Facile Fabrication of Defect-Controlled Graphene Oxide Membrane through Shear-Induced Alignment for Regulating Ion Transport.

Two-dimensional (2D) membranes enable ion-sieving through well-defined subnanoscale channels. In particular, graphene oxide (GO), a representative 2D material with a flexible structure, can be manufactured into various types of membranes, while defects such as pores and wrinkles are readily formed through self-aggregation and self-folding during membrane fabrication. Such defects provide a path for small ionic or molecular species to be easily penetrated between the layers, which deteriorates membrane performance. Here, we demonstrate the effect of shear-induced alignment with continuous agitation on GO membrane structure during pressure-assisted filtration. The shear stress exerted on the GO layers during deposition is controlled by varying the agitation rate and solution viscosity. The well-stacked 2D membrane is obtained via the facile shear-controlled process, leading to an improved salt rejection performance without additional physicochemical modifications. This simple approach can be extensively utilized to prepare the well-ordered structure of other 2D materials in various fields where the defect control is required.

Introduction

Selectively permeable membranes enable us to control a mass transport of small species such as monovalent ions due to their unique pore or channel structures on the basis of physical sieving and electrical interaction.[1,2] They offer a wide range of applications in energy and environment, encompassing battery, fuel cell, and water purification, where the membrane efficiency is one of the decisive factors.[3−5] Hence, the development of a new class of membrane is critical for advancing a relevant performance in engineering applications, and it can be achieved by elaborately tailoring the structure of membranes in an atomic or molecular level on the basis of a structure–property relationship, which ultimately governs mass transport.

Although it is difficult for a conventional membrane with nanoscale-pores to control the transport of angstrom-scale species such as monovalent ions, the development of two-dimensional (2D) material-based assembly methods makes it possible to effectively manipulate the penetration of such small atomic and ionic species through subnanometer interlayer channels.[6−8] Graphene oxide (GO), one of the 2D building blocks, is widely utilized for the membrane separation process due to the structural versatility as well as subnanoscale interlayer channels.[9,10] In addition, as the GO layer has a high surface area with flexible 2D graphitic structures, the GO membrane can be easily fabricated by simple filtration methods via self-assembly.[11,12] For the application in desalination and water purification, the GO membrane allows water molecules to be easily agglomerated in the hydrophilic region via hydrogen bonding with oxygen functional groups, while simultaneously they can rapidly pass through the hydrophobic region with almost no friction, resulting in superior water permeability.[13,14] Although most studies have focused on the physicochemical modification of GO sheets by introducing additional functional groups or electrostatic moieties, the separation performance of the GO membrane is highly affected by the layer-stacked structures.[15,16] Therefore, it is imperative to develop stacking methods for the well-aligned laminated structure without nanopores and wrinkles that can deteriorate the membrane selectivity.

A variety of fabrication methods for GO membranes have been devised, which significantly influence their stacked structure and also membrane properties on the basis of a structure–property relationship.[17−19] Accordingly, in a well-aligned GO membrane, a key membrane property such as tensile strength and selectivity is remarkably enhanced.[20,21] For example, the compact GO membrane was obtained by reducing the electrical repulsive force between GO layers by adding opposite charged ions and forming cross-linked structures through chemical reactions in the nanochannels.[22−24] However, this method permanently changes the internal structure and the physicochemical environment of nanochannels, and the chemical reaction is limited only for specific molecules with functional groups. Also, the arrangement of the GO membrane can be improved with the help of external forces such as doctor blading or water floating; however, it is difficult to control the alignment of the assembled GO layers as the membrane thickness increases, because the force could be only applied to the membrane surface.[24−28] Likewise, since GO layers are continuously compressed by external pressure during pressure-assisted filtration, the folded-structure in the GO laminate could be stretched unlike vacuum filtration, leading to ordered membrane structures.[29,30] Nonetheless, when the GO layer is folded in the initial stage of stacking, the fully wrinkled structure is obtained in the final membrane architecture. Therefore, a continuous driving force is required to prevent self-folded structures during membrane fabrication, and it is anticipated that the more aligned GO membranes can be successfully obtained without structural defects.

