Graphene Oxide-Based Membrane for Liquid and Gas

GRAPHENE OXIDE-BASED MEMBRANE FOR LIQUID AND GAS

SEPARATION

A Dissertation

Presented to

The Graduate Faculty of The University of Akron

In Partial Fulfillment

of the Requirements for the Degree

Doctor of Philosophy

Han Lin

August 2020 GRAPHENE OXIDE-BASED MEMBRANE FOR LIQUID AND GAS

SEPARATION

Han Lin

Dissertation

Approved: Accepted:

Advisor Department Chair Dr. Jiahua Zhu Dr. Michael Cheung

Committee member Interim Dean of the College Dr. George Chase Dr. Craig Menzemer

Committee member Dean of the Graduate School Dr. Jie Zheng

Committee member Date Dr. Siamak Farhad

Committee member Dr. Xiong Gong

ii Dedicated to My beloved parents and family members who helped me all things great and small.

Without your endless love and support, none of my success would be possible.

iii ACKNOWLEDGEMENTS

I would like to express my special appreciation and deepest thanks to my advisor

Dr. Jiahua Zhu, for the patient guidance, encouragement and advice he has provided in the past four years. Your advice on both research as well as on my career have been invaluable.

I would also like to thank my committee members, Dr. George Chase, Dr. Jie

Zheng, Dr. Siamak Farhad, and Dr. Xiong Gong for their valuable comments and suggestions that improve the quality of the work.

I am grateful to all of labmates, Dr. Liwen Mu, Dr. Long Chen, Dr. Tuo Ji, Dr.

Nitin Mehar, Dr. Marjanalsadat Kashfipour, and Mr. Yifan Li for their help during my struggling time. I really enjoyed time we worked together, as well as had coffee together.

I would like to thank my best friends, Jiahui Wang, Chengwei Polly Zhou, and

Hao Zhang for our unbelievable friendship for past 12 years. I will never forget the most precious time that we studied together at the library. I would also like to thank Dr.

Tao Liu for the great suggestions and timely encouragement on both research and daily life.

Last but not least, I would like to express my vehement protestations of gratitude to my parents, Jianyu Lin and Chengxia Wu, for their endless love and unconditional support. Without their sacrifices, I could never receive the best education and finish this dissertation.

iv ABSTRACT

Graphene oxide (GO), a 2-dimensional material, has attracted great attention in membrane research in the past decades due to its intrinsic physicochemical properties, such as the tunable surface functionality, excellent chemical inertness, and thermal stability. With a similar 2-D structure as graphene, the rich surface functional groups at both edge area and basal plane enable a vast opportunity to design the membrane functionality. However, there are still a few key challenges need to be addressed before

GO based membranes can be practically used including but not limited to weak mechanical strength at nanoscale thickness, low membrane integrity due to swelling and limited approaches to tune d-spacing. Although it has been demonstrated that d- spacing, the interlayer distance of GO nanosheets, is a very important parameters that influence the separation performance of GO based membranes, the d-spacing, surface charge, wrinkling effect, and size of GO nanosheets are also important factors for excellent separation performance.

Here, we presented three different approaches, that were utilized for fabricating stable GO based membranes with excellent gas and liquid separation performance. First, epoxide ring-opening reaction and subsequent modification with oxalic acid (OA) were proceeded to enrich the functional groups on the basal plane of GO sheets. By facilitating the plane-plane connection of GO sheets with activated in-plane groups, better packing (reduced wrinkling) of GO membrane would be expected and thus superior separation performance. GO membranes were fabricated by pressure-assisted

v filtration method on porous polymer support. Membrane thickness was simply controlled by the amount of filtration solution. Cross-linking was introduced in the membrane to improve the mechanical strength as well as reduce swelling in aqueous media. The GO modification, cross-linking reaction, microstructure and wrinkle structure of GO membrane were systematically characterized. The prepared membranes were tested in various separation applications including H2/CO2 separation, organic dye separation from water, and desalination. The prepared membrane showed good permselective H2/CO2 separation with a separation factor of 16.14 and H2 permeance of 13.6 × 10-9 mol.m-2s-1Pa-1. In MB separation tests, the rejection rates of 70 ppm MB solution of OAGO/EDA-1 and OAGO/EDA-2 membranes achieved nearly 100% under

+ + - 2- 2+ 300 psi. The overall high ions (Na , K , Cl , and SO4 and Mg ) rejection rates of >

98.1% were also obtained in a pervaporation desalination system. This work offers a new strategy to fabricate modified GO membrane with strong plane-plane interaction and better ordered microstructure.

Secondly, a controlled pre-cross-linking method is developed to address the swelling issue and d-spacing control simultaneously. Cross-linking and GO reduction are coupled in the pre-cross-linking reaction. Hybrid GO/rGO membranes with well- patterned layer stacking structure is fabricated by vacuum filtration method. The crosslinking reaction and microstructure evolution of the GO membranes are systematically characterized by X-ray photoelectron spectroscopy, Raman spectroscopy, scanning electron microscope, and X-ray diffraction. The properties of the prepared membranes are tested in both gas and liquid separation systems. The H2/CO2 separation is measured with a flow-work gas separation system, and desalination test was carried out in a pervaporation system. The hybrid membrane shows excellent permselective H2/CO2

vi - separation with a separation factor of 22.93±1.57 and H2 permeance of 2.46±0.01× 10

8 -2 -1 -1 + + 2+ - 2− mol.m s Pa . Extremely high ion (Na , K , Mg , Cl , and SO4 ) rejection rate of >99% is also obtained in a pervaporation desalination system. This work offers a new strategy to fabricate hybrid GO/rGO membrane with small d-spacing and excellent stability particularly in aqueous environment.

Thirdly, amino boron nitride (a-BN) nanosheets were prepared from bulk boron nitride (BN) by ball milling method with urea as agent. By cross-linking GO nanosheets with a-BN nanosheets, the a-BNGO hybrid membrane with narrow d-spacing and long- term stability in liquid separation applications would be expected and thus advanced methylene blue (MB) separation performance. The a-BNGO hybrid membranes were fabricated by pressurization filtration method on porous polyethersulfone (PES) substrate with polydopamine (PDA) coating. Membrane thickness can be simply controlled by the amount of GO and a-BN nanosheets. A two-step cross-linking process was processed to ensure the covalent bonds connection, one was applied in a-BNGO mixture solution before membrane fabrication and another was applied after membrane was formed. The a-BN modification, cross-link reaction, microstructure and surface morphology and properties were characterized systematically. The prepared a-BNGO membranes were tested in MB separation from water under different pH conditions.

The a-BNGO membrane showed the best result with rejection rates of 99.98% and water flux of 4.15 LMH within first 3-hour period and only slight drops were observed for the following second and third 3-hour periods. The a-BNGO membranes also showed a wide pH range of operation, from 4.0 to 10.0, without sacrificing any performance. This work provides a new choice of cross-linker to fabricate non-swelling

GO membranes.

vii At last, the effect of GO nanosheet lateral size on membrane microstructure and liquid separation performance was investigated systematically. The EDA cross-linked

GO membranes were fabricated by three groups of GO nanosheets with different lateral sizes, which was simply separated by centrifugation method. The differently sized GO nanosheets form unique stacking patterns that leads to varied separation performance, especially in different pH conditions. All the membranes showed excellent long-term

(24 hours) stability in neutral (pH=7), acidic (pH=4) and basic (pH=10) conditions. MB separation performance has been found closely related to GO lateral size at specific pH environments. Specifically, the highest rejection rate in GO/EDA-S, GO/EDA-M, and

GO/EDA-L membranes has been found at pH=4, 7 and 10 conditions, respectively. For

MO, the separation is less effective relatively as compared to MB, while rejection rate of > 96.0% can be still achieved on all three membranes. Rejection rate of > 99.0% for

Cr (VI) was accomplished on GO/EDA-S and GO/EDA-M membranes taking advantage of narrowed d-spacing and enlarged ion size in acidic environment.

viii TABLE OF CONTENT LIST OF TABLES ...... xi I. INTRODUCTION ...... 1 II. BACKGROUND ...... 4 2.1. GO synthesis and its chemical property...... 4 2.2. Separation mechanism of GO-based membranes ...... 5 2.3. Fabrication of GO based membranes ...... 7 2.3.1. Filtration method ...... 7 2.3.2. Layer-by-layer assembling method...... 10 2.4. Swelling issue and its strategies ...... 12 2.4.1. GO swelling ...... 12 2.4.2. Swelling control by non-covalent ...... 13 2.4.3. Swelling control by covalent bonds ...... 15 2.4.4. Physically confined GO with vertical alignment ...... 16 2.5. Fundamentals of GO Membrane Separation...... 17 2.6. Gas Separation ...... 19

