Swelling of Graphene Oxide Membranes in Aqueous Solution

Jun 1, 2017 - Graphene oxide (GO) has recently emerged as a promising 2D nanomaterial to make high-performance membranes for important applications. H...
15 downloads 26 Views 6MB Size
Download PDF
Swelling of Graphene Oxide Membranes in Aqueous Solution: Characterization of Interlayer Spacing and Insight into Water Transport Mechanisms Sunxiang Zheng,† Qingsong Tu,† Jeffrey J. Urban,‡ Shaofan Li,† and Baoxia Mi*,† †

Department of Civil and Environmental Engineering, University of California, Berkeley, California 94720, United States The Molecular Foundry, Lawrence Berkeley National Laboratory, Berkeley, California 94720, United States

‡

S Supporting Information *

ABSTRACT: Graphene oxide (GO) has recently emerged as a promising 2D nanomaterial to make high-performance membranes for important applications. However, the aqueous-phase separation capability of a layer-stacked GO membrane can be significantly limited by its natural tendency to swell, that is, absorb water into the GO channel and form an enlarged interlayer spacing (dspacing). In this study, the d-spacing of a GO membrane in an aqueous environment was experimentally characterized using an integrated quartz crystal microbalance with dissipation and ellipsometry. This method can accurately quantify a d-spacing in liquid and well beyond the typical measurement limit of ∼2 nm. Molecular simulations were conducted to fundamentally understand the structure and mobility of water in the GO channel, and a theoretical model was developed to predict the d-spacing. It was found that, as a dry GO membrane was soaked in water, it initially maintained a d-spacing of 0.76 nm, and water molecules in the GO channel formed a semiordered network with a density 30% higher than that of bulk water but 20% lower than that of the rhombus-shaped water network formed in a graphene channel. The corresponding mobility of water in the GO channel was much lower than in the graphene channel, where water exhibited almost the same mobility as in the bulk. As the GO membrane remained in water, its d-spacing increased and reached 6 to 7 nm at equilibrium. In comparison, the d-spacing of a GO membrane in NaCl and Na2SO4 solutions decreased as the ionic strength increased and was ∼2 nm at 100 mM. KEYWORDS: graphene oxide, membrane, interlayer spacing, swelling, water transport along with ultrafast water transport7,8 and high mechanical strength.9,10 Analogous to the cylindrical pores in traditional polymeric membranes, the space between any two neighboring GO nanosheets within a GO membrane forms a nanosized channel that allows water to pass through while rejecting unwanted species. Like the pore size for traditional membranes, the interlayer spacing (commonly termed as d-spacing), which is defined as the center-to-center distance between two adjacent carbon planes (Figure 1), plays a key role in determining the separation properties of a GO membrane.11−14 Therefore, in order to achieve the full potential of a GO membrane for highperformance aqueous-phase separation, its nanostructure, in particular the d-spacing, must be accurately characterized.

A

s one of the most interesting derivatives of graphene, the two-dimensional (2D) atom-thick graphene oxide (GO)1 has recently drawn much attention for its capability of forming layer-stacked membranes that hold great promise for many important applications such as water treatment,2 desalination,3 batteries,4,5 and energy storage.6 The presence of abundant oxygenated functional groups on GO makes the material very hydrophilic and thus endows it with a high tendency to absorb water and swell in humid or aqueous environments, greatly deteriorating its targeted performance. Accurate characterization and a fundamental understanding of GO swelling are expected to provide key information for the optimal synthesis of GO membranes with enhanced performance. However, such information is lacking in the literature. Taking the aqueous-phase separation as an example, the GO membrane has exhibited an extraordinary separation capability © 2017 American Chemical Society

Received: May 1, 2017 Accepted: June 1, 2017 Published: June 1, 2017 6440

DOI: 10.1021/acsnano.7b02999 ACS Nano 2017, 11, 6440−6450

Article

www.acsnano.org

Article

ACS Nano

Figure 1. Schematic illustration of d-spacings of graphene, dry GO, and GO soaked in water.

Figure 2. Method for accurately characterizing the d-spacing of a GO membrane in aqueous solution. Integration of QCM-D and ellipsometry (A). Preparation of a GO-coated gold sensor (B). AFM characterization of the thickness of a GO film coated on a gold sensor and direct SEM observation of bare and GO-coated sensor surfaces, respectively (C).

The d-spacing of a GO membrane in a wet condition can be significantly larger than that in a dry condition. As illustrated in Figure 1, the d-spacing of a pure graphene channel is 0.34 nm, and the d-spacing of a dry layer-stacked GO film is 0.8 ± 0.1 nm, measured by X-ray diffraction (XRD).15−17 The latter value is close to twice the sum of the carbon−oxygen hydrogen bond length (∼0.22 nm) and van der Waals radius of oxygen (0.17 nm),18,19 totaling 0.78 nm. This dry configuration, if kept unchanged, would create a free spacing of ∼0.44 nm between two carbon planes, ignoring the protruding oxygenated groups. Such free spacing is ideal for admitting only water molecules (∼0.25 nm) and rejecting most hydrated ions (e.g., Na+ with a hydrated diameter of 0.5 nm),20 thereby achieving high water productivity and ion selectivity. However, the actual aqueousphase separation performance of a GO membrane is far below this ideal performance.21 A main reason is the swelling of the GO membrane in a wet condition (the highly hydrophilic GO nanosheets attract water molecules into the interlayer space of the GO membrane,22,23 thereby hydrating the GO nanosheets and increasing the d-spacing), which significantly impairs the separation efficiency. On the other hand, the swelling of GO, a property not possessed by other 2D materials such as graphene, can be advantageously used to precisely control the d-spacing of a GO membrane on a sub-nanometer scale. As an example, an approach for making GO membranes with tunable ionic sieving capability was developed to adjust the size of the GO channel by controlling the relative humidity condition and then fix the d-spacing using epoxy glue.24 Recently, the effects of GO membrane swelling in an aqueous environment on membrane performance have been reported, and characterization of such swelling has been attempted. For example, the observation of decreased rejection as a GO membrane encountered a cationic solution hinted at the membrane swelling by cations.25 The overall degree of swelling was assessed by examining the difference between the

weights of the GO membrane in dry and water-saturated conditions, respectively. 26−28 Due to the difficulty in completely removing water from the GO membrane surface, the weight of water contained in the GO membrane cannot be accurately measured, and thus membrane swelling was only roughly estimated. In addition, scanning force spectroscopy was used to quantify the change (up to 15%) in the thickness of a GO film with different levels of relative humidity in air.29 XRD was also used to directly measure the d-spacing (up to ∼1.3 nm) of wet GO samples.30,31 However, measuring d-spacing up to several nanometers in an aqueous environment presents a practical technical challenge for XRD measurement. For example, XRD was previously employed to show that the GO membrane swelled more in a diluted aqueous solution than in a concentrated solution;32 however, the d-spacing of the GO membrane in solutions more diluted than 0.5 M NaHCO3 was not reported. This is because severe swelling likely disrupts the alignment of stacked GO layers, and thus the XRD peaks for a GO membrane with a large d-spacing would be lost. Lack of suitable methods for accurately quantifying the d-spacing of a GO membrane in aqueous solution has not only impeded the fundamental understanding of water and solute transport through GO membrane but also adversely affected the synthesis of GO membranes with precisely controlled d-spacing for target contaminant removal. To address the above issue, we herein present an experimental method to characterize the d-spacing and the associated swelling of a GO membrane immersed in aqueous solutions. This method integrates two characterization toolsa quartz crystal microbalance with dissipation (QCM-D) and ellipsometryto simultaneously monitor the changes in the mass and thickness of a layer-stacked GO thin film. As a result, the density and d-spacing of a GO membrane that swells in aqueous solution can be accurately quantified. Molecular dynamics (MD) simulations were carried out to numerically 6441