In this work, morphologically controlled GO membranes were fabricated to demonstrate the effect of agitation-induced shear stress during pressure-assisted filtration on the laminated structure. The continuous shear stress is applied to GO layers by introducing a simple stirring during a stacking process, which not only prevents self-aggregation of the GO nanosheets but also facilitates to form a densely packed structure. In the case of the GO membrane assembled without shear stress, the nanopores and defects can be generated in the structure due to the self-aggregation of GO nanosheets, while the shear-controlled fabrication could remove the wrinkled structure by stretching the flexible GO layer with the continuous stirring process, as schematically presented in Figure a. The agitation process generates the shear stress, which can be varied with the stirring rate and the viscosity of the GO dispersion. As the nanopores and wrinkles within the membrane are known to deteriorate the separation performance, it is anticipated that the shear-induced well-stacked GO membranes exhibit an improved separation performance for small species such as monovalent ions.[31] With this simple pressure-assisted membrane fabrication method combined with continuous shear alignment, a well-ordered 2D laminated structure can be readily obtained without any complicated chemical reaction or additional treatment.

Figure 1

(a) Schematic illustration of the effect of shear stress on the membrane structure during the fabrication; red circles indicate the formation of pores and wrinkles. The shear stress profile depending on the relative position from the stirring obtained from (b) the mathematical equation and (c) CFD simulation.

Result and Discussion

Preparation of Defect-Controlled GO Membranes via Shear Stress

In order to understand the effect of shear stress on the ordered architecture, the GO membranes were fabricated by varying the stirring rate and the viscosity of the GO solution (the detailed membrane fabrication procedure is described in the Experimental Section and Figure S1). The stirring rate was set to 0, 100, 200, and 300 rpm, and the viscosity of the GO solution was adjusted by adding glycerol into water—the viscosity of which is 1400 times higher than that of water. The viscosity of the mixture of water and glycerol was measured by varying the ratio, as shown in Figure S2a,b. The maximum shear stress applied to the GO membrane is calculated via a mathematical equation according to the solution composition and stirring rate (Figure S2c and the calculation details are described in the Supporting Information). When the ratio of water to glycerol is 2:1, the GO membrane fabricated at 200 rpm is subjected to a shear stress similar to that of the GO membrane created at 300 rpm in deionized water (DIW); thus, the ratio of water to glycerol was set to 2:1 for comparing the effect of the same shear stress induced at different agitating rates on the membrane structure. Several types of membranes were manufactured under different conditions, and they are referred to as X-Y-GOM, where X indicates the type of the used solvents—such as DIW or 2:1, which means water and glycerol are mixed in a ratio of 2:1—and Y stands for the stirring rate during membrane deposition. For example, a GO membrane with a mixture of water and glycerol in a 2:1 ratio at 200 rpm is identified as 2:1-200-GOM.

As shown in Figure b,c, the shear stress of the pure water and water–glycerol mixture depending on the position from the center of the stirring at various shear rates was calculated using the empirically derived mathematical equation as well as computational fluid dynamics (CFD) simulation (Figure S3 and the details are explained in the Supporting Information). As the stirring rate increases, it is found that the shear stress applied to the dispersed GO layers is also intensified. The simulation presents a similar tendency with the calculated results where the shear stress is proportional to the stirring rate. However, there exists a discrepancy in the absolute values, where those obtained from the mathematical equation are approximately 5 times larger than the results from the simulation. It is deduced that such difference is ascribed to the influence of pressure applied during the fabrication and GO nanosheets, which were not considered in the equation, even though the shear stress in the solution is also caused by the solute. In addition, when the applied pressure increases in the simulated system, the generated friction between fluid layers is intensified, reducing the development of the flow; thus, the applied shear stress to the GO nanosheets would decrease. Despite such difference, both calculation results support the effect of the stirring rate and the viscosity of the solution on shear stress within the range of our study; hence, a series of GO membranes were prepared under the conditions used in the calculation.

