High-Performance Membrane Capacitive Deionization Based on Metal-Organic Framework-Derived Hierarchical Carbon Structures.

Membrane capacitive deionization (MCDI) is a simple and highly energy efficient method to convert brackish water to clean water. In this work, a high-performance MCDI electrode architecture, which is composed of three-dimensional graphene networks and metal-organic frameworks (MOFs)-derived porous carbon rods, was prepared by a facile method. The obtained electrode material possesses not only the conducting networks for rapid electron transport but also the short diffusion length of ions, which exhibits excellent desalination performance with a high salt removal capacity, i.e., 37.6 mg g-1 at 1.2 V in 1000 mg L-1 NaCl solution. This strategy can be extended to other MOF-derived MCDI electrodes.

Introduction

The shortage of fresh water has emerged as a global challenge nowadays. Capacitive deionization (CDI), a novel water purification technique, has attracted increasing attention due to its low cost and high energy efficiency.[1] In the CDI process, an electrical potential is applied across two porous carbon (PC) electrodes and salt ions can be adsorbed and stored by forming the electrical double layers (EDL) on the electrodes. However, the ions adsorbed on the electrode simultaneously expel the co-ions during the adsorption process, which leads to a decreased desalination efficiency.[2−6] An effective way to alleviate such a problem is to integrate the ion-exchange membranes (IEMs) with typical CDI systems to construct a membrane capacitive deionization (MCDI).[7−10] The MCDI has exhibited enhanced performance in terms of the electrosorption capacity as well as charge efficiency.[11−15]

Based on the working principle of MCDI, its performance strongly depends on the properties of electrode materials. In theory, the electrode materials are supposed to have high electrical conductivity, hierarchically porous structure, good wettability, and large surface areas.[16] Various carbon materials have been investigated as electrode materials, such as graphene,[17−24] carbon nanofibers,[25−27] mesoporous carbon,[28−30] carbon nanotubes,[31] carbon aerogels,[32,33] and carbon composites.[34−37] Until now, many efforts have been devoted to design novel electrode materials to further improve the desalination capacity and efficiency.

Recently, metal–organic frameworks (MOFs), a type of porous crystals with ordered pore structure, have been demonstrated as versatile precursors for the synthesis of porous carbon (PC).[38−40] Due to its high specific surface area and porosity, MOF-derived PC holds great promise for high-performance CDI electrodes.[41−47] For instance, Pan et al. reported that a PC electrode obtained via pyrolysis of ZIF-8 showed a high desalination capacity of 13.86 mg g–1.[48] Our recent work demonstrated that the PC, prepared by using bimetallic MOFs as a precursor, delivered an excellent desalination capacity of 45.62 mg g–1 when used as MCDI electrodes.[49] However, MOF-derived PC electrodes usually suffer from poor electrical conductivity due to the low crystallinity of the MOF-derived carbon.[50,51] Incorporating graphene into MOF-derived PC is an efficient strategy to further improve the electron transport.[52,53]

In this work, we report a novel and high-performance MCDI electrode based on a hierarchically porous carbon composite, which is constructed by an interconnecting three-dimensional (3D) graphene network and MOF-derived porous carbon rods/reduced graphene oxide (PC/rGO). The fabricated PC/rGO composite electrode exhibits excellent desalination performance, e.g., 37.6 mg g–1 at 1.2 V in 1000 mg L–1 NaCl solution.

Results and Discussion

Scheme illustrates the synthetic process of MOF-derived porous carbon/reduced graphene oxide (PC/rGO) composites. The presynthesized Fe–MOF crystals (Figure S1) are quickly added into GO solution and vigorously mixed for 5 min (Scheme , step 1). In this process, Fe–MOF crystals and GO nanosheets are immediately self-assembled into a 3D structure. This is a result of the break in the charge balance of GO dispersion. After a freeze-dry process, Fe–MOF/GO composite aerogels are obtained (Scheme , step 2). The prepared Fe–MOF/GO composites are subjected to the subsequent annealing and etching process, which results in the formation of PC/rGO composites (Scheme , steps 3 and 4). To investigate the optimal composition of PC/rGO composites used for MCDI electrode, a series of Fe–MOF/GO composite aerogels were obtained by adjusting the weight ratios of Fe–MOF to GO. The obtained Fe–MOF/GO composite aerogels are designated as Fe–MOF/GO-n and their derived porous carbon/graphene composites are designated as PC/rGO-n, where n stands for the weight ratio of Fe–MOF to GO. For comparison, Fe–MOF-derived PC and rGO aerogel were also prepared.