DOI: 10.1021/acsnano.7b02999 ACS Nano 2017, 11, 6440−6450

Article

ACS Nano

Figure 3. Characterization of GO membrane swelling in DI water. GO membrane mass (measured by QCM-D) and thickness (measured by ellipsometer) as a function of soaking time (A). Density of a GO membrane calculated by eq 1 using the mass and thickness measurements (B). d-spacing of a wet GO membrane calculated by eq 3 (C).

an average value of ∼11 nm (Figure 2C), consistent with the thickness measured by ellipsometry. Also, both AFM and SEM images show a fairly homogeneous GO coating on the sensor surface (Figure 2C). Knowing that the graphite structure (sp2 hybridization) is typically associated with a d-spacing (dgraphite) of 0.34 nm, the average d-spacing of the dry GO membrane can be calculated as

investigate the water network structure near a GO surface to help understand the critical GO−water interactions that possibly affect the d-spacing and water transport. Both deionized (DI) water and representative solutions were used to comprehensively investigate the swelling of a GO membrane in an aqueous environment. Besides, the theoretical d-spacing was derived and compared to the experimental results. Finally, the implications of GO swelling for the synthesis and nanostructure optimization of high-performance GO membranes for aqueous separation are discussed.

ddry ‐ GO =

RESULTS AND DISCUSSION Characterization of GO Membranes by Integrated QCM-D and Ellipsometry. QCM-D can sensitively measure the mass of a thin film by monitoring the changes in the frequency and dissipation of an acoustic resonator, while ellipsometry is an optical technique that measures the thickness of a thin film by monitoring the change in the polarization of light upon reflection. By integrating QCM-D and ellipsometry (Figure 2A, with a detailed description of the integrated analysis in the SI and Figures S1 and S2), we were able to simultaneously measure the mass and thickness of a GO membrane in an aqueous environment and then quantitatively derive its density and d-spacing. To this end, a gold sensor was uniformly coated with a GO film by first drying the sensor on and then peeling it off from a freshly prepared wet GO film (Figure 2B). The GO membrane was first characterized in dry air to establish a baseline for the subsequent swelling study. In general, based on the surface mass M measured by QCM-D and thickness δ obtained from ellipsometry, the density of a GO membrane is determined as ρGO = M/δ

ρgraphite ⎛ M w,O ⎞ ⎟⎟dgraphite a⎜⎜1 + R ρdry ‐ GO ⎝ M w,C ⎠

(2)

where R is the oxygen-to-carbon (O/C) ratio, which is calculated to be 0.77 based on the atomic ratios measured from the X-ray photoelectron spectroscopy (XPS) spectrum (Figure S3). Mw,O and Mw,C are the molecular weights of oxygen (16) and carbon (12), respectively. The overall percent defect-free area of GO nanosheets, a, depends on the specific degree of oxidation, and its accurate value may be obtained using high-resolution transmission electron microscopy (TEM) while taking into account the potential interference from contaminants.34 On the basis of the experimental d-spacing of 0.76 nm from our XRD characterization of an oven-dried GO membrane, the value of a is back-calculated by eq 2 to be 88%. To study the swelling of a GO membrane in pure water, we continuously monitored for several days the changes in the mass and thickness of the membrane soaked in DI water. As shown in Figure 3A, there was a prompt initial increase in membrane mass from 2.0 to 4.2 μg/cm2 in less than 5 min after the GO membrane was soaked in water, most likely because water was quickly absorbed onto the surface and filled the free spaces in the GO membrane due to the high hydrophilicity of GO nanosheets. In contrast, the membrane thickness did not increase abruptly at the beginning, indicating that it took longer for the intercalated water layers to push the neighboring GO nanosheets away from each other to enlarge the d-spacing and cause swelling. As more water gradually entered into the spaces between GO layers, the membrane thickness steadily increased from a dry thickness δdry‑GO of 11.2 nm to a wet thickness δwet‑GO of ∼70 nm after 3 days and eventually reached a final value at a much lower speed. As a result of such unsynchronized initial increases in mass and thickness, the corresponding density of the GO membrane (Figure 3B) sharply reached a peak value of 3.75 g/cm3 and then quickly fell to a value slightly lower than that of a dry GO membrane within just 1 day of soaking. After that, the density gradually decreased and stabilized at ∼1.0 g/cm3, practically the same as

(1)

The surface mass and thickness of the dry GO membrane were measured to be 2.0 μg/cm2 and 11.2 nm, respectively. Thus, the density of the dry GO membrane, ρdry‑GO, is calculated to be 1.79 g/cm3, much lower than the graphite density (ρgraphite) of 2.26 g/cm3,33 mainly due to the enlarged d-spacing by the extrusion of oxygenated functional groups from the carbon plane. Note that in order to accurately measure the thickness of the relatively thin GO film, we also separately prepared thicker GO films and used a multisample analysis method (described in the SI) to eliminate the possible error caused by the periodic phase shift. In addition, we characterized the thickness of the coated GO film by atomic force microscopy (AFM) and found 6442

DOI: 10.1021/acsnano.7b02999 ACS Nano 2017, 11, 6440−6450

Article

ACS Nano

30 min air-drying process the XRD peaks quickly shifted to the right as the d-spacing decreased from 1.04 nm to 0.78 nm, indicating a relatively fast water evaporation rate as the wet GO membrane was air-dried. Therefore, it is recommended that XRD be carefully used to quantify the d-spacing of a wet GO membrane because, without proper sealing of wet samples to prevent water evaporation, considerable errors could be introduced in the d-spacing measurement. Understanding of Water Structure and Transport in a GO Channel. In order to explain the initial sharp increase in the density of a GO membrane immediately after it was soaked in water (Figure 3B), MD simulations were conducted to fundamentally examine the water structure inside a GO channel at this stage. Based on the XPS results (Figure 4B), a GO model was constructed with an O/C ratio of 0.23 and a composition of hydroxyl (60%) and epoxy (40%) groups that were randomly distributed on the carbon plane. Note that the carboxyl and carbonyl groups, which mainly reside at the edges of GO, were ignored in the MD simulations in order to focus on GO surface−surface interactions, but their masses were added back to correct the density using eq S3 after simulation. Such a strategy has been commonly adopted in similar studies.35,36 As seen in Figure 5A, the calculated densities of a wet GO membrane based on experimental data and MD simulations, respectively, agree on the general trends that a high density was associated with a small d-spacing when the sample was first soaked in water and that the density decreased with increasing d-spacing during the prolonged soaking. Therefore, it is qualitatively confirmed that, by absorbing water into its structure without expanding the d-spacing (as demonstrated in Figure 3A by an immediate increase in the membrane mass but not thickness), a wet GO membrane was actually denser than both a dry GO membrane (1.79 g/cm3) and bulk water (1 g/ cm3) at the initial stage of soaking. The continuing swelling gradually decreased the membrane density, which eventually reached a value equal to that of bulk water. However, the highest density of 2.3 g/cm3 from the simulation is much lower than 3.75 g/cm3, which was repeatedly measured with ±13% accuracy in the experiment, indicating the likely existence of other contributing factors. It is believed, as discussed in the SI and illustrated in Figure S5, that the water loosely associated with the rough surface of the GO membrane led to experimental overestimation of the mass and density of a wet GO membrane, especially in the initial soaking stage when the d-spacing was relatively small and thus the surface roughness effect was more pronounced. As an example, after the surface roughness effect was deducted using eq S5, the experimental density decreased from 3.75 g/cm3 to 2.3 g/cm3, perfectly matching the simulated density. The MD simulations indicated that a d-spacing of 0.8 nm corresponds to the minimum space that allows a continuous water layer to form between two GO sheets, since a smaller dspacing would lead to discrete water patches. It was further revealed that, with a d-spacing of 0.8 nm, the density of the formed water layer was 1.3 g/cm3 (Figure 5B), 30% higher than the bulk water density of 1 g/cm3. Besides, it was found that, with a larger d-spacing, the density of water near the GO plane also increased but quickly decreased to the bulk water density at a perpendicular distance of ∼1 nm away from the center of the GO plane (Figure S6). Since no external pressure was applied in the present study, the increase in water density near the GO plane was not due to the squeezing by two adjacent GO planes. The almost symmetric profile of water density near