2.6.1. CO2/N2 Separation ...... 21

2.6.2. CO2/CH4 Separation...... 22

2.6.3. CO2/H2 Separation ...... 22 2.7. Liquid Separation ...... 24

III. PERMSELECTIVE H2/O2 SEPARATION AND DESALINATION OF HYBRID GO/rGO MEMBRANES WITH CONTROLLED PRE-CROSSLINKING ...... 30 3.1. Outline...... 30 3.2. Introduction ...... 30 3.3. Experimental Procedures ...... 33 3.3.1. Materials ...... 33 3.3.2. GO synthesis ...... 33 3.3.3. Membrane fabrication ...... 34 3.3.4. Characterization ...... 36 3.4. Results and Discussion ...... 39

ix 3.5. Conclusion ...... 52 IV. REDUCED WRINKLING IN GO MEMBRANE BY GRAFTING BASAL- PLANE GROUPS FOR IMPROVED GAS AND LIQUID SEPARATIONS ...... 54 4.1. Outline...... 54 4.2. Introduction ...... 54 4.3. Experimental Procedures ...... 57 4.3.1. Materials ...... 57 4.3.3. Membrane fabrication ...... 58 4.3.4. Characterization ...... 60 4.4. Results and Discussion ...... 62 4.5. Conclusion ...... 71 V. GRAPHITE OXIDE/BORON NITRIDE HYBRID MEMBRANE: THE ROLE OF CROSS-PLANE LAMINAR BONDING FOR A DURABLE MEMBRANE WITH LARGE WATER FLUX AND HIGH REJECTION RATE ...... 73 5.1. Outline...... 73 5.2. Introduction ...... 73 5.3. Experimental Procedures ...... 77 5.3.1. Materials ...... 77 5.3.2. Preparation of GO and a-BN nanosheets ...... 77 5.3.3. Membrane fabrication ...... 78 5.3.4. Characterization ...... 81 5.4. Results and Discussion ...... 82 5.5. Conclusion ...... 94 VI. CROSS-LINKED GO MEMBRANES ASSEMBLED WITH GO NANOSHEETS OF DIFFERENTLY SIZED LATERAL DIMENSIONS FOR ORGANIC DYE AND CHROMIUM SEPARATION ...... 96 6.1. Outline...... 96 6.2. Introduction ...... 96 6.3. Materials and Methods ...... 99 6.3.1. Materials ...... 99 6.3.2. Preparation of GO with different lateral sizes ...... 100 6.3.3. Membrane fabrication ...... 101

x 6.3.4. Characterization ...... 102 6.4. Results and Discussion ...... 104 6.5. Conclusion ...... 124 REFERNECES ...... 126

xi LIST OF FIGURES

Figure 1. Representative structure of graphene oxide…………………………...……..5

Figure 2. (A) The pathway of water molecules and small ions in GO membrane. Larger molecules are blocked. (B) Ajustable nanochannal size for separation purposes. (C) Several GO membrane fabrication methods that have been reported…….….…………7

Figure 3. (A) Schematic of pressure-assisted self-assembly method to fabricate GO/mPAN membranes. (B) Surface SEM images of mPAN substrate (a) and GO/mPAN membranes (b, c); cross-section SEM images of corresponding three samples (d-f). Used GO amount: (b, and e) 8.6×10-5 g*cm-2; (c, and f) 17.3×10-5 g*cm-2………………………………………………………………..………………..8

Figure 4. Schematics of PASA, VASA, and EASA methods for GO/mPAN membrane fabrication………………………………………………………………………….…..9

Figure 5. Digital images, surface and cross-section SEM images of GO/mPAN membranes fabricated by PASA (a, d, g), VASA (b, e, h) and EASA (c, f, i) method…9

Figure 6. Schematic of (a) a procedure to fabricate GO membrane by repeated LbL deposition, (b) the structures of polydopamine and TMC, and the reaction mechanism between them, and (c) the reaction mechanism between GO and TMC……………..11

Figure 7. Schematic of (a) conventional LbL self-assembly process and (b) applied electric field involved LbL self-assembly process………………………………..….12

Figure 8. The schematic of multiple interfacial reactions in hybrid GO membrane….13

Figure 9. (a) The comparison of interlayer spacings of GOMs after immersing in water and 0.25 M salt solutions. (b) The comparison of Na+, Ca2+ and Mg2+ permeation rates of untreated GO-750 and KCl treated GO-750 membrane. Dashed lines are the detection limits of Na+, Ca2+ and Mg2+ cations…………………………………….15

ix Figure 10. Structural diagram of GO and three composite GOF membranes (GO-EDA, GO-BDA, and GO-PPD). Each GOF was produced by cross-linking GO with a diamine (EDA, BDA, or PPD)………………………………………………………………....16

Figure 11. (a) Schematic of separation mechanism of physically confined GO membranes. (b) digital images of a PCGO sample that glued in a plastic disk. Scale bar, 5 mm. (c) optical cross-section micrograph of PCGO. GO laminates (black) were embedded in epoxy (light yellow). (d) SEM image of GO laminates that marked in red rectangle in (c). Scare bar, 1 µm. (e) interlayer-spacing under different humidity. Inset: XRD results that used to calculate interlayer-spacing………………………………..17

Figure 12. (a) Structures of HPEI and TMC, and the reaction between them. (b) permeances of CO2 and N2 in term of GPU. (c) the relation between the selectivity of CO2/N2 and the GO content………………………………………………………….21

Figure 13. (a) Schematic illustration of fabrication 2D channels by an external force. (b) the comparison on H2/CO2 separation performance between EFDA-GO membranes and other state-of-the-art gas separation membranes…………………….23

Figure 14. (A) Separation result of 50:50 H2/CO2 gas mixture and (B) separation result of 50:50 H2/N2 gas mixture…………………………………………………….……24

Figure 15. (a) Schematic of the fabrication process of polycation/GO composite membrane on PAN substrate by LbL self-assembling method. (b) 100 mg/L methyl blue solution separation performance of PDDA/GO and PDDA/PAA membranes under different dye concentrations at 5 bar………………………………………………….26

Figure 16. (a) Ion permeation fluxes of diamines cross-linked GO membranes, (b) separation factor of K+/Mg2+ of diamines cross-linked GO membranes…………….27

Figure 17. (a) Schematic of synthesizing graphene oxide framework. (b) water flux and ion rejection results of 3.5 wt.% seawater desalination……………………………….28

Figure 18. The comparison of antibacterial performance between GO/Ag membrane (grey) and pure GO membrane (blank)……………………………………………….29

x Figure 19. Schematic diagram of membrane fabrication process. (a) pure alumina substrate, (b) PDA coated alumina substrate, (c) vacuum filtration method and partial cross-linked GO/EDA dispersion, (d) structure of partial cross-linked GO/EDA, (e) GO/EDA membrane on PDA coated alumina substrate and (f) mixed stacking of GO/rGO sheets……………………………………………………………………….36

Figure 20. Schematic diagram showing the apparatus for H2/CO2 separation……...38

Figure 21. Digital images of GOEDA solution with pre-crosslinking for (a) 1, (b) 2, (c) 6, (d) 24 h and (e) solutions after 24 h settlement…………………………………….40

Figure 22. XPS wide scan of (a) pure GO, (b) GOEDA-0, (c) GOEDA-1and (d) GOEDA-2 membranes…………………………………………………………….....42

Figure 23. Schematic of reaction between GO and EDA. (a) pure GO; (b) GO/EDA-0; (c) GO/EDA-1; (d) GO/EDA-2……………………………………..………………...42

Figure 24. XPS elemental analysess of (a) pure GO, (b) GOEDA-0, (c) GOEDA-1, and (d) GOEDA-2 membranes in C 1s……………………………………………………44

Figure 25. Raman spectra of (a) pure GO, (b) GO/EDA-0, (c) GO/EDA-1 and (d) GO/EDA-2………………………………………………………………………...…45

Figure 26. Digital images of (a) pure GO, (b) GO/EDA-0, (c) GO/EDA-1 and (d) GO/EDA-2…………………………………………………………………………...46

Figure 27. XRD profiles of (a) GO (b) GO/EDA-0, (c) GO/EDA-1 and (d) GO/EDA-2. Table summarizes the d-spacing value of each diffraction peak……………………..47

Figure 28. Reactions involved in the membrane fabrication of (a) GO/EDA and (b) OAGO/EDA……………………………………………………………………….....59

Figure 29. Scheme of membrane fabrication procedure. (a) PES membrane before and after dopamine polymerization, (b) separation cell with O-rings, (c) sealed PES film and prepared GO solution in separation cell under pressure, and (d) final GO/EDA (or OAGO/EDA) membrane………………………………………………………….….60

xi Figure 30. Cross section SEM images of (a) GO/EDA-0.5, (b) GO/EDA-1, (c) GO/EDA-2, (d) OAGO/EDA-0.5, (e) OAGO/EDA-1, and (f) OAGO/EDA-2……...62

Figure 31. XPS wide scan of (a) pure GO, (b) GO/EDA-2, and (c) OAGO/EDA-2 membranes. The atomic percentages of each membrane (C, N and O) are at the top right corner………………………………………………………………………………....63

