Emulsion Assembly of Graphene Oxide/Polymer Composite Membranes

Pickering Emulsion Formation.

To demonstrate how the GO-polymer interaction affects the stability of the Pickering emulsions, oil-in-water emulsions were made. 25 Oil-in-water emulsions refer to oil droplets dispersed in water, whereas a water-in-oil emulsion is water droplets dispersed in oil. 26 Pickering emulsions are produced when the oil-water interface is stabilized by solid particles, which in our case is the GO-polymer complex. 27 To prepare the oil-in-water emulsion, we first dissolved TripPEES-CH 2 NH 2 :HCl polymer in an solution of DMF and ortho-dichlorobenzene (o-DCB), or DMF and chloroform, and then we dispersed this oil phase in a 0.04 wt % GO water dispersion. The oil phase of 0.01 wt % polymer and o-DCB:DMF at a 1:4 ratio yielded stable Pickering emulsions of ∼5 μm in diameter (Figure 2B). We obtained a large distribution of emulsion sizes because we used a homogenizer to make the emulsions instead of a microfluidic device, due to the particulate nature of the GO that will clog the latter. Factors that can influence the distribution are the emulsification method, GO concentration, and polymer concentration. 0.01 wt % polymer and DMF:CHCl 3 at a 1:1 ratio as an oil phase were also used to make oil-in-water emulsions, but this produces less stable emulsions than o-DCB emulsions (Figure S7). Larger oil-in-water emulsions were made to observe their interface more closely (Figure S9). There are wrinkles forming at the interface, indicating membrane formation.

Water-in-oil emulsions were also made with a 0.04 wt % GO dispersion as the water phase and a 0.01 wt % polymer solution in chloroform (oil) phase, which yielded stable emulsions of ∼20 μm in diameter. Emulsions with wrinkled interfacial membranes are observed, which are stable for extended periods (Figure 2C). The formation of stable water-in-oil and oil-inwater emulsions with GO:polymer composites confirm the robust nature of this method. To determine the role that the GO and polymer interaction plays in stabilizing the emulsions, we made control emulsions with the same conditions as above, but without the polymer in the oil phase. The chloroform emulsions were unstable with only GO in the water as the stabilizing agent.

Water Filtration Membranes. Figure 3A schematically illustrates membrane fabrication from oil-in-water emulsions. The emulsions, after the excess GO dispersion is removed, Scheme 1. Synthesis of TripPEES-CH 2 NH 2 :HCl form a viscous composition which is spread on the polytetrafluoroethylene (PTFE) polymer substrate with a glass pipet. The emulsions are left to dry in the air for up to 3 h wherein they become unstable and coalesce. Our expectation was that the resulting membrane would have the polymer as the spacer between the GO sheets as a result of the partitioning of the polymer in the particle's interior. Figure 3B shows the optical images of the Pickering emulsion before the top phase of the aqueous GO dispersion is removed. After spreading on the substrate, the emulsion particles are mobile and assemble into an even surface coating. As the water evaporates, the emulsion particles coalesce and evaporation of the oil phase creates a dried film. After drying, the membrane has a smooth surface and observations with an optical microscope confirm the formation of a continuous film with no sign of the original emulsion structure.

We evaluated the role of the oil composition in the Pickering emulsions and the quality of membrane films. When chloroform was used as the oil phase with DMF as cosolvent, we observed bubbling during the drying process and large round features with examination under the microscope (Figure S8). Chloroform has a lower boiling point than o-DCB, and the bubbles observed are the result of evaporation. We have also noticed that chloroform-containing emulsions also tend to be less stable than those with o-DCB, and they likely burst before coalescence. The problem is mitigated by using a vacuum filtration method to produce the membrane. The application of the vacuum removes liquid chloroform before it vaporizes, which prevents bubbling. To verify that both methods yield the same membrane's composition with the same emulsions, we collected the filtrate from the vacuum filtration and analyzed for any polymer that had passed through the filter. Figure S6 shows that no polymer is detected.

The DMF and o-DCB oil composition was optimized to create a homogeneous and pinhole-free film. Figure 4 shows photographs of fabricating membranes/films wherein the oil composition of o-DCB:DMF ranges from 1:1 to 1:2, 1:3, and 1:4. The polymer substrate is PP-backed PTFE, and with increasing DMF there are fewer cracks and eventually the film is homogeneous at the 1:4 ratio. This is observed for deposition on both hydrophobic and hydrophilic PTFE substrates (Figure S10). A similar trend was observed microscopically as well. In Figures S11 andS12, as we increased the DMF concentration, we observed less and less of the vestiges of the original emulsions. The cracks may form in part as a result of the lack of wettability of the oil solution on the substrate, although we could not confirm this hypothesis with contact angle measurements because of the rapidity at which the emulsion particles spread. However, it has been observed previously that compositions that produce uniform crack-free films are those that best wet the substrate, and increased interfacial tension between the solution and the substrate can produce cracked films. 28,29 To make membranes for water filtration testing, we used emulsions with an oil composition of 1:4 o-DCB:DMF. For the casting method, the solvent evaporation rate is 0.5 mL/h at 21 °C in a wellventilated area.