that of pure water, when the electrostatic repulsion force slowly became dominant and pushed adjacent GO nanosheets apart to finally reach an equilibrium position. Recognizing that the total number of GO layers remains the same before and after membrane swelling, the average dspacing of a wet GO membrane is calculated as d wet ‐ GO = ddry ‐ GO

δwet ‐ GO δdry ‐ GO

(3)

As shown in Figure 3C, the d-spacing of a wet GO membrane in pure water increased steadily and reached 6.2 nm after 6 days of soaking, about 8 times the initial d-spacing of 0.76 nm for a dry GO membrane. Such a large d-spacing is very difficult, if possible at all, to probe by XRD, the most commonly used tool to characterize the d-spacing of 2D materials. As demonstrated in Figure 4A,

Figure 4. Characterization of the structural and chemical properties of GO. XRD spectra of a GO membrane in oven-dried, wetted, and air-dried conditions (A) and XPS spectra of the composition of carbon peaks (B).

XRD failed to measure the d-spacing of a GO membrane after 30 min of wetting in DI water. The rising Fourier transform infrared (FTIR) peaks at 1640 and 3500 cm−1 during the wetting process (Figure S4) confirmed the quick absorption of water by the GO membrane within 30 min of soaking. By drying the wet GO membrane in air for only 5 min, an XRD peak started to form. It is interesting to observe that during the 6443

DOI: 10.1021/acsnano.7b02999 ACS Nano 2017, 11, 6440−6450

Article

ACS Nano

Figure 5. MD simulation results for membranes soaked in pure water. Densities of wet GO with different d-spacings (A). Density profile of water between GO sheets with a d-spacing of 0.8 nm (B). Radial distribution functions of water in bulk, a graphene channel, and a GO channel, respectively (C). Side and top views of 1250 water molecules that formed a well-aligned rhombus-shaped crystal layer in a graphene channel with a d-spacing of 0.75 nm (D), a typical hexagonal structure of ice crystals in bulk water (E), and a less aligned rhombus-shaped crystal layer in a GO channel with a d-spacing of 0.8 nm (F). Alignment of water molecules along the armchair and zigzag directions of the graphene carbon lattice (G). Mean square displacements of water molecules in bulk, the graphene channel, and a GO channel, respectively (H).

both surfaces of a GO plane indicates that the water network near the GO plane was different from that in the bulk. To understand the effect of the GO plane on the water network, MD simulations were used to study the structures of bulk water and the water layer formed near a graphene sheet,

respectively, as a comparison. The radial distribution functions (RDFs) for water in bulk, the graphene channel, and the GO channel are compared in Figure 5C. It is observed that water between GO sheets and water between graphene sheets have similar RDF peaks (4.3 vs 4.4) and are higher than that of bulk 6444

DOI: 10.1021/acsnano.7b02999 ACS Nano 2017, 11, 6440−6450

Article

ACS Nano

Figure 6. Understanding the effects of various interaction forces on the d-spacing of a GO membrane in ionic solutions. GO membrane swelling, in terms of changes in equilibrium thickness and d-spacing, in various solutions of different ionic strengths (A). Experimental and theoretical d-spacings as well as Debye length of a GO membrane at equilibrium in NaCl solutions of different ionic strengths (B). Schematic illustration of repulsive and attractive forces between GO sheets (C). Distribution of electric potential ϕ and osmotic pressure π between two GO sheets in NaCl solutions of 1 mM (D), 10 mM (E), and 100 mM (F). van der Waals (VW) and Coulombic attractions (negative value) and electrostatic/osmotic repulsion (positive value) between two GO sheets with different d-spacings in a 1 to 100 mM NaCl solution (G).

water (3.7) for the first hydration layer, indicating more densely packed water molecules in both GO and graphene channels than in the bulk. Note that only the graphene-sandwiched water demonstrated a crystal feature, because the repeating small RDF peaks at longer distances can be generated only by a lattice with fixed distances between water molecules. In addition, there are no apparent RDF peaks at longer distances for GO-sandwiched or bulk water, indicating a macroscopically amorphous water network. The water structures formed in the bulk and in graphene and GO channels were compared in Figure 5D−F. Shown in Figure 5D are the side and top views of a water layer formed between two graphene sheets with a d-spacing of 0.75 nm, the minimum space allowing the formation of a continuous water layer. It is seen that water molecules formed a well-aligned rhombus (and even a perfect square)-shaped crystal layer with a thickness of only ∼0.2 nm and a density of 1.62 g/cm3. In Figure 5E, when the same number (1250) of water molecules as that between graphene sheets reached an equilibrium in bulk water with an identical top-view area, the thickness of water molecules was 0.38 nm, almost twice that of the water layer formed between graphene sheets, indicating a much looser network and also verifying a lower density of bulk water. The unique rhombus-shaped crystal structure formed between graphene sheets is very different from the typical hexagonal structure of ice crystals formed in bulk water. Recently, a similar square-shaped ice structure formed in a graphene liquid cell was reported37 and attributed to the squeezing effect of a very high lateral pressure applied to the two graphene sheets. Although the experimentally observed

crystal structure by TEM37 was questioned by some researchers,38 the reported simulation results37 agree well with our findings as shown in Figure 5D. However, unlike the graphene liquid cell applied with a high pressure,37 our simulation box was under atmospheric pressure only. As a result, the rhombus-shaped water crystal structure and the higher density observed in our simulated graphene channel were not caused by an external pressure. Hence, we hypothesize that the alignment of water molecules was mainly induced by the van der Waals forces between water molecules and graphene sheets. This hypothesis is supported by a high correlation between the water structure and the carbon lattice; as observed in Figure 5G, water alignment followed the armchair and zigzag directions of the carbon lattice, and the side of rhombus is equal to √3 times the length of a carbon− carbon bond. Figure 5F shows that the water molecules between GO sheets with a d-spacing of 0.8 nm still tended to form rhombusshaped units, similar to the structure between graphene sheets. However, the alignment of these rhombuses was obviously disturbed by the extruding oxygenated groups on GO, leading to a density of 1.3 g/cm3 for the water network formed near the GO plane (Figure 5B), 20% lower than 1.62 g/cm3 in the graphene channel. Nevertheless, the water molecules between GO sheets were still much denser and more patterned than in the bulk environment. The oxygenated functional groups on GO not only disrupt the alignment of water molecules but also significantly affect the mobility of water molecules in the GO channel. Water mobility was studied in terms of mean square displacement of water 6445