Figure 32. XPS elemental analyses of (a) pure GO, (b) GO/EDA-2, and (c) OAGO/EDA-2 membranes in C1s…………………………………………………....64

Figure 33. XRD results of (a) pure GO, (b) GO/EDA-2, and (c) OAGO/EDA-2 membranes…………………………………………………………………………...65

Figure 34. Schematic of microstructure of (a) GO/EDA and (b) OAGO/EDA membrane………………………………………………………………………...…..66

Figure 35. Surface AFM images of (a) GO/EDA-0.5, (b) GO/EDA-1, (c) GO/EDA-2, (d) OAGO/EDA-0.5, (e) OAGO/EDA-1, and (f) OAGO/EDA-2 membranes………67

Figure 36. Height profiles of selected membrane area in Figure 35………………….67

Figure 37. MB separation results of GO/EDA membranes as a function of pressure (a- b) and membrane thickness (c-d)……………………………………………………..69

Figure 38. MB separation results of OAGO/EDA membranes as a function of (a-b) pressure and (c) membrane thickness………………………………………………...69

Figure 39. AFM images of (a) BN nanosheets and (b) GO nanosheets, and (a1,a2,b1,b2) their corresponding height profiles in nm……………………………………………..78

Figure 40. The schematic structure of (a) GO, (b) a-BN nanosheets, (c) crosslinking reaction between GO and a-BN, and (d) scheme of compact laminar structure of a- BNGO membrane…………………………………………………………………….80.

Figure 41. Schematic process of membrane fabrication. (a) PES membrane before and after PDA coating, (b) separation cell with PDA/PES between two O-rings, (c) sealed

xii PDA/PES and prepared a-BNGO solution in separation cell under 300 psi pressure, and (d) final a-BNGO membrane…………………………………………………..….….81

Figure 42. The FT-IR spectrum of (a) BN and a-BN, and (b) a-BN, GO and a- BNGO………………………………………………………………………………...82

Figure 43. (a) XPS wide scan of pure GO and a-BN2.0GO. The XPS elemental analyses of (b) a-BN2.0GO in N 1s, (c) pure GO in C 1s, and (d) a-BN2.0GO in C 1s……….84

Figure 44. The digital images of (a) pure a-BN membrane and (b) 70 ppm MB solution before and after separation. The surface SEM image of pure a-BN membrane……...85

Figure 45. The digital images of a-BN1.0GO and a-BN2.0GO solutions (a) before and (b) after pre-cross-linking and (c) the corresponding membranes…………………….86

Figure 46. Cross-section SEM images of (a) pure GO, (b) a-BN0.25GO, (c) a-BN0.5GO, (d) a-BN1.0GO, (e) a-BN1.5GO, and (f) a-BN2.0GO……………………………….87

Figure 47. XRD results of (a) pure GO, (b) a-BN0.5GO, (c) a-BN1.0GO, (d) a- BN1.5GO, and (e) a-BN2.0GO……………………………………………………….88

Figure 48. AFM morphology images and line profiles of (a) a-BN0.5GO, (b) a- BN1.0GO, (c) a-BN1.5GO, and (d) a-BN2.0GO…………………………………….89

Figure 49. Contact angle result of pure GO, a-BN0.25GO, a-BN0.5GO, a-BN1.0GO, a-BN1.5GO, and a-BN2.0GO………………………………………………………...91

Figure 50. Rejection rate (black) and water flux (blue) of 70 ppm MB separation tests of hybrid a-BNGO membranes. The numbers of 3, 6, and 9 represent tine period of 0- 3 hours, 3-6 hours, and 6-9 hours…………………………………………………….93

Figure 51. Rejection rate (black) and water flux (blue) of 70 ppm MB separation tests of a-BN1.0GO membrane. The numbers of 3, 6, and 9 represent tine period of 0-3 hours, 3-6 hours, and 6-9 hours……………………………………………………………..94

xiii Figure 52. The digital images of GO solutions after centrifugation (a) 500 rpm for 10 mins, (b) 2000 rpm for 20 mins, and (c) 5000 rpm for 30mins. And correspond solutions after dispersing in DI water correspond (d-f)……………………………………….101

Figure 53. Scheme of membrane fabrication process. The digital images of (a) pure PES, (b) PDA coated PES, and (c) GO/EDA (dark brown) on PDA/PES. The PDA/PES was fixed by two O-rings, then GO/EDA solution was added in the cell. After draining with 300 psi pressure, supported GO/EDA membrane was fabricated……………..102

Figure 54. AFM morphology of (a) GO-S, (b) GO-M and (c) GO-L. (d) Distribution of the largest length measured in 2D dimension of GO-S, GO-M and GO-L. The AFM images were collected by contact mode with scanning dimension of 45*45 μm. Sample size: 15………………………………………………………………..……………..105

Figure 55. The FT-IR spectrum of (a) GO-S, (b) GO-M, and (c) GO-L…………….106

Figure 56. High resolution XPS C 1s analysis of pristine GO nanosheets. (a) GO-S, (b) GO-M, and (c) GO-L. The four deconvoluted peaks are C=O, C-O-C, C-O, and C=C…………………………………………………………………………………107

Figure 57. XPS scan of pristine GO nanosheets and GO/EDA membranes. (a) GO-S, (b) GO-M, (c) GO-L, (d) GO/EDA-S, (e) GO/EDA-M, and (f) GO/EDA-L. The atomic percentages of C, O, and N were listed below each curve…………………………..109

Figure 58. High resolution XPS N 1s analysis of (a) GO/EDA-S, (b) GO/EDA-M, and (c) GO/EDA-L. The four deconvoluted peaks are NH3+-C, N-C(O), H2N-C, and N- H……………………………………...……………………………………………..110

Figure 59. Cross section SEM images of (a) GO/EDA-S, (b) GO/EDA-M, and (c) GO/EDA-L. Thickness was controlled to be about 500 nm by adjusting the amount of GO/EDA solution…………………………………………………………………...110

Figure 60. Surface morphology images by AFM (contact mode) of (a) GO/EDA-S, (b) GO/EDA-M, and (c) GO/EDA-L; and corresponding height profiles of the selected area (red and green lines). Marked lines were selected to across both brightest and darkest (highest and lowest) part of each sample………………………………………..….112

xiv Figure 61. XRD results of three GO/EDA membranes (a) at dry state, and after 24 hours soaking in (b) pH=7, (c) HCl solution, pH=4, and (d) KOH solution, pH=10. The corresponding d-spacing values were marked in plot, which was calculated by Bragg’s Law………………………………………………………………………………….114

Figure 62. Scheme of GO/EDA membrane with (a) compacted laminar structure at dry state and (b) enlarged d-spacing at aqueous environment. (c) Scheme of length of EDA…………………………………………………………………………………115

Figure 63. Rejection rate (black) and water flux (blue) of MB removal in dead-end filtration system at room temperature under different pH conditions. (a) pH=7, (b) pH=4, and (c) pH=10. The pH was adjusted by HCl and KOH. Initial concentration of MB solution is 70 ppm. 3, 6, 9 in plots represents time period of 0-3 h, 3-6 h and 6-9 h, respectively. Digital images were provided in Figure 64……………………………117

Figure 64. Digital images of MB solutions before and after (3, 6, and 9 hours) separation tests under (a) pH=7, (b) pH=4, and (c) pH=10 conditions………………………….117

Figure 65. Digital image of MB solution before (70 ppm) and after (6 h and 9 h) separation test……………………………………………………………………….118

Figure 66. Rejection rates (black) and water flux (blue) of long-term (24 hours) MB separation tests under various pH conditions. (a) pH=4, (b) pH=7, and (c) pH=10. The pH was adjusted by HCl and KOH. The numbers represent the time periods of separation tests (e.g. 3 means 0-3 h, 6 means 3-6 h)…………………………….….118

Figure 67. Rejection rate (black) and water flux (blue) of MO removal in dead-end filtration system at room temperature under different pH conditions. (a) pH=7, (b) pH=4, and (c) pH=10. The pH was adjusted by HCl and KOH. Initial concentration of MO solution is 70 ppm. Digital images of solutions were provided in Figure 68……….120

Figure 68. Digital images of MO solutions before and after (3, 6, and 9 hours) separation tests under (a) pH=7, (b) pH=4, and (c) pH=10 conditions………….….120

Figure 69. Rejection rate (black) and water flux (blue) of Cr (VI) removal in dead-end filtration system at room temperature under different pH conditions. (a) pH=7, (b) pH=4,

xv and (c) pH=10. The pH was adjusted by HCl and KOH. Initial concentration of Cr (VI) solution is 1 ppm. Digital images of colored Cr (VI) solutions were shown in Figure 70…………………………………………………………………………………....123

Figure 70. Digital images of colored Cr (VI) solutions before and after (3, 6, and 9 hours) separation tests under (a) pH=7, (b) pH=4, and (c) pH=10 conditions……...123

xvi LIST OF TABLES

Table 1. Gas separation membranes for CO2 separation reported by references…...…20