A key structural feature of any graphene-based membrane is the intersheet spacing that typically manifests a characteristic X-ray diffraction (XRD) peak. Figure 5A (and Figure S13) shows the intersheet spacing in membranes with different polymer concentrations. With increasing polymer concentration, the intersheet spacing increases, which is consistent with our structural model of the GO:polymer composite membranes, 2 wherein the polymer behaves as an intersheet spacer. Polymer concentrations in the oil phase above 0.1 wt % result in changes in the membrane's integrity and morphology. Films with the most homogeneous morphology are produced with concentrations less than 0.015 wt %. At higher concentrations, membranes start to show lightly colored spots and can delaminate from the substrate. The polymer is not soluble in water, and as mentioned before, this leads to GO aggregation when the polymer is mixed in a GO dispersion. At higher polymer concentrations, phase separation appears to occur and the light spots are a polymer-rich phase. 30 The membrane's thickness was also investigated with varying polymer concentrations. The amount of the emulsions spread over the support was kept constant at 0.126 mL/cm 2 , and the thickness was measured by the scanning electron microscopy (SEM) imaging of a cross-section prepared by freeze-fracturing a membrane. More details and enlarged SEM images are shown in Figure S14. As detailed in Figure 5B, the film thickness increases with increasing polymer concentration. The changes in both the intersheet spacing and thickness confirm that the polymer is fully incorporated into the membrane and acts as a spacer between GO sheets. As the concentration increases from 0 to 0.01 wt %, the thickness increases linearly with the concentration. However, we see a large jump in thickness as we reach 0.015 wt %, which is consistent with the onset of phase separation.

Another observation from Figure 5A,B is that when we added 0.01 wt % polymer to the membrane, the thickness increased more than 100% but the interlayer spacing increased by 10%. The discrepancy is from the phase separation of the polymer and not all polymer is intercalated between the GO sheets. The XRD measurement shows us how much interlayer spacing is changed due to the intercalation of the polymer but does not quantify how much the polymer is phase separated. The film thickness does reveal the additional material that is incorporated into the membrane but does not reveal the thickness increases resulting from polymer intercalation between the GO sheets relative to the polymer-rich phase.

Homogeneous membrane films were evaluated for the water filtration test using a dead-end cell. We initially analyzed a solution of 100 ppm Rose Bengal (∼1 kDa) at pH 12 to mimic the separation conditions of black liquor from pulp paper manufacturing. 31 All membranes were made with an emulsion loading of 0.126 mL/cm 2 . Figure 6A shows results from the membrane made with 16 mg GO and 0.6 mg polymer loading (0.01 wt %) on a hydrophilic PTFE substrate. The membrane produced a 65% rejection and 2500 g m -2 h -1 bar -1 , and the permeance decreased over time as a result of material buildup in the membrane. 32 Rejection is the percentage of solute removed from the feedwater in a membrane filtration process. It was calculated according to the eq 1 in the Membrane Testing section. The rejection increases steadily throughout the experiment. A polymer concentration higher than 0.01 wt % was used to fabricate thicker membranes. However, the polymer has low solubility in o-DCB, so the increased concentration of polymer over the course of the emulsion coalescence and drying resulted in a nonuniform membrane. From the performance we displayed from Figure 6A, we can confirm that our method can produce a homogeneous and functional filtration membrane.