Investigation of the Structure of Morphology-Controlled GO Membranes

To determine the effect of shear stress for controlling wrinkles and nanopore, cross-sectional and top-down scanning electron microscopic (SEM) images of the prepared membranes were acquired (Figure and Figure S4). Evidently, DIW-0-GOM, which was not stirred during deposition, has a folded structure that can form nanopores as reported in previous studies, and the wrinkles are also observed on the surface.[20] On the contrary, the shear-controlled DIW-100-GOM and DIW-200-GOM exhibit a well-laminated structure as well as a smooth surface morphology without noticeable wrinkles. In both cases, the self-folded GO layer is not observed in the cross-sectional image and the membrane surface becomes flat, implying that a defect formation could be effectively prevented by shear stress generated by the stirring process. However, DIW-300-GOM under the influence of the more intensive shear stress during fabrication has a crumpled structure, and it is consistent with the previous study in which the membrane deformation occurs at shear stress above the critical point (Figure d).[32] In addition, the deposited structure of the GO membrane depending on the position is examined, in which the membrane architecture fabricated at the same agitation rate is almost similar regardless of the position (Figure S5). Therefore, we demonstrate that the well-ordered structure can be obtained at the optimum shear stress, and the deformation occurs beyond such condition. The similar tendency is revealed in 2:1-Y-GOMs. For the 2:1-0-GOM without the stirring process, the self-folded structure is observed due to the self-agglomeration of GO layers, similar to DIW-0-GOM. However, the 2:1-100-GOM, which is under the influence of similar level of shear stress with DIW-100-GOM or DIW-200-GOM according to the calculation (Figure b,c), shows a neat appearance without self-constructed crease on the cross-sectional and surface images. In the case of 2:1-200-GOM, which is manufactured under a similar shear stress to the case of the DIW-300-GOM, the membrane deformation evolves again, which makes crumples and vesicles. For the 2:1-300-GOM, as the strongest shear stress is applied to the GO nanosheets, numerous wrinkled structures are found in the membrane surface, suggesting that shear stress above a critical point actually degrades the laminated architecture. Apparently, the optimum shear stress not only prevents wrinkle formation caused by self-folding but also hinders the defect formation by flatting the flexible GO nanosheet and forming a lamellar structure. Furthermore, the additional SEM images of the edge and central sites were measured to confirm the overall laminated structure of the glycerol inserted GO membranes (Figure S6). The stacked architecture of 2:1-GOMs is almost similar regardless of the deposited position, indicating that the shear-induced alignment method allows us to regulate the overall membrane structure.

Figure 2

Cross-sectional SEM images of (a) DIW-0-GOM, (b) DIW-100-GOM, (c) DIW-200-GOM, (d) DIW-300-GOM, (e) 2:1-0-GOM, (f) 2:1-100-GOM, (g) 2:1-200-GOM, and (h) 2:1-300-GOM; the inset shows the top-down surface image of each GOM (the scale bar of the cross-section and inset images are 1 and 100 μm, respectively).

In order to quantitatively evaluate the influence of shear stress on the wrinkle removal and the formation of the stacked structure for the GO membrane, each membrane was analyzed by atomic force microscopy (AFM) as shown in Figure (additional images are also presented in Figure S7). DIW-0-GOM and 2:1-0-GOM without the stirring process have roughnesses with Rq values of 63.9 and 63.4 nm, respectively, and some wrinkles exist on the surface due to the flexibility of the GO nanosheet. However, for DIW-100-GOM, DIW-200-GOM and 2:1-100-GOM, the Rq values are reduced to 40 nm approximately and wrinkles disappeared, since shear stress less than a certain value can flatten the surface by preventing self-aggregation as aforementioned. Nevertheless, the alignment of the GO membrane can be collapsed by forming wrinkled structures when the shear stress further increases. In the case of DIW-300-GOM and 2:1-200-GOM, which are deformed under a similar level of shear stress, the membrane surface shows a number of crumples so that the Rq rises to 63.6 and 58.4 nm, respectively (Figure d,g). 2:1-300-GOM modified by the strongest shear stress has the largest Rq value of 67.8 nm, which indicates that deforming the GO layer with a significantly intensive force results in more corrugated structures.

Figure 3

AFM images of (a) DIW-0-GOM, (b) DIW-100-GOM, (c) DIW-200-GOM, (d) DIW-300-GOM, (e) 2:1-0-GOM, (f) 2:1-100-GOM, (g) 2:1-200-GOM, and (h) 2:1-300-GOM (scan size: 20 μm × 20 μm). For comparison, images are colored in the same range, and Rq means the root-mean square of the roughness.