Table 2. XPS elemental analyses of pure GO and GO with pre-crosslinking…………43

Table 3. H2/CO2 binary gas separation results with pure GO membrane, GO/EDA-0, GO/EDA-1, and GO/EDA-2, respectively…………………………………………...49

Table 4. Radii of hydrated ions existing in seawater………………………………….50

Table 5. Comparison of desalination with pervaporation and nanofiltration on GO/EDA-2…………………………………………………………………………...50

Table 6. Pervaporation desalination of GO/EDA-1 and GO/EDA…………………...51

Table 7. Pervaporated ions removal results of GO/EDA-2 with membrane regeneration………………………………………………………………………..…51

Table 8. Desalination with 3.5 wt% sea salt solution by using GO/EDA-2…………..52

Table 9. H2/CO2 binary gas separation results with pure GO, GO/EDA-2, OAGO/EDA-2 membranes…………………………………………………………...70

Table 10. Pervaporation desalination of GO/EDA-2 and OAGO/EDA-2…………....71

Table 11. Comparison of organic dye separation efficiency with GO based membranes in this work and literatures…………………………………………………………..121

xi CHAPTER I

INTRODUCTION

With the fast development of modern industry and vastly growing demands of natural resources, human society is facing ever challenging environmental and energy crisis, such as air pollution, shortage of clean water, global warming, etc. Shortage of drinking water is one of the most urgent challenges in this century. According to World

Health Organization (WHO), 884 million people are still suffering from the shortage of drinking-water in 2017. It is projected that half of the world’s population will be living in water-stressed areas by 2025[1]. To alleviate drinking water issue, separation of drinkable water from sea water or polluted water has practiced for decades where different technologies have been developed such as distillation[2, 3], ion exchange[4], membrane separation[5, 6], freezing desalination[7, 8], solar desalination[9, 10], etc.

Among these technologies, membrane separation is considered as a promising technology taking advantages of its design flexibility, easy scaling, large treatment capacity and good selectivity.

Based on the selected materials, membranes can be roughly divided into two categories: organic (mostly polymer[11-16]) and inorganic[17-21] (mostly metal, metal oxide, ceramic and carbon). Over the past decades, polymer membranes have been successfully commercialized for gas and liquid separation applications. The relatively low manufacturing cost, easy installation and chemical inertness of polymer membranes are the most attracting features, while the small operation window of temperature, swelling issues in organic solvent and fast degradation rate are the major

1 issues still need to be addressed. Inorganic membranes usually show excellent mechanical strength and thermal stability that qualify them to be used in sever temperature and pressure conditions. However, inorganic membranes also have their intrinsic drawbacks. For example, the major issue of ceramic membrane is the large temperature gradient developed along the cross-section that results in membrane cracking[22]; metallic membranes can be easily poisoned by hetero-element at the surface[23].

Carbon membranes are well known for their successful application as carbon molecular sieve (CMS) in gas separations[24-27]. The pore size dependent separation mechanism has been demonstrated robust for achieving long-term separation efficiency, while the flux is usually low due to the resistance of flow in the microchannels[28].

Therefore, CMS is often supported on porous substrates as ultrathin membranes to minimize the transport resistance[29]. Besides the excellent separation performance, the high manufacturing cost of CMS is one of the major hurdles for its practical use. In recent years, another form of carbon, graphene and its oxidized form graphene oxide

(GO) have attracted ever increasing attention in the membrane separation field[30-32].

In 2012, Nair et al. presented an interesting study on the gas and liquid separation with submicrometer-thick GO membrane. The study showed that GO only allowed water penetration through the membrane while all other species are blocked even gas molecules[33]. Realizing the great potential of GO in membrane separation, research activities were trigged immediately as evidenced by the dramatically increased number of journal publications and patents in past a few years.

Overviewing the GO related membrane research, GO was mainly used in two different ways. One is to mix with matrix materials and form a composite membrane,

2 the other is to fabricate pure GO membrane supported on porous substrate. In the former approach, GO is usually incorporated into polymer matrix to form the so-called mixed matrix membranes (MMMs). The addition of GO into polymer matrix could not only improve the mechanical strength, but also regulate the hydrophilicity of the composite membranes for enhanced water flux[34-36]. In most of the reported studies, only small portion (< 1 wt%) of GO was used in MMMs. That is said, the large water flux property of GO has not been fully utilized and thus the matrix property still dominates the separation performance. To enable the greatest benefit of GO in membrane separation,

GO sheets need to be connected and arranged/stacked in desired manner, i.e. GO membrane. However, a freestanding GO membrane usually cannot meet the requirement of mechanical strength in practical use. Therefore, most of the reported

GO membranes were fabricated on top of a porous support where the support offered the mechanical strength and the top GO layer provided the separation function. Based on this concept, different methods have been developed to fabricate supported GO membrane include filtration[5, 37-41], layer-by-layer assembling[42-51] and spin- coating[52].

CHAPTER II

BACKGROUND

2.1. GO synthesis and its chemical property

Like graphene, GO is a 2D material that can be synthesized from graphite by wet chemical method. Though GO does not acquire excellent thermal and electric properties as graphene, researchers found that the unique layered structure with plenty functional groups set GO a unique position in many other fields. For example, the rich surface functional groups of GO facilitate ultrafast water transportation and thus enable its promising application in membrane separation. The binding energy of functional groups is different depending on their composition and location[53]. The plenty of oxygenated functional groups also enable its excellent dispersion capability in water and other polar solvents. A well dispersed GO solution makes the fabrication of quality

GO membrane more convenient by different approaches.

GO synthesis can track back to 19th century. In 1859, Brodie synthesized GO by using graphite, chlorate of potash and fuming nitric acid. The mixture was placed at

60 °C for 4 days. After washing and drying, Brodie found that the product contained carbon, oxygen, and hydrogen with increased weight. Then, he successfully calculated the ratio of C: H: O (61.04: 1.85: 37.11) in the oxidized form of graphite.

In 1958, Hummer and Offeman reported a rapid and relatively safe method to synthesize GO from graphite by using sulfuric acid, sodium nitrate and potassium permanganate[54]. The C: O ratio of GO is between 2.1 and 2.9. Marcano and Kosynkin developed a modified Hummer method, which is more efficient in terms of

4 oxidation[55]. More oxygenated functional groups can be generated that is beneficial for subsequent chemical treatment such as surface modification and cross-linking.

Moreover, no toxic gases were generated. They also found an increased number of isolated aromatic rings in the final product, indicating the less disruption on the basal plane than Hummer method. The Hummer’s method and the modified ones are now the most popular methods to synthesize GO.

Although tremendous research has been carried out to understand the GO structure, it still remains a challenge to precisely describe its accurate structure[56, 57].

In general, GO sheet is considered as a graphene basal plane with oxygenated functional groups. Typically, epoxide groups in the basal plane, and carboxyl and hydroxyl groups located at the edge area. The mostly accepted and simplified GO structure is presented in Figure 1[58, 59].

Figure 1. Representative structure of graphene oxide.

2.2. Separation mechanism of GO-based membranes

The 2D sheet structure of GO qualifies itself as effective filler in MMMs. Not only structural reinforcement, but also the induced hydrophilicity of the composites for

enhanced water flux. Enhanced separation performance has been reported in MMMs after incorporating[60, 61], which did demonstrate the effectiveness of GO as filler in

MMMs. However, the distinct property of GO cannot be fully explored in such membrane structure.

In 2014, Mi reviewed the mechanism of GO membrane for ionic and molecular separation[62]. This review emphasized the key factor of interlayer d-spacing in determining the separation performance of GO membranes. Figure 2 summarizes the separation of different sized species by using GO membranes of different d-spacing values. In theory, GO membrane can be used for a broad range of separation applications as long as the d-spacing can be precisely controlled. For desalination or hydrofracking, the d-spacing needs to be controlled within the range of 0.3-0.7 nm; d- spacing of 0.7-2 nm is good for water, fuel, or chemical purification; and large d- spacing of ≥ 2 nm is suitable for biomedical separations.

6 Figure 2. (A) The pathway of water molecules and small ions in GO membrane. Larger molecules are blocked. (B) Ajustable nanochannal size for separation purposes. (C) Several GO membrane fabrication methods that have been reported.[62]

To maximize flux, thinner GO membrane is often preferred. However, thinner membranes often sacrifice its mechanical strength. Therefore, GO membrane is often prepared on a mechanically strong porous support. The commonly used supports are porous polymers or ceramics including but not limit to polyacrylonitrile (PAN)[48, 63-

65], polyethersulfone (PES)[66-68], polysulfone (PSf)[40, 42, 43], and alumina[6, 38,

50, 69].