Scheme 1

Schematic Illustration of the Synthetic Process of PC/rGO Composites

The morphology of the as-prepared Fe–MOF/GO-20 composite and its derived PC/rGO-20 composite was characterized by scanning electron microscopy (SEM) and transmission electron microscopy (TEM). Figure a shows a 3D structure constructed by rodlike Fe–MOF crystals and interconnected GO nanosheets, in which Fe–MOF crystals were well dispersed on the GO nanosheets. Some wrinkles were observed on the surface of Fe–MOF, indicating that Fe–MOF rods are partially wrapped by GO nanosheets with thin thickness. The characteristic peaks of Fe–MOF were shown in X-ray diffraction (XRD) pattern (Figure S2),[54] whereas the peak of GO (usually at ∼10.5°) was not observed due to the low weight percentage of GO in the composite.

Figure 1

(a) SEM image of Fe–MOF/rGO-20 composite. Inset: magnified SEM image of Fe–MOF/rGO-20 composite. (b) SEM and (c) TEM images of Fe–MOF/rGO-20 composite after annealing. Inset in (c): HRTEM image of a Fe2O3 particle. (d) SEM and (e, f) TEM images of PC/rGO-20 composite. Inset in (d): magnified SEM image of PC/rGO-20. Inset in (e): HRTEM image of PC/rGO-20.

After the annealing process, GO was reduced to rGO and Fe-MOF/GO-20 composite was converted to a composite of Fe, Fe2O3, and rGO. This is confirmed by XRD characterization (Figure S3). The diffraction peaks corresponding to (220), (311), (320), (421), (511), and (440) planes for Fe2O3 (JCPDS: 39-1346) were observed.[55] In addition, the XRD pattern of the annealed Fe-MOF/GO-20 also shows the (110), (200) peaks of Fe (JCPDS: 06-0696), indicates the existence of Fe residues.[56]Figure b indicates the annealed Fe–MOF crystals still preserve a rodlike shape similar to that of Fe–MOFs. Some large particles with sizes of several hundred nanometers were also observed on the annealed Fe–MOF crystals. The TEM observations in Figure c are consistent with SEM. High-resolution TEM (HRTEM) image shows a lattice distance of ∼0.26 nm (inset of Figure c) corresponding to the (310) plane of Fe2O3.[57]

The PC/rGO-20 composite was obtained after the subsequent etching process of Fe/Fe2O3/rGO composite. Figure d shows the 3D structure of PC/rGO-20 composite composed of interconnecting graphene networks and rod-shaped PC. From a close-up view of the rod part in inset of Figure d, the large particles were removed by acid, forming porous carbon rods. TEM images in Figure e,f indicate the PC rod is highly porous with both micro- and nanosized pores. Moreover, thin layers of rGO sheets were observed on the surface of a PC rod. The XRD patterns of PC/rGO-20, PC, and rGO all present two broad peaks at ∼25 and ∼44°, which can be indexed to the (002) and (101) planes of graphitic carbon, respectively (Figure a). The obtained PC and PC/rGO-20 composite were further characterized by Raman spectroscopy (Figures b and S4) and X-ray photoelectron spectroscopy (XPS) (Figure c,d). As shown in Figure b, the two characteristic peaks of graphene, i.e., D and G bands, were observed in PC/rGO-20, indicating the presence of rGO.[58] The increase in the intensity ratio of D to G band (ID/IG) from 0.97 in Fe–MOF/GO-20 to 1.04 in PC/rGO-20 confirms the reduction of GO during the annealing process. The XPS spectra of PC and PC/rGO-20 were presented in Figure c. Both samples show the presence of C 1s peak at ∼283 eV, N 1s peak at ∼400 eV, and O 1s peak at ∼531 eV.[35,59] There are no Fe in the XPS spectra, which is consistent with the XRD results. High-resolution N 1s spectrum suggests two main peaks at ∼397 and 400 eV, which correspond to pyridinic-N and graphitic-N, respectively (Figure d).[53]

Figure 2

(a) XRD patterns of rGO, PC, and PC/rGO-20 composite. (b) Raman spectra of Fe–MOF/GO-20 and PC/rGO-20 composites. (c) XPS spectra of PC and PC/rGO-20. (d) High-resolution N 1s XPS spectra of PC and PC/rGO-20.