We were interested to evaluate how the polymer and graphene oxide loading affects the permeance and rejection characteristics. Figure 6B details the performance of films made with and without the polymer. In the composite membranes, the polymer:GO mass ratio is kept constant at 10:1. We used chloroform and DMF for the oil phase, since the polymer has higher solubility in chloroform. A vacuum filtration setup was used to dry the film instead of air-drying because of chloroform's low boiling point as mentioned earlier. The polymer concentration was held constant at 0.015 wt %, since we observed that this yields the most consistent morphology with this oil composition. We tested all the membrane compositions in the dead-end setup for up to 7 days. During this period, we did not observe any large change in rejection or permeance (Figure S15). This shows that all of them maintained working conditions during this period. For the composite membranes, the rejection increased from 30% to 70% as the GO and polymer loading increased. Permeance, on the other hand, decreases from 80 to 25 g m -2 h -1 bar -1 with an increase in GO and polymer loading. Since the polymer concentration is constant for all the composite membranes, the intersheet spacing is invariant in this series. Hence, the trend we observed is the result of the increased thickness of the membrane and increasing layers of GO and polymer. 33,34 We observed a large decrease in rejection going from the films without polymer (50%) to films with polymer (30%). There is a significant increase in permeance in transitioning from GO membranes without polymer (15 g m -2 h -1 bar -1 ) to composite GO:polymer membranes (85 g m -2 h -1 bar -1 ). The performance differences between a pure GO membrane and composite membranes are related to the change in intersheet spacing (Figure 5A) and thickness. In addition, it is likely that the anionic dye, Rose Bengal, adsorbs to our the cationic polymer through electrostatic interactions. 35,36 As we introduce more polymer, the intersheet spacing increases as well as the diffusion path and dye adsorption capacity of the membrane, which leads to a lower rejection and higher permeance as reported. 34,37 To better understand the composite properties, we measured pure water permeance of the composites and GO membranes. Figure 7 compares the permeances between the composite membrane and GO in dye solution and pure water.

Orange lines show the permeance of composite membranes with the same GO:polymer 10:1 mass ratio. For pure water, the permeance does not show a significant variation with increasing material loading. The addition of dye causes the permeance to decrease gradually. This indicates that the water diffusion through the composite membrane is fast and independent of the thickness of the composite membrane. When the dye is added, permeance decreases by more than 30% because absorption of bulky dye molecules prevents water molecules from diffusing through the membrane. As mentioned, Rose Bengal is an anionic dye that can bind to the cationic polymer in the composite membranes through electrostatic adsorption. 35 This further blocks water diffusion pathways in the membrane, leading to a significant reduction in water permeance.

We compared the permeance of pure water between GO membranes and composite membranes (Figure 7, lines with diamond markers). Figure 7 shows that the more open and porous structure of composite membranes allows the water to diffuse faster in comparison to the GO membranes. 35 GO membranes have a lower permeance when compared to composite membranes. Figure 7 reveals that the porous and open structure of the composite membranes allows water molecules to diffuse through the membrane.

■ METHODS Materials. Graphene oxide was purchased from Graphenea in 0.4 wt % 250 mL bottle. Hydrophobic PP-backed PTFE substrates were purchased from Sterlitech, and hydrophilic polypropylene-backed PTFE substrates were purchased from Allpure. The substrate has a pore size of 0.1 μm. Anhydrous toluene was obtained from an INERT PureSolv MD5 solvent purification system and stored under Ar over 4 Å molecular sieves. Anhydrous solvents including dimethylacetamide (DMAc) and 1,1,2,2-tetrachloroethane were purchased from Sigma-Aldrich in SureSeal bottles. Anthracene was recrystallized from ethanol/toluene. Benzoquinone was purified by eluting through a plug of silica using CH 2 Cl 2 as the eluent. All other chemicals were purchased from commercial sources and used as received.

Synthesis. Synthesis of TripPEES.