2.3. Fabrication of GO based membranes

2.3.1. Filtration method

Vacuum filtration has been used for a long time to fabricate supported GO membranes due to its advantages in easy operation, convenient thickness control and adjustable filtration rate by vacuum pressure[5, 37, 38, 63, 70-73]. A wide range of porous substrates can be selected to support the GO membrane and the membrane

thickness and structure can be controlled by processing conditions. For example, Hung et al. successfully prepared flexible GO membrane on modified PAN (mPAN) substrate by vacuum filtration technique[63]. The PAN membrane was firstly modified in NaOH

(2 M) to improve its surface hydrophilicity. After modification, highly-packed layer structure and smooth surface can be obtained, as seen in Figure 3[63].

The processing method has huge influence on the GO membrane structure. In a recent study by Tsou et al.[65], three different processing methods were adopted to fabricate GO membranes, i.e. pressure-assisted (PASA), vacuum-assisted (VASA), and evaporation-assisted self-assembly (EASA), Figure 4. The main difference of these methods is the way how pressure applies to the filtration solution. PASA leads to a highly ordered membrane that shows a clear packed laminar structure; VASA results in a highly ordered structure at bottom but random on the top surface; EASA forms a loosely packed membrane structure without packing orientation. Both surface

morphology and cross-section structure demonstrate the effectiveness of PASA method in fabricating highly packed GO membranes, Figure 5.

Figure 4. Schematics of PASA, VASA, and EASA methods for GO/mPAN membrane fabrication.[65]

Figure 5. Digital images, surface and cross-section SEM images of GO/mPAN membranes fabricated by PASA (a, d, g), VASA (b, e, h) and EASA (c, f, i) method.[65]

Not only the membrane structure itself, the operation conditions of separation also affect the membrane structure and correspondingly the separation efficiency. The stacking of GO sheet in the fabricated membrane could be modulated by the pressure applied to the membrane during separation. In a recent study, Wei et al.[74] found that

9 high pressure led to a reduced interlayer distance of GO membrane and thus changed the separation performance. Specifically, at 1.0 MPa, water flux of GO membrane decreased by 75% after compaction; while at 1.5 MPa, the rejection of sodium sulfate increased from 21.32% (before compaction) to 85.84% (after compaction). Both reduced water flux and increased rejection rate implied the reduced d-spacing of GO membrane.

2.3.2. Layer-by-layer assembling method

Layer-by-layer (LbL) is another commonly used method in thin film fabrication[42, 43, 45, 47, 48, 50, 75-77]. Traditionally, LbL technique uses electrostatic attraction as driving force to assemble oppositely charged materials into layered film structures. Depending on the number of repeating cycles of assembling, the thickness of the thin film can be precisely controlled in nanometer range. Hu et al. fabricated GO membrane by electrostatic interaction between negatively charged GO and positively charged poly (allylamine hydrochloride) (PAH)[45]. The water flux of the GO membrane (2.1-5.8 LMH/atm) is much higher than that of commercial forward osmosis membrane (0.36 ± 0.11 LMH/atm). At low ionic strength solution, the

GO/PAH membrane remained its tight stacking structure and high sucrose rejection rate of 99%.

Despite the electrostatic attraction, covalent bonding has also been utilized to bridge the interlayer connection of GO sheets. Covalent bonds have larger bonding energy than electrostatic attraction and therefore robust membrane structure could be expected. Hu et al. reported a method to synthesize supported GO membrane by LbL deposition of GO nanosheets and 1,3,5-benzenetricarbonyl (TMC)[43]. The detailed synthetic procedure is shown in Figure 6, where polydopamine (PDA) coated PSf

10 membrane is immersed into TMC solution and GO solution alternatively. The PDA is used as a transition layer to provide strong bonding between PS substrate and GO membrane. This membrane showed outstanding separation performance for

Rhodamine-WT (93 - 95%).

There is no doubt that chemistry plays the major role in LbL assembly, some other factors may also influence the LbL process. Recently, Zhao et al. successfully fabricated GO/polyethyleneimine (PEI) membrane with the assistance of an external electric field. As shown in Figure 7, the PEI chain was highly disordered without the electric field, which induced the random packing of GO sheet. After introducing the electric field during assembling, a tightly packed membrane structure can be obtained,

Fig 7 (b). The effect of electric field on membrane structure is evidenced by the reduced surface roughness at increased voltage. The other benefit of using electric field is to expedite the manufacturing processing. Comparing to membrane fabricated by

11 traditional LbL method, G O membrane synthesized under electric field has a better packing structure of GO sheets[51].

Figure 7. Schematic of (a) conventional LbL self-assembly process and (b) applied electric field involved LbL self-assembly process.[51]

2.4. Swelling issue and its strategies

2.4.1. GO swelling

GO has plenty of oxygen-containing groups on the basal plane as well as at the edge area of the sheet. The d-spacing of graphene membrane is ~0.34 nm, and the d- spacing of GO membrane at dry state is 0.8 ± 0.1 nm[55, 78, 79]. The hydrophilic GO membrane is easy to be hydrated and the d-spacing will change dramatically depending on the degree of hydration. To correlate the degree of hydration and d-spacing of GO membrane, Zheng et al. performed a study in aqueous environment[80]. The d-spacing of both dried and hydrated GO membrane was accurately measured by using an integrated quartz crystal microbalance and ellipsometry. The swelling of the GO membrane can be precisely quantified. The d-spacing of wet GO membrane is reported within the range of 6-7 nm, which is larger than the ions with the hydration layer. In other words, the swelling of GO membrane in an aqueous environment will dysfunction the size exclusion mechanism for ionic separation, it will also lead to instability of the

12 GO membrane in water. Therefore, strategies to control GO swelling become critical to remain GO membrane function in the long term.

2.4.2. Swelling control by non-covalent

As mentioned above, GO membrane tends to trap water molecules between GO sheets and leads to an increased d-spacing. To restrict swelling, stronger inter- connection of GO sheets is demanded. The abundant negatively charged groups on the

GO sheet allow strong electrostatic attraction by inserting positively charged species.

For instance, Choi et al. reported an LbL assembly method to fabricate a supported GO membrane with negatively charged GO and aminated GO (AGO) with the positive charge. By depositing alternating AGO/GO layers, a GO membrane was successfully fabricated[42]. The membrane shows good stability in separating NaCl. Later, Zhao et al.[76] also used an LbL method to fabricate hybrid GO/gelatin membranes. Three different driving forces were responsible for the inter-connection between gelatin and

GO nanosheets, including electrostatic attraction, hydrogen bond, and hydrophobic interaction. As shown in Figure 8, ionized carboxyl groups of GO and amino groups in gelatin formed electrostatic attraction; carbon backbone of GO and the hydrophobic side chain of gelatin form hydrophobic interactions; polar groups in gelatin and GO form hydrogen bonds. In general, the strength of non-covalent bonds is weaker than that of covalent bonds.

Figure 8. The schematic of multiple interfacial reactions in hybrid GO membrane.[76]

Naturally, GO is not stable in water due to its hydrophilic property. However, the remarkable stability of supported GO membranes in water was reported[81, 82].

Using porous anodized aluminum oxide (AAO) as support, Yeh et al. found that Al3+ acted as a cross-linker that led to a good stability of GO membrane in water. More recently, Chen et al.[83] studied the effect of different cations on the interlayer spacing of the GO membrane. Experimental results indicated that GO membrane by KCl treatment has the smallest interlayer spacing, Figure 9(a). The ion permeation rate of

GO-KCl is much lower than that of untreated GO, Figure 9(b). To better understand the separation mechanism of the GO membranes cross-linked by different cations, molecular dynamics simulation was performed and revealed that the presence of both oxygen functional group and aromatic ring were required to keep stable cation position in GO membrane. The molecular orbital analysis demonstrated the coupling between empty orbitals of cation and lone pair of electrons of GO (oxygen atoms from the functional group and electron cloud from benzene ring). Thus, both cation-π state

interaction and cation-oxidized groups interaction helped to fix the d-spacing of GO membrane and thus achieved excellent separation performance.

2.4.3. Swelling control by covalent bonds

The swelling issue of GO membrane can be either alleviated by peeling off the surface hydrophilic groups or embed strong connections between GO sheets. Bridging

GO sheets by strong covalent bonds are one of the most effective approaches. For example, Feng et al. reported a vacuum filtration method that used 1,4-phenylene diisocyanate (PDI) as cross-linker to synthesize 3-D GO framework[6]. By adding PDI into GO suspension, carbamate esters were formed between GO flakes. The separation results revealed a high water-flux, good ion rejection rate, and good stability. Some other researchers also reported the improved stability of GO membrane after introducing covalent bonds[84, 85].

The d-spacing control determines the permeation of specific ions/molecules through the membranes. Adjustment of d-spacing through hydration is not an option since it is not able to control the d-spacing at desired value. Therefore, cross-linker needs to be inserted between GO sheets to regulate the d-spacing. For example, Hung et al.[86] reported a pressure-assisted self-assembly method to fabricate GO

membranes that cross-linked by three molecules of different lengths, ethylenediamine

(EDA), butylenediamine (BDA), and p-phenylenediamine (PPD). The structural diagram of these membranes is provided in Figure 10. These membranes were designed to separate water/alcohol mixture and all of them showed good stability. Theoretically, the water flux is supposed to increase with the increase of d-spacing, while it is not in this case. The cross-linker also changed the microstructure and hydrophilicity of the

GO membrane, which affected the water flow across the membrane.