DOI: 10.1021/acsnano.7b02999 ACS Nano 2017, 11, 6440−6450

Article

ACS Nano

It is reasonable to attribute the repulsive forces between two GO sheets mainly to the short-distance hydrophilic repulsion (i.e., hydration force) and long-distance electrostatic repulsion, as schematically illustrated in Figure 6C. The hydration force was strong within the first hydration layer, which had a thickness of 0.27 nm based on simulation and was much smaller than the measured d-spacing of more than 2 nm (Figure 6B). With such a d-spacing, simulation results showed that the shell of hydrated ions was not perturbed in the GO channel. Therefore, the dominant repulsive force was the electrostatic repulsion, which can be derived by analyzing the potential distribution in the electrolyte solution between two charged GO sheets. Derivation of relevant equations is provided in the SI. Since the potential between two GO sheets does not decay to zero, their charges are not fully neutralized by the intermediate counterions in between. Thus, a repulsive force should still exist at the middle plate (between the two GO sheets) and is equivalent to the osmotic pressure generated by ions at the same location tending to pull water from the bulk and push the two GO planes away from each other. Therefore, the repulsive pressure can be estimated by the osmotic pressure difference π between the solution at the middle plate and the bulk solution. Derivation of relevant equations can be found in the SI. To reach an equilibrium d-spacing, an attractive force between the two negative GO sheets is needed to overcome the electrostatic/osmotic repulsive force. Since the clear separation distance, Ds, between two GO sheets in the present study is in general beyond the effective range of van der Waals attraction, the dominant attractive force for each GO sheet must be the Coulombic attraction between the negatively charged GO and the intermediate counterions in the solution (Figure 6C). Note that the Coulombic attraction is determined by the electric field and charge density, both of which are affected by the overlapping potential of the two GO sheets and thus cannot be directly calculated without knowing Ds first. To circumvent this difficulty, it is assumed that the potential between two GO sheets is equal to the sum of the potential generated by each individual sheet. As a result, the solution to the original problem is simplified by doubling the attractive force generated by one sheet (with a detailed calculation in the SI), as schematically illustrated in Figure 6D. The attractive electrostatic pressure and the repulsive osmotic pressure are expected to cancel out each other at equilibrium, where the two GO sheets are no longer pulled toward or pushed away from each other. Therefore, the separation distance Ds can be obtained by equating the repulsive osmotic pressure π (eq S9) to the Coulombic attractive pressure P (eq S10). Recall that d-spacing is defined as the center-to-center distance between two adjacent GO sheets. Thus, it consists of the thickness of a whole GO sheet (0.76 nm in a dry state), the first hydration layer on each GO surface (0.27 nm), and the clear separation distance Ds between the two charged GO surfaces. Therefore, given the Ds of two hydrated GO sheets, the theoretical d-spacing can be calculated as

molecules. As shown in Figure 5H, the mobility of water molecules in the graphene channel was almost the same as that in bulk water, indicating that water did not experience any resistance from the graphene sheet during its transport, an observation consistent with the large slip length and theoretical high water transport rate in graphene channels.39−44 In comparison, the mobility of water molecules in the GO channel was much lower than in the graphene channel, indicating considerable resistance to water transport in the GO channel due to the existence of oxygenated functional groups.45 Experimental and Theoretical Study of GO Membrane Swelling in Ionic Solutions. The swelling of the GO membrane in representative ionic solutions, including monovalent ions (NaCl), divalent cations (CaCl2, MgCl2), and anions (Na2SO4), was studied. The level of swelling was represented by the thickness and d-spacing of a GO membrane after the swelling had reached an equilibrium in each solution. As the ionic strengths of NaCl and Na2SO4 increased from 1 mM to 100 mM, the equilibrium thickness and d-spacing of the GO membrane decreased significantly (Figure 6A), as did its mass (Figure S7). At the same time, however, the equilibrium density of the GO membrane increased only slightly, by less than 10% (Figure S8). The above phenomena can be explained using the theory of electrostatic double layer compression by two negatively charged plates (e.g., GO sheets in the present study). First, the Debye length, λD, a measure of the thickness of the diffusive ionic double layer that covers a charged surface in aqueous solution, was calculated using eq S6 as a function of ionic strength and plotted in Figure 6B. Since the corresponding electrostatic double layers were mainly composed of counterions (Na+) in both NaCl and Na2SO4 solutions, neither the Debye length nor d-spacing was affected by different anions (Cl− and SO42−). The overall trend that the Debye length decreases with increasing ionic strength is consistent with that of GO swelling, as shown in Figure 6B. Note that the d-spacing of a GO channel at an ionic strength of 100 mM was ∼2.1 nm, almost twice the Debye length of 1 nm at the same ionic strength, suggesting the two adjacent GO sheets were separated by the electrostatic double layers on the two charged GO surfaces. However, in a highly diluted solution of 1 mM NaCl, the experimental d-spacing of 7.1 nm was much smaller than twice the corresponding Debye length of 9.6 nm, indicating a significant overlap of the electrostatic double layers between the two adjacent GO sheets. In order to mechanistically elucidate the swelling of a GO membrane in aqueous solution, the theoretical d-spacing between two GO sheets was also derived in the present study. For this purpose, it is necessary to completely understand the governing interaction forces between 2D nanomaterials, a key piece of information lacking in the literature. Traditionally, the interactions between two charged particles are described by the balancing effects of van der Waals attraction and electrostatic repulsion, well known as the Derijaguin, Landau, Verwey, Overbeek (DLVO) theory, which has been widely applied to study colloidal stability.46 However, the standard DLVO theory neglects other interaction forces such as Coulombic attraction through intermediate counterions47,48 and hydrophilic repulsion introduced by the hydration potential of a hydrophilic surface.49 In the present study, a parallel-plate model was used to quantitatively describe the repulsive and attractive forces existing between two adjacent GO sheets.

d ‐spacing = Ds + (0.76 nm + 0.27 nm × 2)

(4)

The theoretical d-spacings between two GO sheets in, for example, 5, 50, and 100 mM NaCl solutions are thus calculated to be 5.6, 2.6, and 2.2 nm, respectively, which agree very well with the experimental d-spacings of 5.7, 2.8, and 2.1 nm, respectively (Figure 6B). Such agreement not only demon6446