The synthesis of TripPEES was adapted from a previously reported procedure. 23 Anthracene (3.56 g, 20 mmol, 1 equiv), benzoquinone (2.59 g, 24 mmol, 1.2 equiv), and xylenes (20 mL) were refluxed under N 2 for 4 h. After cooling to rt, the solids were collected by vacuum filtration and washed with cold toluene and methanol. The product was recrystallized from xylenes (∼120 mL) to obtain 9,10dihydro-9,10-[1,2]benzenoanthracene-13,16-dione as a light yellow crystalline solid (4.87 g, 85% yield). All glassware was oven-dried. 9,10-Dihydro-9,10-[1,2]benzenoanthracene-13,16-dione (4.8702 g, 17.009 mmol, 1 equiv), bis(4-fluorophenyl)sulfone (4.3245 g, 17.009 mmol, 1 equiv), and K 2 CO 3 (5.40 g, 39.1 mmol, 2.3 equiv) were placed in a two-neck flask and purged with Ar. The flask was fitted with a short-path distillation apparatus. Anhydrous DMAc (50 mL) and toluene (12 mL) were added. The mixture was heated at 150 °C for 6 h, during which the toluene and water formed from K 2 CO 3 were azeotropically distilled off. Afterward, the distillation apparatus was replaced with a reflux condenser, and the mixture was refluxed at 180 °C for 18 h. The mixture was precipitated in warm water, and the solids were collected by vacuum filtration, washed with water, and dried in a vacuum oven at 60 °C. The dried polymer was redissolved in DMAc and precipitated in warm water again, and the solids were collected by vacuum filtration, washed with water, and dried in a vacuum oven at 60 °C. The dried polymer was dissolved in chloroform and precipitated in methanol, and the solids were collected by vacuum filtration, washed with methanol, and dried in a vacuum oven to obtain an off-white solid (7.75 g, 91% yield). The chloromethylation of TripPEES was performed following a previously reported procedure. 38 TripPEES (1.50 g, 3.0 mmol, 1 equiv of repeating units) was dissolved in anhydrous 1,1,2,2-tetrachloroethane (50 mL) under Ar. Chloromethyl methyl ether (6.84 mL, 90 mmol, 30 equiv) was added, followed by ZnCl 2 (1.23 g, 9 mmol, 3 equiv) under an Ar flow. The mixture was stirred at 60 °C for 5.5 h for a functionalization density (x) of 1.2 (x can be tuned by the reaction time). After the reaction, the polymer solution was precipitated in methanol, and the precipitate was collected by vacuum filtration and washed thoroughly with methanol, water, and then methanol again. The polymer was dried under vacuum to give an offwhite solid (90-95% yields typically obtained). The level of functionalization (x) was determined using the NMR integration ratios of the ArCH 2 Cl protons and the bridgehead CH protons. 1 H NMR (400 MHz, CDCl 3 ): δ 7.90 (m, 4H), 6.92 (m, ∼ 11H), 6.69 (br s, 2H), 5.42 (m, 2H), 4.37 (m, 2.4H) (for x = 1.2; integration values vary based on x).

Gabriel Amine Synthesis to Obtain TripPEES-CH 2 NH 2 .

TripPEES-CH 2 Cl (1.54 g, x = 1.2, 2.75 mmol of polymer backbone repeating units, 3.3 mmol of benzyl chloride groups) was dissolved in anhydrous DMAc (60 mL) under Ar. Potassium phthalimide (1.83 g, 9.9 mmol, 3 equiv vs benzyl chloride) and NaI (49 mg, 0.33 mmol, 0.1 equiv vs benzyl chloride) were added under Ar flow. The mixture was stirred at 45 °C for 18 h. After the reaction, the mixture was precipitated in 10:1 water/methanol, and the precipitate was collected by vacuum filtration, washed with water and methanol, and dried under vacuum to give TripPEES-CH 2 phthalimide as a white solid that was used directly for the next step. TripPEES-CH 2 phthalimide was dissolved in dioxane (50 mL). To the polymer solution was added a solution of hydrazine monohydrate (2.4 mL, 50 mmol, 15 equiv vs benzyl chloride) in dioxane (5 mL). The mixture was heated at 90 °C for 18 h, during which it became a white dispersion. After the reaction, the mixture was precipitated in 10:1 0.1 M NaOH/methanol, and the precipitate was collected by vacuum filtration to give TripPEES-CH 2 NH 2 as a white solid (1.32 g, 92% yield over two steps). The product was stored in the freezer to impede gradual cross-linking. 367 mg of TripPEES-CH 2 NH 2 was dissolved in ∼7 mL of DMF, and 1.6 mL of 1 M HCl (∼2 equiv relative to amines) was added dropwise with stirring. The solution was then dialyzed against water and lyophilized to give a slightly brown fluffy solid. Emulsion Preparation. Oil-in-water emulsions were made with 0.04 wt % GO as the water phase. The oil phase was a solution of o-DCB and DMF in volume ratios of 1:1, 1:2, 1:3, and 1:4 or a solution of DMF and chloroform in a volume ratio of 1:1. The polymer was dissolved in DMF at different weight concentrations. To make the dispersions, 8 mL of oil was added to 40 mL of 0.04 wt % GO. Homogenizer (Tissue Tearor Model 985370) was used to mix the two phases.

Membrane Fabrication. The emulsion was left to settle for 10 min. Most of the particles fell to the bottom of the water solution. The excess water phase was removed, which created a viscous paste. This was spread onto the substrate with a pipet and left to dry in air.

Membrane Testing. Dead-end cells (HP4750) were purchased from Sterlitech. 250 mL of a 0.01 M NaOH solution (pH = 12) and a 100 ppm Rose Bengal solution were poured slowly into the cell to prevent breakage of the membrane due to fast water flow. Nitrogen gas with a pressure of 75 psi was used. The permeate was collected in a 20 mL vial. To calculate the permeance, the weight of permeate was (2)

where I is the UV-vis absorbance at ∼548 nm, the conductivity, or the refractive index, V is the volume, n is the sample number, p is the permeate, r is the remainder, and f is the feed.