Figure 10. Structural diagram of GO and three composite GOF membranes (GO-EDA, GO-BDA, and GO-PPD). Each GOF was produced by cross-linking GO with a diamine (EDA, BDA, or PPD).[86]

2.4.4. Physically confined GO with vertical alignment

It has been experimentally demonstrated that the d-spacing of hydrated GO membranes is larger than 1.35 nm, which is not suitable for ionic separation since the diameter of most hydrated ions is less than 1.0 nm[82, 87]. To enable ionic separation by GO membrane in aqueous environment, swelling should be well controlled. Besides the covalent bond approach, reduction of GO membrane into graphene membrane is another option. Sun et al. fabricated a GO/Titania hybrid membrane. Titania was used as a photocatalyst that reduced GO to r-GO under ultraviolet irradiation. The oxygenate functional groups were removed after the photocatalytic reaction, which effectively prevented water accumulation in the membrane and thus reduced swelling. Meanwhile, the hydrophilic membrane turned into hydrophobic that sacrificed the water flux through the membrane.

Recently, Abraham et al. reported a physical confinement approach to restrict

GO membrane swelling[88]. Specifically, GO membrane was vertically embedded into polymer matrix with a cross-section structure facing to the separation media, Figure

11(a). The GO membrane was vertically-fixed in a polymer matrix and the swelling became impossible since there was no room for volume expansion. The d-spacing of

GO membrane can be controlled by exposing the dried membrane into the humid environment and the distance was measured as from 6.4 (RH-0%) to 9.8 Å (RH-100%).

The cross-section structure of the membrane was characterized and presented in Figure

11(b-d). By using the GO membrane with d-spacing of 6.4 Å, ion permeation was not detected after 5 days of separation test.

2.5. Fundamentals of GO Membrane Separation

The separation function of GO membrane is based on its size of d-spacing and the size of molecules to be separated. If the hydration diameter of molecules or ions is

larger than d-spacing, they will be blocked; otherwise, molecules or ions will penetrate through the GO membranes. In general, the d-spacing of GO membrane will be affected by a few factors including the stacking pattern, density of functional groups, amount of intercalated water molecules, etc. As a result, different d–spacing values from 0.4 - 1.0 nm were reported[89, 90].

Gas permeance and selectivity are the two main parameters to evaluate membrane separation performance. The gas permeance (J) is a function of gas flux and partial pressure difference across the membrane, Equation (1)[91].

푔푎푠 푓푙푢푥 퐽 = (1) 푝푎푟푡푖푎푙 푝푟푒푠푠푢푟푒 푑푖푓푓푒푟푒푛푐푒 where the partial pressure difference is the pressure difference between upstream and downstream of GO membrane. The unit of permeance of i (Ji) was reported as GPU (1

GPU = 10-6 cm3(STP)/(cm2·s·cmHg) = 3.35×10-10 mol/(m2·s·Pa)).

More specifically, Karunakaran et al.[92] derived Equation (2) to calculate permeance:

푉∗22.4 푝 −푝 퐽 = 푙푛 ( 퐹 0 ) (2) 푅∗푇∗퐴∗푎∗푡 푝 −푝 퐹 푃(푡) where V is volume of permeate gas in L; R is ideal gas constant, 0.0831 bar·L·mol-1·K-

1; T is temperature of operation in K; A is effective membrane area in m2; t is operation time in second; pF is pressure of feed; p0 is initial pressure of permeate side; pP(t) is pressure of permeate side at time t. All pressure values are in unit of bar.

Then selectivity (α) of gas species (i and j) can be calculated by Equation (3).

퐽 α = 푖⁄ (3) 퐽푗

For liquid separation, permeation flux (F) and rejection rate (R) are often used to evaluate separation performance. F and R can be calculated by Equations (4) and (5), respectively.

푀 F = (4) 푡∗퐴∗푃

퐶 R = (1 − 푝) × 100% (5) 퐶푓 where M is total mass in kg; t is time in h; A is the effective area of membrane in m2; P is operation pressure in bar. Cp and Cf are concentrations of solutions in permeate and feed streams.

2.6. Gas Separation

CO2 is well-known greenhouse gas. Separate, collect and reuse CO2 in desired manner is one of the most studied research areas now[93]. Efficient CO2 separation technology is urgently demanded in industry. In this section, we reviewed the CO2 separation from three different mixtures (CO2/N2, CO2/CH4, and CO2/H2) by using GO membrane. Table 1 briefly summarized the different methods and materials that have been reported on CO2 capture and separation.

Table 1. Gas separation membranes for CO2 separation reported by references.

Membrane Gas Permeance/permeability Selectivity Reference material mixture

-7 Silicalite-1 CO2/N2 CO2: 7.0 × 10 68 Guo [94] mol/m2 ·s· Pa

-6 SAPO-34 CO2/N2 CO2: 1.2-1.5 × 10 21-32 Li [95] mol/m2 ·s· Pa

Multiwalled CO2/N2 CO2: 741.67 GPU 40.17 Ahmad [96] carbon nanotubes in cellulose acetate matrix/polymer

ZIF-8 in CO2/N2 CO2: 18 GPU 44 Dai [97] polyetherimide matrix

GO-ionic liquid CO2/N2 CO2: 37 GPU 130 Karunakaran composite [92] membrane

-7 SAPO-34 CO2/CH4 CO2: 4.0 × 10 115 Li [98] mol/m2 ·s· Pa

−10 Supported ionic CO2/CH4 CO2: 5 × 10 to 5 × 5-30 Iarikov [99] liquid membrane 10−9 mol/m2 ·s· Pa

GO in PSf matrix CO2/CH4 CO2: 86.6 GPU 25 Zahri [100]

Borate cross- CO2/CH4 CO2: 650 GPU 75 Wang [85] linked GO membrane

-6 Organic CO2/H2 H2: 1.7 × 10 17.4 Ying [41] framework (mol/m2 ·s· Pa) membrane

Multiwalled CO2/H2 CO2: 836 Barrer 43 Zhao [101] carbon nanotubes in PVA matrix

-7 Ultrathin GO CO2/H2 H2: 1.0 × 10 3400 Li [102] membrane mol/m2 ·s· Pa

2.6.1. CO2/N2 Separation

The gas separation of graphene and GO membranes started with simulation studies[103]. After demonstrating the effectiveness of such membranes by simulation, a great amount of experimental work was carried out. For example, Dong et al.[91] fabricated a GO-polyethyleneimine (HPEI)/trimesoyl chloride (TMC) membrane on

PSf substrate and used it for CO2/N2 (10:90 v:v) separation, as illustrated in Figure

12(a). It was revealed that GO coated membrane showed CO2 flux of 9.7 GPU and selectivity of 80, which is much higher than pure polymer membrane. Figure 12 shows the relation between CO2 permeance and GO content. Highest CO2 permeance and

CO2/N2 selectivity were both achieved at GO content of 0.33 wt.%.

Figure 12. (a) Structures of HPEI and TMC, and the reaction between them. (b) permeances of CO2 and N2 in term of GPU. (c) the relation between the selectivity of CO2/N2 and the GO content.[91]

Better CO2/N2 selectivity of GO membrane was reported by Karunakaran et al[92]. They fabricated ultrathin GO/ionic liquid (IL) hybrid composite membrane on porous PAN substrate and achieved high CO2 flux and CO2/N2 selectivity by coating

GO/IL. Two commercial ionic liquids were used in this study: 1-ethyl-3- methylimidazolium acetate ([EMIM][Ac]) and 1-ethyl-3-methylimidazolium tetra fluoroborate ([EMIM][BF4]). The IL/PAN membranes without GO showed poor

CO2/N2 selectivity, which was 20 for [EMIM][Ac] and 22 for ([EMIM][BF4]). The addition of GO created extra separation channels and resulted in a much better separation performance with the CO2 flux of 37 GPU and high selectivity of 130.

2.6.2. CO2/CH4 Separation

In natural gas and syngas purification, efficient CO2/CH4 separation is highly demanded. Wang et al.[85] fabricated an ultrathin GO membrane with outstanding

CO2/CH4 separation performance. In this work, borate was used as both cross-linker and transport carrier of CO2. With borate in the membrane, nanochannels could be formed between GO layers that resulted in a size sieving capability of GO membrane.

The small pore size allowed diffusion of relatively smaller CO2 molecules, but restricted the permeation of larger CH4 molecules. The ultrathin GO membrane exhibited outstanding CO2 flux of 650 GPU and good selectivity of 75. Without crosslinking, GO membrane showed much lower CO2 flux of 105 GPU and poor selectivity of 16. The overall separation performance of such supported GO membranes is much better than MMMs such as GO/PSf (flux: 86.8 GPU; selectivity: 25)[100].