DOI: 10.1021/acsnano.7b02999 ACS Nano 2017, 11, 6440−6450

Article

ACS Nano

Besides, it has been shown that the oxygenated functional groups on GO significantly affect the structure and transport of water molecules in the GO channel. Therefore, in order to achieve a high-flux GO membrane, it is critical to control both the number and distribution of oxygenate functional groups on GO. However, complete reduction of GO tends to decrease the d-spacing and hence make the membrane almost as impermeable as a pure graphene membrane. Therefore, novel reduction approaches are needed to optimize the presence of oxygenated functional groups remaining on GO, forming a precisely controlled d-spacing for water transport and solute removal.

strates the accuracy of the established theoretical model for calculating the spacing between 2D nanomaterials but also confirms the capability of the integrated QCM-D/ellipsometry method for experimentally quantifying the d-spacing of a membrane made of such 2D nanomaterials. Note that, for a highly diluted solution (e.g., 1 mM NaCl), the theoretical dspacing may differ noticeably from the experimental value, possibly due to the experimental errors stemming from, for example, (i) the extremely slow swelling kinetics that took place in low-concentration NaCl solutions, compared to the fast swelling kinetics at higher ionic strength (Figure S9), and (ii) various factors (e.g., gravity) that overwhelmed the small interaction forces in diluted solutions. The potential and osmotic pressure profiles associated with NaCl solutions of different ionic strengths are plotted in Figure 6D−F. The magnitudes of the electric potential between two GO sheets are very similar for a wide range of ionic strengths between 1 and 100 mM NaCl. However, the osmotic repulsive pressure π at the middle plate between two GO sheets increases significantly from 0.5 to 61 atm as the ionic strength increases from 1 to 100 mM. The electrostatic/osmotic repulsion, Coulombic attraction, and van der Waals attraction (calculated by eq S11) are compared in Figure 6G, with repulsion considered positive and attraction considered negative. It is observed that, within the studied range of d-spacing, the van der Waals attraction is negligible and the interaction force is dominated by the long-distance electrostatic/osmotic repulsion and Coulombic attraction between GO sheets and intermediate counterions. The behavior of GO swelling in CaCl2 and MgCl2 solutions is very different from that in NaCl and Na2SO4 solutions. As shown in Figure 6A, the GO membrane at equilibrium had a constant d-spacing of 6 nm in CaCl2 and 7 nm in MgCl2 solution as the ionic strength increased from 1 mM to 100 mM. This is most likely because the charge screening effect of Ca2+ and Mg2+ ions is much weaker than that of Na+ (counterion);50,51 thus the GO exhibited almost the same level of swelling as in DI water. In addition, Ca2+ and Mg2+ ions likely formed a complex with the carboxylate functional groups on GO and exhibited cation−π interactions with the graphene surface,52−57 thus helping maintain a large d-spacing, regardless of ionic strength. Note that the different coordination complexes formed by Ca2+ and Mg2+ ions may be responsible for the difference in the d-spacings (6 vs 7 nm) of GO membranes soaked in the two types of solutions. Implications for GO Membrane Synthesis. Results from the present study have significant implications for the synthesis of future high-performance GO membranes. It has been revealed that the d-spacing of a GO membrane can be dramatically increased when soaked in aqueous solution, considerably impacting the membrane selectivity and integrity. To resolve this problem, GO membranes can be synthesized in nonaqueous solution in order to minimize the initial membrane swelling and achieve a target d-spacing at the fabrication phase. In addition, strategies should be sought to control the d-spacing (i.e., counteract GO swelling) and stabilize the GO membrane via appropriate cross-linking and synthesis conditions. In particular, because the GO membrane in dry conditions has a typical d-spacing of ∼0.8 nm, a potential approach to making a desalination GO membrane, which requires a d-spacing of ∼0.8 nm, is to cross-link the stacked GO sheets with rigid (as opposed to flexible) spacers using, for example, chemical vapor instead of water as a solvent during membrane synthesis.

CONCLUSIONS The integrated QCM-D/ellipsometry method has demonstrated its capability of accurately measuring the d-spacing of any size within a layer-stacked GO membrane especially when fully soaked in aqueous solutions, while the existing methods (e.g., XRD) can measure a d-spacing of only up to ∼2 nm for a sample typically in dried/semidried conditions. It has been discovered that, when a GO membrane is soaked in pure water, its d-spacing dramatically increases from an initial value of ∼0.8 nm and eventually reaches an equilibrium with a size of as large as 6 to 7 nm. The presence of NaCl and Na2SO4 in an aqueous environment introduces a charge screening effect that compresses the electrostatic double layer, thus dramatically decreasing the d-spacing as the ionic strength increases (e.g., ∼2 nm at 100 mM). In addition, theoretical analysis based on a complete interaction force model has indicated that the equilibrium separation distance between GO sheets is governed by the balance between the osmotic repulsion (in the middle plate) and the Coulombic attraction (between GO and intermediate counterions), while the van der Waals attraction typically considered by the DLVO theory is negligible due to its short-distance nature. It has also been found that the equilibrium d-spacing of a GO membrane in CaCl2 and MgCl2 solutions, irrespective of ionic strength, remains almost the same during the test, suggesting that Ca2+ and Mg2+ generate a much weaker charge screening effect. Furthermore, MD simulations have revealed that the oxygenated functional groups on GO prevent water molecules from forming a perfectly aligned rhombus-shaped crystal structure, which is typically created within a pure graphene channel. Instead, a semiordered water network, with a density of 30% higher than that of bulk water but 20% lower than that in the graphene channel, is formed in the GO channel. In addition, the presence of oxygenated groups on GO has been found to be responsible for much less mobility of water molecules in the GO channel than in the graphene channel, where water mobility is almost the same as that in bulk water. These findings suggest that an optimal strategy for GO membrane synthesis is to cross-link dry GO sheets with rigid spacers to maintain the ideal d-spacing of ∼0.8 nm, followed by GO reduction to create graphene channels for fast water transport. METHODS Chemicals. All chemicals were used as received from SigmaAldrich (St. Louis, MO, USA) unless noted otherwise. The chemicals used in the present study included HCl, H2O2, H2SO4, MgCl2, CaCl2, NaCl, NaNO3, NaOH, Na2SO4, NH3OH, and graphite. Integrated QCM-D/Ellipsometry Measurement. QCM-D (E-1, Q-sense, Sweden) was engineered to be placed on the sample stage of an ellipsometer (FS-1 Multi-wavelength, Film Sense, Lincoln, NE, 6447