The interlayer distance of the as-prepared membranes subject to different shear stress was analyzed by X-ray diffraction (XRD) patterns (Figure ). The channel sizes of DIW-0-GOM, DIW-100-GOM, DIW-200-GOM, and DIW-300-GOM were measured as 7.9, 7.9, 8.2, and 8.3 Å, respectively. The d-spacing of the GO membrane becomes broader as the stirring rate increases, since the locally folded structure and micropores in the membrane are formed more when a stronger shear stress is applied during the fabrication. However, the moderate level of shear stress can help the GO layers to be stretched in the stacking process, so that the channel size of DIW-100-GOM is similar to that of DIW-0-GOM. When the mixture of water and glycerol was used for the GO dispersion, the interlayer distance of the resulting membrane appears to be changed. The d-spacings of 2:1-0-GOM, 2:1-100-GOM, 2:1-200-GOM, and 2:1-300-GOM were determined to be 8.8, 8.8, 8.9, and 9.1 Å, respectively. Presumably, the broader channel size is derived from the inserted glycerol, which could interact with oxygen functional groups via a hydrogen bond.[33,34] Although the channel size of the GO membranes does not dramatically increase depending on the stirring rate, the wrinkles and pores present in the layered structure can deteriorate on the membrane separation performance. In addition, the change of the channel size in a wet state was investigated, in which case, the interlayer spacing of the GO membranes is rather broadened as the water molecules are intercalated into the GO layers (Figure S8). Nonetheless, DIW-100-GOM presents the narrowest channel size, which indicates that the optimized shear-induced alignment can maintain the well-stacked architecture.

Figure 4

XRD patterns and the corresponding d-spacing values of (a) DIW-Y-GOM and (b) 2:1-Y-GOM.

Effect of Shear-Induced Morphology Control of GO Membranes on Ion Rejection and Water Permeance

To investigate the effect of the modified laminated structures by shear stress on the mass transport behavior, ion permeation tests were conducted by a dead-end filtration method using the wrinkle-controlled GO membranes. As shown in Figure a, DIW-0-GOM prepared without shear stress presents a NaCl rejection rate of 39.5%. On the contrary, in the line with the aforementioned structural changes, the GO membranes fabricated under the influence of shear stress by stirring shows the different salt rejection behaviors. DIW-100-GOM exhibits the highest NaCl rejection of 53.8% among the defect-controlled DIW-Y-GO membranes, which is 14.3% higher than that of DIW-0-GOM. Even though there is little difference in terms of the nanochannel size between DIW-0-GOM and DIW-100-GOM as confirmed from the XRD patterns, the self-aggregated GO layers in the DIW-0-GOM lead to defects in the laminated channel, deteriorating the rejection performance of the membrane. In addition, the NaCl rejection rate of DIW-200-GOM is 50.8%, similar to that of DIW-100-GOM since the internal crease or pores are removed by shear stress. However, for DIW-300-GOM, the NaCl rejection rate decreases to 43.7%, induced by the increased channel size that originated from the folded structure of the GO layers. The ion rejection results deviate from the general relationship regarding physical sieving between the interlayer channel size and the ion transport behavior, in which case, the ion rejection is inversely proportional to the d-spacing of the membrane. It indicates that pores and wrinkles in the membrane structure have a critical impact on the membrane separation performance. Furthermore, DIW-0-GOM and DIW-300-GOM with relatively more wrinkled structures show a higher water permeance than DIW-100-GOM and DIW-200-GOM, since a larger amount of water molecules can easily pass through the created defects. On the contrary, the water permeance is slightly reduced in DIW-100-GOM and DIW-200-GOM with few pores and defects, because the stretched GO layers enable to form well-ordered channel structures where water molecules can rapidly pass through the hydrophobic channel.

Figure 5

Water permeance and NaCl rejection performance for (a) DIW-Y-GOM and (b) 2:1-Y-GOM. Water permeance and Na2SO4 rejection performance for (c) DIW-Y-GOM and (d) 2:1-Y-GOM.