2.6.3. CO2/H2 Separation

GO membrane can be used for H2 separation from CO2 in petrochemical streams[41]. By using a spray-evaporation technology, Guan et al.[104] fabricated GO

-8 2 membrane that reached H2 permeance of 2.7×10 mol/(Pa·m ·s) (80.6 GPU) and

H2/CO2 selectivity of 20.9. The H2 selectivity is better than GO-assisted ultrathin covalent framework membrane (17.4)[41]. Shen et al.[89] also reported a precisely structured subnanometer 2-D GO membrane by controlling the external forces. As described in Figure 13(a), the tight packing of GO membrane formed by external force driven assembly (EFDA) was mainly controlled by three forces: intrinsic force

(repulsive electro-static interactions between GO sheets), “outer” external forces

(compressive force, centrifugal force and shear force), and “inner” external force

(GO/polymer interaction). Figure 13(b) shows the H2/CO2 selectivity of the and

comparison with other reported membranes. Obviously, the H2/CO2 selectivity of GO membrane is much higher (~30) than other membranes. The separation performance exceeded permeability-selectivity upper-bound of polymeric membranes that was presented by Robeson in 2008[105].

High H2/CO2 selectivity of 240 was reported by Chi et al.[52] with ultrathin GO membranes (2-3 layers) on top of the alumina substrate. The H2 permeance reached

3.4×10-7 mol/(Pa·m 2·s) (1014.93 GPU). For comparison, the authors also fabricated

GO membrane by vacuum filtration. Although H2 permeance of this membrane is 35% higher than the prior one, H2/CO2 selectivity dramatically decreased to 51. The highest

H2/CO2 selectivity of 3400 was reported by Li et al. with GO membrane thickness down to 1.8 nm[102]. The H2/CO2 and H2/N2 separation performance were presented in

Figure 14. The separation factors of H2/CO2 and H2/N2 reached to 3400 and 900 at 20 °C, respectively. For both gas separation experiments, the separation factors decreased dramatically when the temperature increased to 100 °C, the red line in Figure 14. This can be explained by Arrhenius law, equation (6), where Ed is diffusion activation energy and ∆퐻푎푑푠 is the heat of adsorption, and both units are kJ/mol. The Ed-∆퐻푎푑푠 value of

H2 is 6.9 kJ/mol while the value of CO2 is 60.2 kJ/mol, which means CO2 permeance is much more sensitive to temperature than H2. However, this paper did not mention

how operating temperature change affect the micro-structure of GO membranes, which also is a likely reason that affected gas separation performance.

−(퐸푑−∆퐻푎푑푠) 푃푒푟푚푒푎푐푛푒 ∝ 푒 푅푇 (6)

Figure 14. (A) Separation result of 50:50 H2/CO2 gas mixture and (B) separation result of 50:50 H2/N2 gas mixture.[102]

2.7. Liquid Separation

The outstanding liquid separation performance of graphene and GO membranes is firstly predicted by simulation studies. In 2012, David et al. predicted that nanoporous graphene has a great advantage in water purification due to the high water flux in the laminar structure and they demonstrated a linear relationship between water flux and pore area[106]. It has been understood that the good hydrophilicity of pores is the reason for the increased water permeation since oxygen functional groups are most distributed at the edge area of the pores. It is well accepted that water permeation is dependent on the membrane thickness of traditional membranes. However, this is not true in GO membranes as evidenced by an experimental study by Hu et al.[43]. In this study, the LbL method was used to fabricate GO membranes of different thicknesses by controlling the assembled layers. The testing results indicated that water flux was not related to the thickness of the GO membrane. In other words, the free motion of

water molecules between the GO layers cut-off the diffusion resistance along the path and thus the membrane offers a thickness-independent flow behavior.

Until now, GO based membranes have been tested in a wide range of liquid separation applications including organic dye removal[107], desalination[108], sterilization, etc. Nam et al.[109] used branched polyethylene-imine as cross-linker to fabricate a laminated GO membrane. The membrane showed a high rejection rate (>

90%) for different dye molecules such as methylene blue, rose bengal, and brilliant blue.

With a defect-free GO nanohybrid membrane prepared by LbL technology, Wang et al.[44] reported a much higher rejection rate of 99.5% for congo red and 99.3% for methyl blue. In a similar work, polycation/GO nanofiltration membrane was also fabricated by the LbL method, Figure 15(a), and a high rejection rate of 99.2% was reported for methyl blue[49]. Dye rejection of poly(diallyl dimethylammonium chloride) (PDDA)/GO and PDDA/poly (acrylic acid) (PAA) membranes was comparatively investigated in Figure 15(b). The retention of two membranes was about the same, while PDDA/GO showed higher flux and better stability in a wider range of dye concentration from 100 to 1500 mg/L. Different from Han’s work[110], this study found that the thickness of the GO membrane had a huge influence on both the water flux and retention rate. By increasing layer number from 1 to 10, retention rate increased from 92.5% to 99.6% and water flux decreased from 93.1 kg/(m2·h·bar) to

2.89 kg/(m2·h·bar).

Ion transport across GO membranes has been studied widely due to their potential for desalination applications. To achieve good ionic separation, the d-spacing of GO membrane should be small enough to exclude the diffusion of ionic species. The typical size of ions is smaller than 1.0 nm. In 2015, Coleman et al.[111] reported that ion transportation in GO membrane could be dominated by two different mechanisms: pore- and slit-dominated transport. In thin membranes, ion transportation is dominated by relatively larger pores with a diameter of > 1.75 nm, while slits (< 1.42 nm) dominate the ion transportation in thicker membranes. Jia et al.[112] synthesized GO membranes with adjustable d-spacing by adding different cross-linkers, where the selectivity of

K+/Mg2+ is only 6.1. A year later, the same group[84] reported GO membranes with diamines as cross-linkers. The d-spacing of GO membrane is increased with increasing the chain length of cross-linkers. The new designed GO membranes exhibited higher water fluxes and K+/Mg2+ selectivity of 7. The separation performance of GO membranes with different cross-linker is summarized in Figure 16. Ion fluxes of all the membrane show similar order, which is K+ > Na+ > Ni2+ > Mg2+.

Figure 16. (a) Ion permeation fluxes of diamines cross-linked GO membranes, (b) separation factor of K+/Mg2+ of diamines cross-linked GO membranes.[84]

To remain the integrity of GO membrane in liquid media, cross-linker is often added between GO sheets. Feng et al.[6] reported a GO framework on alumina tubes with 1,4-phenylene diisocyanate (PDI) as cross-linker. The GO sheets were strongly connected with carbamate and amide functionalities, Figure 17(a). The ion rejection rate reached 99.9% with a large water flux of ~7 kg/(m2·h-1), Figure 17. Increasing operation temperature to 90 °C resulted in an increase of water flux to 11.4 kg/(m2·h-1) without sacrificing the ion rejection rate. With the temperature increasing, the water vapor pressure of the feed side increases, while the vapor pressure on the permeate site remains constant, which offers a larger driving force across the membrane. Also, high temperature leads to a faster thermal motion of water molecules, which results in the accelerated diffusion rate. The stable ion rejection rate indicated that the microstructure was not changed by the increased temperature. The membrane remained outstanding separation performance at 75 °C within the testing period of 120 hours.

Figure 17. (a) Schematic of synthesizing graphene oxide framework. (b) water flux and ion rejection results of 3.5 wt.% seawater desalination.[6]

Not only has d-spacing, the surface charge of the GO membrane also affected the ion separation performance[113]. The surface modification is also on the hot topic of GO membrane for liquid separation[114, 115]. With the exposed positive charge in polyethyleneimine/GO membranes, Nan et al. reported a high Mg2+ rejection rate of

93.9%[48]. Comparative investigation on GO membrane and reduced GO (rGO) membrane also revealed the important role of surface charge in ionic separation[116].

GO and rGO membranes show different characteristics: (1) d-spacing of GO membrane is relatively larger (0.8 nm) than rGO membrane (0.7 nm); (2) GO membrane has more surface oxygenate functional groups than rGO membrane that exposes more negative charge in liquid media. In metal ion (Hg2+, Cu2+, and Pb2+) separation tests, GO membrane exhibited better (~50% higher) rejection rate compared to rGO membrane.

According to the Donnan effect and the charge neutrality requirement, both anions and cations were repulsed by the negative charge on the membrane surface. In this case, although d-spacing of GO membrane is relatively larger, the intrinsic surface charge of

GO membrane dominated the metal ion separation process. In fact, for other separation tests, both surface charge and d-spacing could be the dominant factors. Therefore, adjust surface charge and control d-spacing are the major considerations when GO membrane is designed for ionic separation.