DOI: 10.1021/acsnano.7b02999 ACS Nano 2017, 11, 6440−6450

Article

ACS Nano USA). As shown in Figure S1, a sensor was installed in an ellipsometry-specified chamber that has a glass lens on each side to allow the light incident from the source to reflect back to the detector. First, bare gold sensors were tested in the QCM-D chamber to obtain a dry-state baseline in DI water and different testing solutions (NaCl, MgCl2, CaCl2, and Na2SO4). Testing solutions were prepared by dissolving different ionic species in DI water. The solution pH was adjusted by adding a negligible amount of 0.1 mol/L HCl and 0.1 mol/L NaOH standard solutions. A testing solution was flowed at a cross-flow rate of 1 mL/min through the chamber using a peristaltic pump, which was stopped after the chamber had been filled with the testing solution, providing a static aqueous environment to preserve the integrity of the GO membrane by preventing it from partially dissolving into water. The stability of the GO membrane in various solutions is discussed in the SI and demonstrated in Figure S10. QCMD and ellipsometry data were recorded simultaneously with the sensor in a static ambient environment. Then, the sensor was taken out of the chamber, rinsed with DI water to remove residual ionic species, coated with GO, and installed back into the chamber to characterize the swelling properties by soaking the GO-coated sensor in the testing solution. A dry GO-coated sensor was also characterized to evaluate the coating quality. The QCM-D data were analyzed to obtain the mass of the sample by using Q-Tools software (Q-sense, Sweden). The ellipsometry data were analyzed by establishing an optical model. Briefly speaking, data collected for the bare gold sensor were fitted to obtain the dielectric constants (i.e., refractive index and extinction index), and data collected for the GO-coated sensor were fitted using Cauchy’s equation to determine the thickness. More information about data analysis is provided in the SI. Preparation of GO-Coated QCM-D Sensor. GO was prepared from graphite flakes using a modified Hummers method with a detailed procedure described in our previous publication.2 The resulting GO suspension in water was diluted to 0.1 g/L with the pH adjusted to 4. The coating procedure was similar to that described in our earlier work58 with some modifications. As schematically illustrated in Figure 2B, 1 mL of the 0.1 g/L GO water suspension was filtrated through a commercially available poly(ether sulfone) (PES) membrane substrate (Sterlitech, Kent, WA, USA) under a vacuum of 0.8 atm to form a GO thin film. A gold-coated QCM-D sensor (Biolin Scientific, Linthicum Heights, MD, USA) was first cleaned at 75 °C in a solution containing 10 mL of DI water, 1 mL of 30 wt % NH4OH, and 1 mL of H2O2 to remove any potential contaminants, followed by DI rinsing and nitrogen gas drying. The sensor was then placed in a UV ozone chamber (Biolin Scientific, Linthicum Heights, MD, USA) for additional cleaning and to create hydroxyl groups on the sensor surface. The cleaned sensor was wetted with DI water and subsequently placed upside down on the GO-coated PES membrane, with its top surface contacting the GO thin film. The PES membrane along with the sensor was dried in an oven at 60 °C for 15 min. After the water evaporated, the GO thin film bonded strongly to the sensor surface. As the sensor was peeled off from the membrane substrate, the GO thin film was transferred onto the QCM-D sensor. Characterization of GO Thin Film. AFM images were taken to characterize the surface morphology and cross-sectional profile of the GO thin film deposited on the sensor surface. SEM (Ultra-55 FESEM, Zeiss) images were taken for the bare and GO-coated gold sensors, respectively. The zeta potential of GO sheets and surface zeta potential of GO thin film were measured using a Zetasizer Nano-ZSP analyzer (Malvern, Westborough, MA, USA). FTIR spectroscopy (IS50, Thermo Scientific, Fremont, CA, USA) equipped with a specular reflectance accessory (VeeMax III Variable Angle, Pike Technologies, Madison, WI, USA) was used to evaluate the functional groups at different incident angles. XPS (PHI 5400, PerkinElmer) spectra were also obtained to characterize the elemental composition of GO. Molecular Simulations. MD simulations were performed using the Gromacs MD free simulator.59 Supercells used in the simulations consisted of five pairs of graphene or GO sheets with an area of 10 nm × 10 nm and separated by several distances varying from 0.75 to 6.0 nm. All supercells were solvated in a 10 nm × 10 nm × 6 nm simulation box. This system was repeated periodically to represent an

infinite sheet of GO, and water molecules were inserted between the two sheets and the remaining space. The number of epoxy and hydroxyl groups was kept at a ratio of 3:2, and these groups were distributed randomly over both sides of each sheet to give an O/C ratio of 0.23. Note that while the carboxyl and carbonyl groups on the edges were ignored in the simulations, their masses were used afterward to correct the simulated density using eq S3. Water was simulated using the extended simple point charge (SPC/E) water model. The SHAKE algorithm60 was applied to the water model to reduce the high-frequency vibrations of hydrogen bonds. The interactions between GO sheets were modeled using the OPLS-AA force field, which is known to be accurate in modeling a hydrated system.61 The interactions between water and GO were calculated by the Lennard-Jones 6−12 potential between the oxygen and carbon atoms using σC−O = 0.355 nm and εC−O = 0.293 kJ/mol. All electrostatic interactions were calculated using the particle mesh Ewald method.62 The simulations were performed under the NVT canonical ensemble at 293 K. Temperature control was managed using the Nose−Hoover thermostat.63 The system energy was first minimized using the conjugate gradient algorithm and then run at a time step of 0.5 fs.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsnano.7b02999. Additional figures and discussion (PDF)

AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]. ORCID

Jeffrey J. Urban: 0000-0002-6520-830X Baoxia Mi: 0000-0003-3185-1820 Notes

The authors declare no competing financial interest.

ACKNOWLEDGMENTS The material is based upon work supported by the U.S. National Science Foundation under award no. CBET-1565452 and the U.S. Department of Energy under contract no. DEIA0000018. Work at the Molecular Foundry was supported by the Office of Science, Office of Basic Energy Sciences, of the U.S. Department of Energy under contract no. DE-AC0205CH11231. However, the opinions expressed herein are those of the authors and do not necessarily reflect those of the sponsors. The authors thank Dr. Z. Wang at the University of California, Berkeley, for assistance with the XRD analysis. REFERENCES (1) Stankovich, S.; Dikin, D. A.; Dommett, G. H. B.; Kohlhaas, K. M.; Zimney, E. J.; Stach, E. A.; Piner, R. D.; Nguyen, S. T.; Ruoff, R. S. Graphene-Based Composite Materials. Nature 2006, 442, 282−286. (2) Hu, M.; Mi, B. Enabling Graphene Oxide Nanosheets as Water Separation Membranes. Environ. Sci. Technol. 2013, 47, 3715−3723. (3) Nicolai, A.; Sumpter, B. G.; Meuniera, V. Tunable Water Desalination across Graphene Oxide Framework Membranes. Phys. Chem. Chem. Phys. 2014, 16, 8646−8654. (4) Cassagneau, T.; Fendler, J. H. High Density Rechargeable Lithium-Ion Batteries Self-Assembled from Graphite Oxide Nanoplatelets and Polyelectrolytes. Adv. Mater. 1998, 10, 877−881. (5) Huang, J.-Q.; Zhuang, T.-Z.; Zhang, Q.; Peng, H.-J.; Chen, C.-M.; Wei, F. Permselective Graphene Oxide Membrane for Highly Stable and Anti-Self-Discharge Lithium−Sulfur Batteries. ACS Nano 2015, 9, 3002−3011. 6448