Figure b shows the NaCl rejection and water permeance of 2:1-Y-GOMs. 2:1-0-GOM has a NaCl rejection rate of 49.2%, which shows almost 10% of improvement compared to DIW-0-GOM under the same experimental conditions. Even though the interlayer spacing of the glycerol-inserted GO membrane is wider than that of DIW-0-GOM, it is expected that glycerol could form a strong hydrogen bond with oxygen functional groups within the GO nanochannel even after the washing process, possibly suppressing the ion penetration. Likewise, 2:1-100-GOM subjected to similar shear stress as with DIW-100-GOM exhibits an increased rejection rate of 66.1% due to the stretched GO laminate without nanopores and the presence of a glycerol blocking agent. However, the applied shear stress to 2:1-200-GOM is similar to that of DIW-300-GOM at different shear rates due to the increased viscosity of the glycerol mixed dispersion; thus, more defects are created in the DIW-300-GOM structure, which deteriorates the ion rejection performance. In addition, 2:1-300-GOM is strongly deformed during the filtration process, resulting in a low rejection of 42.5%. These results suggest that the membrane alignment is closely regulated by shear stress, and the ion rejection is affected not only by the size of the nanochannel but also by the presence of a defect. 2:1-100-GOM shows the lowest water permeance among the 2:1-Y-GOMs, which is attributed to the ordered structure with few defects and wrinkles in the nanochannels. In general, the water permeance is inversely proportional to the ion rejection rate, so that 2:1-100-GOM shows the lowest water permeance among the GO membranes, which is ascribed to the ordered structure with few defects and wrinkles in the nanochannels.[2] However, at other stirring conditions, the nanopores, which act as the passage of water molecules in a stacked structure, lead to increased water flux.

In order to confirm the effect of GO membrane alignment on the ion rejection, permeation tests were further conducted with Na2SO4 solution (Figure c,d). The ion sieving in the GO membrane follows the proposed mechanism—repulsion of anions by a negative charge on the GO surface and size exclusion by interlayer spacing.[3,35] Divalent anions can be effectively rejected by oxygen functional groups with the help of the Donnan exclusion mechanism. In addition, since sulfate ions with a relatively large hydration radius are more easily excluded by the narrow interlayer spacing, the ion rejection of Na2SO4 is generally higher than that of NaCl. The well-laminated GO membranes induced by shear modification have ordered nanochannels with few defects, thereby allowing the sulfate ions to be removed by physically compact nanochannels as well as the repulsive force caused by closer interaction distance with oxygen groups. Therefore, Na2SO4 rejection is improved by approximately 10% in DIW-100-GOM and 2:1-100-GOM compared with no stirring membrane. However, the Na2SO4 rejection rate of the membrane modified by excessive shear stress is deteriorated by the pore structures. These results show a very similar trend with NaCl rejection, suggesting that the improvement of GO membrane alignment by shear stress can improve performance regardless of ion type.

To examine whether the shear-induced modification method for well-ordered structures depends on the thickness of membranes, a thinner GO membrane was prepared with a less amount of GO, referred to as DIW-Y-0.5 GOMs. The morphology of thinner GO membranes was also confirmed via SEM analysis (Figure S9). Because there is almost no change in viscosity of GO solution, it is reasonable to suppose that DIW-Y-GOMs and DIW-Y-0.5 GOMs are subjected to similar shear stress. DIW-100-0.5 GOM and DIW-200-0.5 GOM also have a neat cross section without wrinkles or pores, and the folding structure is observed in DIW-0-0.5 GOM without agitating and DIW-300-0.5 GOM with deformation. The ion sieving performance was also measured using NaCl and Na2SO4 solutions (Figure S10), and there is no dramatic difference between the DIW-Y-0.5 GOMs and the DIW-Y-GOMs in terms of the trend in the ion rejection and water permeance depending on the stirring rate, presumably caused by the similar viscosity of the dispersion with less GO. The GO membrane fabricated at a stirring rate of 100 rpm has the highest salt rejection, and particularly, the Na2SO4 rejection rate is overall higher than the NaCl rejection rate, which is due to the Donnan exclusion mechanism.[35] Therefore, it is deduced that the shear-induced strategy to improve the separation performance would be fairly effective within a certain concentration of the dispersion.