A series of PC/rGO-n composites (n = 10, 20, 30, and 40) derived from Fe–MOF/GO composites with different weight ratios of Fe–MOF to GO were also prepared based on our method as described in Scheme (Figure S5). The surface areas of PC/rGO composites were further investigated by N2 adsorption–desorption measurements. As shown in Figure S6, the PC/rGO composites display a type IV isotherm with sharp uptakes at low relative pressure (<0.05) and H3 type hysteresis loops at high pressure, indicating the coexistence of micropore and mesopore, respectively. Meanwhile, the rising curves of N2 uptakes at higher pressure from 0.9 to 1.0 indicate the existence of macropores, which is possibly contributed by the void space among interconnected graphene sheets. The PC/rGO-20 composites possess the largest specific surface area (713 m2 g–1) among all the PC/rGO-n composites. This value is also significantly higher than those of PC and rGO (558 and 358 m2 g–1).

To evaluate the electrochemical behavior of the obtained electrodes, cyclic voltammetry (CV) was performed in 1 M NaCl solution with a potential window of −0.4–0.6 V at various scan rates from 2 to 100 mV s–1. The rectangular shape of CV curve for a PC/rGO electrode suggests a typical electrical double-layer capacitance behavior of electrodes (Figure a). As compared with other electrodes, including the PC/rGO composites with varied compositions, PC and rGO electrodes, PC/rGO-20 electrode exhibits the largest integrated area of CV curves (Figure a and Figure S7). This is in agreement with the calculated specific capacitances of all fabricated electrodes (Figure b,c). The PC/rGO-20 electrode shows the highest specific capacitance among all the electrodes (218 F g–1 at 2 mV s–1). The optimal combination of large specific surface area and high conductivity leads to the enhanced performance of PC/rGO-20.

Figure 3

(a) CV curves at a scan rate of 5 mV s–1 and (b) specific capacitances under different scan rates of PC/rGO-10, PC/rGO-20, PC/rGO-30, and PC/rGO-40 composites. (c) Specific capacitances at 2 mV s–1 and (d) Nyquist plots of rGO, PC/rGO-10, PC/rGO-20, PC/rGO-30, PC/rGO-40, and PC electrodes. Inset is the magnified Nyquist plots.

To further understand the electrochemical properties of PC/rGO electrodes, electrochemical impedance spectroscopy (EIS) analysis was carried out. Figure d displays the Nyquist profiles of rGO, PC, and PC/rGO-n electrodes in 1 M NaCl solution. The plots show an inclined line in the low-frequency region and a small semicircle in the high-frequency range. The small diameter of the semicircle indicates a small charge transfer resistance and a high electrical conductivity of the PC/rGO electrode.[60] In the low-frequency region, the inclined part corresponds to the Warburg impedance. The inclined line of the PC/rGO-20 electrode shows the largest slope, suggesting the most efficient ion diffusion in its porous structure.[61]