DOI: 10.1021/acsnano.7b02999 ACS Nano 2017, 11, 6440−6450

Article

ACS Nano

Pervaporation of Toluenein-Heptane Mixtures. J. Membr. Sci. 2014, 455, 113−120. (27) Jia, Z. Q.; Shi, W. X.; Wang, Y.; Wang, J. L. Dicarboxylic Acids Crosslinked Graphene Oxide Membranes for Salt Solution Permeation. Colloids Surf., A 2016, 494, 101−107. (28) Cao, K.; Jiang, Z.; Zhao, J.; Zhao, C.; Gao, C.; Pan, F.; Wang, B.; Cao, X.; Yang, J. Enhanced Water Permeation through Sodium Alginate Membranes by Incorporating Graphene Oxides. J. Membr. Sci. 2014, 469, 272−283. (29) Rezania, B.; Severin, N.; Talyzin, A. V.; Rabe, J. P. Hydration of Bilayered Graphene Oxide. Nano Lett. 2014, 14, 3993−3998. (30) Jia, Z.; Wang, Y.; Shi, W.; Wang, J. Diamines Cross-Linked Graphene Oxide Free-Standing Membranes for Ion Dialysis Separation. J. Membr. Sci. 2016, 520, 139−144. (31) Hung, W. S.; Tsou, C. H.; De Guzman, M.; An, Q. F.; Liu, Y. L.; Zhang, Y. M.; Hu, C. C.; Lee, K. R.; Lai, J. Y. Cross-Linking with Diamine Monomers to Prepare Composite Graphene Oxide-Framework Membranes with Varying d-Spacing. Chem. Mater. 2014, 26, 2983−2990. (32) Klechikov, A.; Yu, J.; Thomas, D.; Sharifi, T.; Talyzin, A. V. Structure of graphene oxide membranes in solvents and solutions. Nanoscale 2015, 7, 15374−15384. (33) Baimova, Y. A.; Murzaev, R. T.; Dmitriev, S. V. Mechanical Properties of Bulk Carbon Nanomaterials. Phys. Solid State 2014, 56, 2010−2016. (34) Dave, S. H.; Gong, C.; Robertson, A. W.; Warner, J. H.; Grossman, J. C. Chemistry and Structure of Graphene Oxide via Direct Imaging. ACS Nano 2016, 10, 7515−7522. (35) Chen, B.; Jiang, H.; Liu, X.; Hu, X. Observation and Analysis of Water Transport through Graphene Oxide Interlamination. J. Phys. Chem. C 2017, 121, 1321−1328. (36) Lin, L.-C.; Grossman, J. C. Atomistic Understandings of Reduced Graphene Oxide as an Ultrathin-Film Nanoporous Membrane for Separations. Nat. Commun. 2015, 6, 8335. (37) Algara-Siller, G.; Lehtinen, O.; Wang, F. C.; Nair, R. R.; Kaiser, U.; Wu, H. A.; Geim, A. K.; Grigorieva, I. V. Square Ice in Graphene Nanocapillaries. Nature 2015, 519, 443−445. (38) Zhou, W.; Yin, K.; Wang, C.; Zhang, Y.; Xu, T.; Borisevich, A.; Sun, L.; Idrobo, J. C.; Chisholm, M. F.; Pantelides, S. T. The Observation of Square Ice in Graphene Questioned. Nature 2015, 528, E1−E2. (39) Kannam, S. K.; Todd, B. D.; Hansen, J. S.; Daivis, P. J. Slip Length of Water on Graphene: Limitations of Non-Equilibrium Molecular Dynamics Simulations. J. Chem. Phys. 2012, 136, 024705. (40) Cicero, G.; Grossman, J. C.; Schwegler, E.; Gygi, F.; Galli, G. Water Confined in Nanotubes and between Graphene Sheets: A First Principle Study. J. Am. Chem. Soc. 2008, 130, 1871−1878. (41) Gordillo, M. C.; Marti, J. Structure of Water Adsorbed on a Single Graphene Sheet. Phys. Rev. B: Condens. Matter Mater. Phys. 2008, 78, 075432. (42) Gordillo, M. C.; Marti, J. Water on Graphene Surfaces. J. Phys.: Condens. Matter 2010, 22, 284111. (43) Nair, R. R.; Wu, H. A.; Jayaram, P. N.; Grigorieva, I. V.; Geim, A. K. Unimpeded Permeation of Water through Helium-Leak-Tight Graphene-Based Membranes. Science 2012, 335, 442−444. (44) Wei, N.; Peng, X.; Xu, Z. Understanding Water Permeation in Graphene Oxide Membranes. ACS Appl. Mater. Interfaces 2014, 6, 5877−5883. (45) Wei, N.; Peng, X.; Xu, Z. Breakdown of Fast Water Transport in Graphene Oxides. Phys. Rev. E 2014, 89, 012113. (46) Bhattacharjee, S.; Elimelech, M.; Borkovec, M. DLVO Interaction between Colloidal Particles: Beyond Derjaguin’s Approximation. Croat. Chem. Acta 1998, 71, 883−903. (47) Ise, N.; Okubo, T.; Sugimura, M.; Ito, K.; Nolte, H. Ordered Structure in Dilute Solutions of Highly Charged Polymer Lattices as Studied by Microscopy. I. Interparticle Distance as a Function of Latex Concentration. J. Chem. Phys. 1983, 78, 536−540. (48) Sogami, I. Effective Potential between Charged Spherical Particles in Dilute Suspension. Phys. Lett. A 1983, 96, 199−203.