Conclusion

In conclusion, the GO membrane with well-aligned structures was fabricated through continuous shear stress along with pressure-assisted filtration without additional physicochemical modification. The shear stress is controlled by varying the agitating rate and solution viscosity, calculated with the empirical equation and CFD simulation. From the obtained membrane structure, it can be deduced that the continuously applied shear stress flattens GO nanosheets during membrane deposition, significantly reducing wrinkles and nanopores in the GO laminate. With this facile alignment method, the NaCl rejection rate of the shear-controlled membrane can be improved by 14.3% compared to that of the GO membrane without shear stress. In addition, we confirm that the well-aligned membrane structure can be acquired under the optimum shear-controlled condition, leading to the enhanced separation performance. As the shear-induced alignment method along with a pressurization process can easily control the laminated GO structure via a continuous shear stress without any physicochemical change of GO layers, it is expected that this simple method can be utilized to obtain well-stacked structures of other 2D-based materials for various applications.

Experimental Section

Chemicals and Materials

Glycerol, NaCl, and Na2SO4 were purchased from Sigma-Aldrich. Single layer graphene oxide (SLGO) was purchased from ACS material. A PES filter with a 0.2 μm pore size was purchased from Hyundai micro Co.

Preparation of GO Dispersion

The 0.5 mg/mL GO stock solution was prepared by dispersing 50 mg of SLGO powder in 100 mL of DI water, followed by sonication for 3 h. Two kinds of GO sample solutions were prepared to examine the effect of viscosity as described. First, the GO sample dispersed in DIW was prepared to fabricate DIW-Y-GOMs by diluting the stock solution, in which case 2 mL of the stock solution was taken and added to 23 mL of DIW. Second, the GO sample dispersed in a mixture of DIW and glycerol—the ratio of DIW to glycerol was set to 2:1—was prepared to fabricate 2:1-Y-GOMs. In a similar way, 14.667 mL of DIW and 8.333 mL of glycerol were added to 2 mL of the GO stock solution. Furthermore, an additional dispersion was prepared to fabricate DIW-Y-0.5 GOMs, in which case 1 mL of stock solution was added to 24 mL of DIW to prepare the GO sample solution. All GO solutions were vortexed before use.

Preparation of GO Membrane

A PES substrate was placed onto the stirring cell, and each prepared sample solution was poured into the stirring cell. The GO membrane was fabricated under 2 bar of N2 gas by varying the stirring rate at 0, 100, 200, and 300 rpm for each case and subsequently washed with 50 mL of DIW twice under 4 bar. The resulting membranes were dried overnight at 80 °C under a vacuum.

Water Permeance and Ion Rejection Performance of GO Membrane

NaCl and Na2SO4 solutions (500 ppm) were prepared to test the separation performance of the GO membrane. Twenty-five milligrams of NaCl or Na2SO4 was added to 50 mL of DI water to produce 500 ppm of each salt solution. The water permeance and ion rejection tests were conducted using a conventional dead-end filtration method.[7] The prepared GO membrane was placed on the bottom of the stirring cell, and the 50 mL of the salt solution was poured in the cell. The permeation test proceeded for 30 min under 1 bar of N2 conditions, and the solution was agitated at 350 rpm in order to prevent the concentration polarization effect. All permeation tests were repeatedly conducted three times using different GO membranes for accuracy. The rejection rate and water permeance of the GO membrane were calculated by the following eqs and 2.[36]where C0 and CI represent the concentration of the permeated solution and the initial solution, respectively.where V is the volume of permeated solution, A is the effective area of the membrane, t is the time for permeation test, and P is the pressure applied during the test.

Characterization and Measurement

The viscosity of solution was measured with a rheometer (MCR 302, Antor Paar). The surface and cross-section morphology of membranes were characterized with field emission-SEM (SU5000, Hitachi). The surface roughness of membranes was analyzed with an AFM-Raman Spectrometer (INNOVA-LABRAM HR800, Bruker). The interlayer spacing of GO membranes was measured using Cu Kα (1.5406 Å) radiation in high-resolution powder X-ray diffractometer (SmartLab, RIGAKU) equipment. AFM, SEM, and XRD measurements were performed with the aid of the KAIST Analysis Center of Research Advancement (KARA). Electrical conductivity was measured by a conductivity meter (HC9021, Walklab).