, however permselectivity values were slightly lower than expected values. 46 This is a likely side-effect of the increased water uptake and IEC, introducing small amounts of Na + ions into the membrane. Nevertheless, a maximum permselectivity value of 0.86 was achieved for the Q20 membrane. A good permselectivity is favorable for minimizing co-ion expulsion during MCDI cycling. 50 3.5. MCDI Desalination Performance. Asymmetric MCDI cells were assembled as depicted in Figure S4, with quaternized AEMs covering the anode only. The electrodes had an approximate area of 1 cm 2 and were separated by a spacer of width 2 mm to open up a flow channel. A flow rate of 5 mL min −1 was chosen to provide a sufficient ion residence time for uptake through the membrane/electrode. A constant salt water concentration of 200 mg L −1 was chosen for desalination tests; this salt content was previously employed in MCDI studies. 19,51 Both salt adsorption capacity and charge efficiency can be influenced by the feed salt concentration, 21,52 therefore a constant salt concentration was maintained throughout all experiments to isolate the effects of the AEM on the MCDI performance.
Prior to desalination, cyclic voltammetry (CV) responses were recorded for all systems. The voltage was varied between 0.0 and 1.2 V at a scan rate of 0.2 V s −1 and the current response was measured. CVs are displayed in Figure S7 and are indicative of the capacitive behavior of all MCDI systems, with no observed oxidation or reduction peaks at any voltage. Figure 5b) shows the conductivity profile and regenerative behavior of the MCDI system using the Q20 membrane. During each adsorption/desorption cycle a conductivity drop is observed during the charging step (1.2 V), followed by an increase in conductivity during short-circuiting (0.0 V). These results are consistent with regeneration profiles found in batchmode MCDI configurations, 51,53,54 indicating that membranes can maintain the characteristic desalination performance over This was done immediately after an initial cycling of 1 h adsorption/desorption cycles (up to 50h total). The profiles depicted in Figures S13 and S14 show consistent cycling and conductivity profiles over this period, demonstrating that AEMs can maintain desalination performance over long-term MCDI operation and for cycles of longer duration. SAC and charge efficiency values for all systems are shown in Figure 5d); all MCDI systems show drastic improvement in SAC compared to CDI without membranes (4.2 mg g −1 ). Additionally, all quaternized polymer blended membranes displayed higher salt adsorption capacity than the Q0 membrane without PEI polymer (9.4 mg g −1 ). The salt adsorption capacity increased with increasing PEI content giving values of 10.6, 11.6, and 13.2 mg g −1 for Q5, Q10, and Q20, respectively. Charge efficiency values showed strong improvement compared with conventional CDI (27.7%). Again, charge efficiency values showed an increase as the PEI mass loading of the membrane increased. The pristine Q0 membrane MCDI system exhibited modest charge efficiency (40.5%), whist the quaternized blend AEMs returned values of 50.0%, 55.7%, and 67.9%, respectively.
Additionally, salt adsorption rate (SAR) and energy normalized to adsorbed salt results are displayed in Table 5.
The lower resistance to ion transfer, combined with abundant fixed charged carriers, increased the salt adsorption rate (SAR) of the MCDI systems. The measured SAR values were 0.16, 0.18, 0.19, and 0.22 mg g −1 min −1 for Q0, Q5, Q10, and Q20 membranes, respectively. This gradual increase further highlights the positive influence that quaternized PEI domains can have on the rate of ion transport into the electrodes. Likewise, ENAS is a crucial performance metric, relating the quantity of salt removed per unit energy supplied during charging. This value increased consistently with increasing PEI content. The Q20 membrane (5.87 μmol J −1 ) displayed more than double the value of the CDI (2.39 μmol J −1 ) system, indicating a system with superior energy efficiency. The highest ENAS value (Q20 membrane) was comparable to some literature values obtained for a symmetric cell employing both a commercial AEM and CEM; 49 however, it was slightly lower than those for polyelectrolyte-coated activated carbon electrodes. 55 The results validated the use of quaternized polymer blends as AEMs in MCDI. The SAC and SAR values obtained for each MCDI system scaled with increasing IEC of the membranes following Q0 < Q5 < Q10 < Q20. The high IEC and permselectivity of the membranes selectively permitted more counterions (chloride) to transport into the electrode pores during adsorption, due to the high volume of cationic ammonium charge carriers in the membrane backbone. Furthermore, the significantly reduced area resistance of the quaternized blended AEMs was beneficial to co-ion effects. During desorption, expelled co-ions were blocked from entering the spacer channel while counterions were readily desorbed across the membranes with low ionic resistance. This also explains the increase in charge efficiency and ENAS in the MCDI systems; corresponding to a decrease in the area resistance between Q0 and Q20 AEMs.
The improvement of all MCDI systems compared with CDI can be rationalized by co-ion effects. Membranes with high permselectivity (MCDI) result in an accumulation of co-ions in the macropores of the carbon electrodes, as the transport of sodium ions back into the spacer channel (co-ion expulsion) is blocked. This draws across more counterions to maintain electroneutrality in the macropore region, increasing the salt adsorption during subsequent adsorption cycles. 56 These coion effects elucidate the significant increase in SAC (3.1 times higher) and charge efficiency (2.5 times higher) of MCDI with the Q20 membrane compared with conventional CDI. Owing to the higher IEC, reduced area resistance and higher permselectivity, the use of the Q20 (SAC 13.2 mg g −1 ) membrane improved the desalination performance by a factor of 40% compared with the Q0 membrane without PEI (9.4 mg g −1 ). Furthermore, the fabricated membranes outperformed similar MCDI systems in the literature as shown in Table S1. The application of quaternized blended AEMs has been shown to positively affect desalination performance in MCDI, even with an unprotected cathode. The superior performance of all blended AEMs compared to Q0 vindicates the incorporation of quaternized PEI. This study has systematically investigated the effect of quaternized PBI/PEI membranes on the desalination performance of MCDI. In future works, focus will be placed on optimization of factors such as membrane structure, polymer blend interactions, dope solution composition, and quaternization conditions. This optimization will further improve the salt removal and energy efficiency of the systems and give insight into the feasibility of the membranes for industrial MCDI application.

CONCLUSIONS
In this work, AEMs for MCDI were fabricated based on quaternized PBI/PEI polymer blends. The membranes were prepared by a phase inversion process followed by heat treatment and quaternization by methylation. The highest performing Q20 membrane exhibited excellent electrochemical properties such as high IEC (2.6 mmol g −1 ), good permselectivity (0.86) and low area resistance (1.46 Ω cm 2 ). We attributed these properties to the abundance of anionselective quaternary ammonium groups introduced in the PBI/ PEI backbone. Quaternized AEMs demonstrated superior performance to the performance of both CDI (without membranes) and the pristine Q0 membrane. The employment of the Q20 membrane as AEM provided a large increase in SAC (13.2 mg g −1 , 3.1 times higher), charge efficiency (67.9%, 2.5 times higher), SAR (0.22 mg g −1 min −1 , 3.1 times higher), and ENAS (5.87 μmol J −1 , 2.4 times higher), compared to CDI without an AEM. Overall, this study outlines a simple preparation method for high performance AEMs for MCDI. The utilization of quaternized polymer blends as AEMs in MCDI has potential to enhance membrane properties and