(6) Cho, E. S.; Ruminski, A. M.; Aloni, S.; Liu, Y.-S.; Guo, J.; Urban, J. J. Graphene Oxide/Metal Nanocrystal Multilaminates as the Atomic Limit for Safe and Selective Hydrogen Storage. Nat. Commun. 2016, 7, 10804. (7) Mi, B. X. Graphene Oxide Membranes for Ionic and Molecular Sieving. Science 2014, 343, 740−742. (8) Joshi, R. K.; Carbone, P.; Wang, F. C.; Kravets, V. G.; Su, Y.; Grigorieva, I. V.; Wu, H. A.; Geim, A. K.; Nair, R. R. Precise and Ultrafast Molecular Sieving through Graphene Oxide Membranes. Science 2014, 343, 752−754. (9) Dikin, D. A.; Stankovich, S.; Zimney, E. J.; Piner, R. D.; Dommett, G. H. B.; Evmenenko, G.; Nguyen, S. T.; Ruoff, R. S. Preparation and Characterization of Graphene Oxide Paper. Nature 2007, 448, 457−460. (10) Zhu, Y. W.; Murali, S.; Cai, W. W.; Li, X. S.; Suk, J. W.; Potts, J. R.; Ruoff, R. S. Graphene and Graphene Oxide: Synthesis, Properties, and Applications. Adv. Mater. 2010, 22, 3906−3924. (11) Guo, F.; Creighton, M.; Chen, Y. T.; Hurt, R.; Kulaots, I. Porous Structures in Stacked, Crumpled and Pillared Graphene-Based 3d Materials. Carbon 2014, 66, 476−484. (12) An, Z.; Compton, O. C.; Putz, K. W.; Brinson, L. C.; Nguyen, S. T. Bio-Inspired Borate Cross-Linking in Ultra-Stiff Graphene Oxide Thin Films. Adv. Mater. 2011, 23, 3842−3846. (13) Perreault, F.; Fonseca de Faria, A.; Elimelech, M. Environmental Applications of Graphene-Based Nanomaterials. Chem. Soc. Rev. 2015, 44, 5861−5896. (14) Joshi, R. K.; Alwarappan, S.; Yoshimura, M.; Sahajwalla, V.; Nishina, Y. Graphene Oxide: The New Membrane Material. Appl. Mater. Today 2015, 1, 1−12. (15) Marcano, D. C.; Kosynkin, D. V.; Berlin, J. M.; Sinitskii, A.; Sun, Z. Z.; Slesarev, A.; Alemany, L. B.; Lu, W.; Tour, J. M. Improved Synthesis of Graphene Oxide. ACS Nano 2010, 4, 4806−4814. (16) Park, S.; Lee, K. S.; Bozoklu, G.; Cai, W.; Nguyen, S. T.; Ruoff, R. S. Graphene Oxide Papers Modified by Divalent Ions - Enhancing Mechanical Properties via Chemical Cross-Linking. ACS Nano 2008, 2, 572−578. (17) Luo, Z. T.; Lu, Y.; Somers, L. A.; Johnson, A. T. C. High Yield Preparation of Macroscopic Graphene Oxide Membranes. J. Am. Chem. Soc. 2009, 131, 898−899. (18) Lee, C.; Wei, X. D.; Kysar, J. W.; Hone, J. Measurement of the Elastic Properties and Intrinsic Strength of Monolayer Graphene. Science 2008, 321, 385−388. (19) Wijewardena, U. K.; Brown, S. E.; Wang, X. Q. Epoxy-Carbonyl Conformation of Graphene Oxides. J. Phys. Chem. C 2016, 120, 22739−22743. (20) Degrève, L.; Vechi, S. M.; Junior, C. Q. The Hydration Structure of the Na+ and K+ Ions and the Selectivity of Their Ionic Channels. Biochim. Biophys. Acta, Bioenerg. 1996, 1274, 149−156. (21) Liu, R.; Arabale, G.; Kim, J.; Sun, K.; Lee, Y.; Ryu, C.; Lee, C. Graphene Oxide Membrane for Liquid Phase Organic Molecular Separation. Carbon 2014, 77, 933−938. (22) Compton, O. C.; Cranford, S. W.; Putz, K. W.; An, Z.; Brinson, L. C.; Buehler, M. J.; Nguyen, S. T. Tuning the Mechanical Properties of Graphene Oxide Paper and Its Associated Polymer Nanocomposites by Controlling Cooperative Intersheet Hydrogen Bonding. ACS Nano 2012, 6, 2008−2019. (23) Putz, K. W.; Compton, O. C.; Segar, C.; An, Z.; Nguyen, S. T.; Brinson, L. C. Evolution of Order During Vacuum-Assisted SelfAssembly of Graphene Oxide Paper and Associated Polymer Nanocomposites. ACS Nano 2011, 5, 6601−6609. (24) Abraham, J.; Vasu, K. S.; Williams, C. D.; Gopinadhan, K.; Su, Y.; Cherian, C. T.; Dix, J.; Prestat, E.; Haigh, S. J.; Grigorieva, I. V. Tunable Sieving of Ions Using Graphene Oxide Membranes. Nat. Nanotechnol. 2017, DOI: 10.1038/nnano.2017.21. (25) Mo, Y.; Zhao, X.; Shen, Y. Cation-Dependent Structural Instability of Graphene Oxide Membranes and Its Effect on Membrane Separation Performance. Desalination 2016, 399, 40−46. (26) Wang, N. X.; Ji, S. L.; Li, J.; Zhang, R.; Zhang, G. J. Poly(Vinyl Alcohol)-Graphene Oxide Nanohybrid ″Pore-Filling″ Membrane for 6449

DOI: 10.1021/acsnano.7b02999 ACS Nano 2017, 11, 6440−6450

Article

ACS Nano (49) Cevc, G. Hydration Force and the Interfacial Structure of the Polar Surface. J. Chem. Soc., Faraday Trans. 1991, 87, 2733−2739. (50) Ohki, S.; Sauve, R. Surface Potential of Phosphatidylserine Monolayers. I. Divalent Ion Binding Effect. Biochim. Biophys. Acta, Biomembr. 1978, 511, 377−387. (51) Bara, M.; Guiet-Bara, A.; Durlach, J. A Qualitative Theory of the Screening-Binding Effects of Magnesium Salts on Epithelial Cell Membranes: A New Hypothesis. Magnesium Res. 1989, 2, 243−248. (52) Lee, S.; Elimelech, M. Relating Organic Fouling of Reverse Osmosis Membranes to Intermolecular Adhesion Forces. Environ. Sci. Technol. 2006, 40, 980−987. (53) Chowdhury, I.; Mansukhani, N. D.; Guiney, L. M.; Hersam, M. C.; Bouchard, D. Aggregation and Stability of Reduced Graphene Oxide: Complex Roles of Divalent Cations, pH, and Natural Organic Matter. Environ. Sci. Technol. 2015, 49, 10886−10893. (54) Yun, J. W.; Tanase, T.; Pence, L. E.; Lippard, S. J. CarboxylateBridged and Phosphodiester-Bridged Dinuclear Magnesium(II) Complexes. J. Am. Chem. Soc. 1995, 117, 4407−4408. (55) Larsen, T. M.; Wedekind, J. E.; Rayment, I.; Reed, G. H. A Carboxylate Oxygen of the Substrate Bridges the Magnesium Ions at the Active Site of Enolase: Structure of the Yeast Enzyme Complexed with the Equilibrium Mixture of 2-Phosphoglycerate and Phosphoenolpyruvate at 1.8 Angstrom Resolution. Biochemistry 1996, 35, 4349− 4358. (56) Sun, P.; Zhu, M.; Wang, K.; Zhong, M.; Wei, J.; Wu, D.; Xu, Z.; Zhu, H. Selective Ion Penetration of Graphene Oxide Membranes. ACS Nano 2012, 7, 428−437. (57) Sun, P.; Zheng, F.; Zhu, M.; Song, Z.; Wang, K.; Zhong, M.; Wu, D.; Little, R. B.; Xu, Z.; Zhu, H. Selective Trans-Membrane Transport of Alkali and Alkaline Earth Cations through Graphene Oxide Membranes Based on Cation−π Interactions. ACS Nano 2014, 8, 850−859. (58) Zheng, S.; Mi, B. Emerging Investigators Series: SilicaCrosslinked Graphene Oxide Membrane and Its Unique Capability in Removing Neutral Organic Molecules from Water. Environ. Sci.: Water Res. Technol. 2016, 2, 717−725. (59) Berendsen, H. J.; van der Spoel, D.; van Drunen, R. Gromacs: A Message-Passing Parallel Molecular Dynamics Implementation. Comput. Phys. Commun. 1995, 91, 43−56. (60) Miyamoto, S.; Kollman, P. A. Settle: An Analytical Version of the SHAKE and RATTLE Algorithm for Rigid Water Models. J. Comput. Chem. 1992, 13, 952−962. (61) Jorgensen, W. L.; Tirado-Rives, J. The OPLS [Optimized Potentials for Liquid Simulations] Potential Functions for Proteins, Energy Minimizations for Crystals of Cyclic Peptides and Crambin. J. Am. Chem. Soc. 1988, 110, 1657−1666. (62) Essmann, U.; Perera, L.; Berkowitz, M. L.; Darden, T.; Lee, H.; Pedersen, L. G. A Smooth Particle Mesh Ewald Method. J. Chem. Phys. 1995, 103, 8577−8593. (63) Nosé, S. A Unified Formulation of the Constant Temperature Molecular Dynamics Methods. J. Chem. Phys. 1984, 81, 511−519.

6450

DOI: 10.1021/acsnano.7b02999 ACS Nano 2017, 11, 6440−6450