To further investigate the desalination performance, batch-mode MCDI experiments were carried out in NaCl solution with an initial concentration of 500 mg L–1. As shown in the scheme of the custom-made MCDI cell (Figure S8), cation- and anion-exchange membranes are placed in front of cathode and anode, respectively. Figure a shows the typical electrosorption–desorption cycle of PC/rGO-20 with an initial NaCl concentration of 500 mg L–1 at a voltage varying from 1.0 to 1.6 V. When a cell voltage is applied, salt ions start to immigrate to and adsorb on the electrodes. The solution conductivity decreases slowly until an equilibrium is reached after ∼50 min. Subsequently, a reverse voltage is applied on the electrodes. The conductivity slowly goes back to the initial value, indicating the regeneration of electrodes and can be used for the subsequent cycle. Clearly, with the increase in the cell potential, larger decrease in conductivity is observed due to the adsorption of more ions. The transient current curve of PC/rGO-20 in Figure b exhibits a similar trend to that of the conductivity. Charge efficiency is used to evaluate the energy consumption of MCDI and analyze its EDL model. Generally, higher voltages are beneficial to enhancing the charge efficiency of MCDI system. Figure c shows the charge efficiencies of the PC/rGO-20 electrode in 500 mg L–1 NaCl solution at voltages from 1.0 to 1.6 V. The electrosorption–desorption behaviors of the other PC/rGO electrodes as well as PC and rGO electrodes were examined (Figures S9 and S10). Their salt removal capacities are calculated and summarized in Figure d. Consistent with the CV results, PC/rGO-20 exhibits the highest electrosorption capacity. The results indicate the synergistic effect between rGO sheets and the PC derived from Fe–MOF. Moreover, the ratio between Fe–MOF and GO played an important role in the enhanced performance of the composite electrodes. The CDI Ragone plot, combing two important parameters of salt adsorption capacity and salt adsorption rate, is an effective measure to evaluate the capacitive deionization behavior of the electrodes.[62] CDI Ragone plots of salt removal rate vs salt removal capacity of PC/rGO-20 composite, PC, and rGO are shown in Figure S11. Clearly, PC/rGO-20 composite located in the upper-right of the Ragone plot, delivers both a higher salt removal rate and a higher salt removal capacity.

Figure 4

(a) Electrosorption behaviors and (b) the corresponding current responses of PC/rGO-20 electrode at various cell voltages. (c) Charge efficiencies of PC/rGO-20 electrode at 1.0–1.6 V. (d) Salt removal capacities of PC/rGO-10, PC/rGO-20, PC/rGO-30, and PC/rGO-40 composite electrodes at various cell potentials.

As shown in Figure a, the PC/rGO-20 electrode shows an excellent stability for 22 cycles without an obvious decay of electrosorption capacity. Further electrosorption experiments were carried out with different initial concentrations from 125 to 1000 mg L–1. Figure b illustrates the electrosorption isotherms of PC/rGO-20 electrode at cell potentials of 1.0, 1.2, 1.4, and 1.6 V, respectively. The Langmuir adsorption isotherm method is adapted to fit our data, which is presented as , where q is the amount of adsorbed NaCl (mg g–1), qm is the maximum adsorption capacity corresponding to complete monolayer adsorption, KL is the Langmuir constant, and C is the equilibrium concentration (mg L–1). As shown in Figure b, with the initial concentration increasing from 125 to 1000 mg L–1, the salt removal capacity of PC/rGO-20 increases accordingly. The PC/rGO-20 composites exhibit a superior salt removal capacity of 37.6 mg g–1 in a 1000 mg L–1 NaCl solution at 1.2 V, which is among the best of the previously reported electrode materials for CDI or MCDI (Table S1). The excellent performance of the PC/rGO composite electrodes can be ascribed to the following reasons: (1) the 3D porous structure constructed by conducting rGO networks provides a highly conductive pathway. (2) The highly porous carbon derived from Fe–MOF contributes to the enhanced electrosorption capacity. (3) The combination of porous carbon and rGO networks with optimum ratio can integrate the advantages of high surface area and high electrical conductivity, leading to the excellent MCDI performance.

Figure 5

(a) Cycle performance of PC/rGO-20 at 1.2 V. Inset: capacity retention of PC/rGO-20. (b) The electrosorption isotherms of PC/rGO-20 electrode at various cell potentials. Red lines are the Langmuir isotherm fitting.

Conclusions

In conclusion, a novel MCDI electrode consists of a highly interconnecting 3D graphene architecture with MOF-derived porous carbon rods, which can not only provide conductive networks for electron transport but also shorten the diffusion length of ions. Therefore, our PC/rGO composite electrode exhibits excellent desalination performance, i.e., 37.6 mg g–1 in a 1000 mg L–1 NaCl solution at 1.2 V. This strategy can also be extended to other MOF-derived carbon electrodes.

Experimental Section

Method

Preparation of Fe–MOF Crystals

FeCl3, 0.6488 g, was dissolved in 10 mL deionized (DI) water to form a yellow solution, which was mixed with another solution composing of 0.4643 g fumaric acid and 10 mL dimethylformamide (DMF). The formed mixture was then transferred into a Teflon-lined steel autoclave, heated at 80 °C for 4 h, and naturally cooled down to room temperature. The collected samples were subsequently washed with DMF and DI water for several times and then dried at 60 °C for 12 h.

Preparation of Fe–MOF/GO Composites

The Fe–MOF/GO composite aerogel was prepared by the method reported in our previous report. The Fe–MOF was added into the GO solution (4 mg mL–1) and vigorously mixed for 5 min using a QL-901 vortex mixer, forming a Fe–MOF/GO composite hydrogel. Fe–MOF/GO composite aerogel was obtained by freeze-dry process. To adjust the weight ratios of Fe–MOFs to GO (10:1, 20:1, 30:1, and 40:1), Fe–MOF/GO-10, Fe–MOF/GO-20, Fe–MOF/GO-30, and Fe–MOF/GO-40 were obtained.

Preparation of PC/rGO Composites

PC/rGO composites were obtained by carbonization of Fe–MOF/GO-n at 800 °C for 1 h with a heating rate of 5 °C min–1 under N2 atmosphere. After annealing, the samples were etched in 5 mol L–1 HCl aqueous solution for 24 h. Finally, the products were washed using DI water until pH close to 7 and then dried at 60 °C overnight.

Characterization

The morphologies of obtained samples were recorded by field-emission scanning electron microscopy (HITACHI SU-8010) and transmission electron microscopy (TEM, JEM-100CX II). The samples were characterized by X’Pert PRO (PANalytical) with Cu Kα irradiation (λ = 15.4 nm). Nitrogen adsorption–desorption was measured by ASAP 2020 (Micromeritics). Multipoint Brunauer–Emmett–Teller is used to calculate the specific surface area from N2 adsorption data. Raman spectroscopy was performed by Renishaw inVia confocal Raman system with the laser wavelength of 532 nm. Cyclic voltammetry (CV) measurements were conducted in 1 mol L−1 NaCl solution by using CHI 760E electrochemical workstation in a three-electrode mode. The electrochemical impendence spectroscopy (EIS) measurements were carried out in 1 mol L–1 NaCl solution by using Autolab PGSTAT302N.

Fabrication of MCDI Electrodes

MCDI electrodes were prepared by mixing samples, conductive carbon black, and poly(vinylidene fluoride) (PVDF) in a mass ratio of 8:1:1. The mixture was stirred uniformly with the addition of n-methyl pyrrolidone (NMP) solution and coated on a graphite paper of 5 × 5 cm2. Finally, the electrodes were dried at 60 °C for 12 h in vacuum oven. The mass loading is around 50 mg.

Electrosorption Measurement

To evaluate the desalination performance of the samples, batch-mode experiments were carried out by a continuously recycling system, which includes a MCDI cell, a peristaltic pump (YZ-1515x, Longer Pump), a conductivity meter (DDSJ-308F, Leici), and a source meter (SMU-2400, Keithley). The IEMs were purchased from Zhejiang China (Hangzhou Grion Environmental Technology Co., Ltd). The type of cation exchange membrane is polyethylene heterogeneous ion exchange membrane (LE-HeCM-1, Type 1), and the type of anion exchange membrane is heterogeneous ion exchange membrane (LE-HeAM-I, Type 1). The anion exchange membrane was placed in front of the positive electrode and the cation exchange membrane in front of the negative electrode. The total volume of the initial concentrations of NaCl solutions is 50 mL with concentrations varying from 125 to 1000 mg L–1, and the flow rate is 30 mL min–1. The electrosorption tests at every concentration were performed at different cell voltages ranging from 1.0 to 1.6 V with an interval of 0.2 V. The conductivity is recorded every 20 s. The electrosorption capacity, Γ (mg g–1), is calculated by eq where Ci is the initial NaCl concentration (mg L–1), Ct is the final NaCl concentration (mg L–1), V is the volume of NaCl solution (L), and M is the total mass of the electrodes (g).

The charge efficiency, Λ (%), is defined by eq where F is Faraday’s constant (96485 C mol–1), Γ is the electrosorption capacity (mol g–1), ∑ is calculated by the integration of charging current curve (C g–1), and MNaCl is the molar mass of NaCl (58.44 g mol–1).