Microorganism fouling of GO membranes is also a big concern in practical separation processes. Membrane fouling not only reduces water flux, but also affects the selectivity. Different strategies for developing antifouling GO membranes have been explored in recent years. For example, by introducing GO layer on top of polyamide active layer through covalent bonds, Perreault et al.[117] have reported that

64.5% of E. coli cells were not active after contacting GO membrane for one hour. This study revealed that GO itself has good antimicrobial properties. Similarly, by grafting

GO on polymer thin-film membrane, Hegab et al.[118] found that the normalized flux of GO modified membrane decreased by only 3% compared to a 14% decrease of the membrane without GO. Taking advantage of the antibacterial property of silver nanoparticles, Sun et al.[5] synthesized GO/silver composite membrane on top of the cellulose acetate substrate. The excellent antibacterial property, high water flux, and strong mechanical strength can be simultaneously achieved in the membrane of composites. As seen in Figure 18, rare E. coli bacterial colonies were active after contacting GO/Ag membrane for 4 hours.

Figure 18. The comparison of antibacterial performance between GO/Ag membrane (grey) and pure GO membrane (blank).[5]

CHAPTER III

PERMSELECTIVE H2/O2 SEPARATION AND DESALINATION OF HYBRID

GO/rGO MEMBRANES WITH CONTROLLED PRE-CROSSLINKING

3.1. Outline

Covalent bonding is widely adopted in graphene oxide (GO) membrane to improve structural integrity and restrict swelling, while it comes with a price of enlarged d-spacing and sacrifices membrane selectivity. This work offers a facile strategy to break the tradeoff between membrane stability and selectivity. Specifically, graphene oxide (GO)/reduced GO (rGO) hybrid membranes were fabricated by a controlled pre-crosslinking method. With this method, restricted swelling by crosslinking and reduced d-spacing by GO reduction can be achieved simultaneously by controlling reaction time.

3.2. Introduction

Membrane separation has been considered as an effective process in several industries, such as H2 purification[119], water purification[120], and seawater desalination[121]. Graphene (or graphene oxide) based membranes have attracted great interest in recent years due to their excellent chemical stability, thermal stability and more importantly the unique two-dimensional (2D) laminar structure that allows structure design of such membranes[33, 103, 122, 123]. Especially graphene oxide

(GO), with a similar 2-D structure as graphene, the rich surface functional groups at both edge area and basal plane enable a vast opportunity to design the membrane

functionality[72, 84, 92, 109]. After the discovery of exclusive and ultrafast water transportation behavior in GO membrane in 2012[33], tremendous research work have been done that enable excellent properties of GO membranes in various gas and liquid separations[52, 124-126]. Till now, GO has been used in two different ways for membrane development: (1) GO as filler in polymer matrix, often called polymer mixed matrix membranes (MMMs)[36, 60, 127]; (2) GO membrane with patterned stacking layers of GO sheets. In first approach, although improved separation performance has been successfully demonstrated in earlier work, the unique 2-D laminar structure of GO has yet to be utilized. In second approach, the mechanical strength of GO membrane becomes a challenge since the membrane thickness are usually at nanometer or submicron meter scale[66, 102, 110]. Therefore, GO membranes are usually supported on porous substrate to integrate the unique separation function of GO membrane and mechanical strength of the supporting substrate[41, 83, 104].

It has been demonstrated that d-spacing, the interlayer distance of GO nanosheets, is one of the most important parameters that influences the separation performance of supported GO membranes[62]. This is not surprising since the “slit” formed by stacked GO sheets is the major path for molecule transportation. The size of

“slit” is determined by the d-spacing of GO membrane. With a size exclusive separation mechanism, the that d-spacing control becomes critically important to design separation functions. The d-spacing of dried GO membrane and graphene membrane is reported as ~0.80 and 0.34 nm, respectively. The larger d-spacing of GO membrane is mainly attributed to the steric hindrance exerted by oxygen functional groups on GO sheets

[80]. Smaller d-spacing seems to be favorable to achieve better separation performance, which can be achieved by eliminating the oxygen functional groups of GO and form a

more compact structure of reduced GO (rGO)[56]. However, once GO is reduced, it loses its hydrophilicity and thus its unique water separation property. Moreover, the dispersion of rGO in water becomes a challenge that usually results in severe aggregation. Once aggregated, a good patterning of GO layer stacking becomes very difficult that is undesired in membrane fabrication.

Another crucial issue needs to be addressed for GO membranes to be used in liquid separation applications is that it tends to swell in aqueous environment due to the presence of rich hydrophilic oxygen functional groups. Swelling causes two main issues, expanded d-spacing and weakened mechanical strength. The d-spacing of wet GO membrane can reach up to 6-7 nm[80], which is much larger than the size of hydrate radius of salt ions. Therefore, swelling control is critical to maintain separation function as well as long term stability[128]. Bridging GO layers with covalent bonds is a widely practiced approach[6, 43, 86], while the d-spacing could be larger due to the introduced chemical linkers in between. The smallest crosslinkers, cations, have been reported recently that successfully confine the swelling of GO membrane and achieves excellent ionic separation properties[83].

In this work, a controlled pre-crosslinking method is developed to address the swelling issue and d-spacing control simultaneously. Crosslinking and GO reduction are coupled in the pre-crosslinking reaction. Hybrid GO/rGO membranes with well patterned layer stacking structure is fabricated by vacuum filtration method. The crosslinking reaction and microstructure evolution of the GO membranes are systematically characterized by X-ray photoelectron spectroscopy, Raman spectroscopy, scanning electron microscope, and X-ray diffraction. The properties of the prepared membranes are tested in both gas and liquid separation systems. The

H2/CO2 separation is measured with a flow-work gas separation system, and desalination test was carried out in a pervaporation system.

3.3. Experimental Procedures

3.3.1. Materials

Graphite power (SP-1) was purchased from Bay Carbon Inc, USA. Potassium persulfate (K2S2O8, ≥ 99.0%) was purchased from Fisher Scientific. Hydrochloric acid

(HCl) was purchased from EMD Millipore Corporation. Potassium permanganate

(KMnO4, ≥ 99.0%), phosphorus pentoxide (P2O5, ≥ 98.0%), sulfuric acid (H2SO4, 95.0-

98.0%), hydrogen peroxide (H2O2, 30 wt% in H2O), ethylene diamine (EDA ≥ 99.0%), dopamine hydrochloride, and Trizma base (≥ 99.9%) were all purchased from Sigma

Aldrich. The dialysis membrane (Spectra/Por 4, Molecular weight cut off: 12-14 kD) was purchased from Spectrum Laboratories, Inc. Macroporous α-alumina disks

(diameter: 2.54 cm; thickness: 1 mm; porosity: 25%) was purchased from Coorstek.

3.3.2. GO synthesis

GO was synthesized by modified Hummers method[129]. A pre-oxidation step was adopted prior the Hummer method. Specifically, 4.0 g K2S2O8 and 4.0 g P2O5 were added into a beaker containing 12 mL H2SO4. The mixture was heated to 80 °C in an oil bath. Then, 3.0 g graphite powder was added into the mixture and the temperature was remained at 80 °C for 6 hours. The mixture was then cooled down to room temperature and diluted with 500 mL water, followed by washing and filtrating until the rising water became neutral. The pre-oxidized product was dried at 40 °C overnight.

After drying, it was mixed with 120 mL concentrated H2SO4 in iced water bath. 15 g

KMnO4 was gradually added into the mixture and the temperature was controlled at below 20 °C. After that, the mixture was stirred at 35 °C for 2 hours and then diluted

with 250 mL DI water. After 15 min, the solution was continuously diluted with 700 mL DI water followed by the addition of 20 mL H2O2 solution. The mixture turned to bright yellow indicating the formation of GO. The GO was filtered and washed by a

1:10 HCl solution (1 L) to remove metal ions. A small amount of DI water was then added to form a viscous brown dispersion, which was then sealed in a dialysis membrane. The dialysis membrane was then soaked in DI water for 7 days to remove remaining Cl- ions.

3.3.3. Membrane fabrication

GO membrane is fabricated on porous alumina substrates by vacuum filtration method. To enhance the interfacial interaction between support and GO membrane, the alumina support was firstly coated by a layer of polydopamine (PDA). Specifically, alumina support was polished on one side with a sand paper (grit# 1000) and then soaked into 30 mL aqueous solution that contained 0.06 g dopamine and 0.036 g tris base. The PDA coating process was carried out in a shaker at room temperature for 24 hours. After that, the coated porous substrate was dried at 80 °C for 1 hour, Figure 19(b).

GO solution for membrane fabrication was diluted from the concentrated suspension, i.e. 0.5 mL GO (12.7g/L) suspension was diluted by 15.4 mL DI water to form a 0.4 g/L GO solution. To improve the GO membrane stability during desalination test, 0.19 g EDA was added into the GO and heated to 85 °C for different time periods (1, 2, 6 and 24 h) for pre-crosslinking, Figure 19(d). After that, the GO/EDA solution was then placed at room temperature over a period of 24 hours to allow further reaction. For membrane fabrication, PDA coated alumina support was sealed in a Buchner funnel by silicone gel, then 1.5 g diluted GO/EDA solution containing 0.3 g GO/EDA solution and 1.2 g DI water was vacuum filtered under 0.1 MPa, Figure 19(c). At last, the

supported GO membrane is further heated at 80 °C for 2 hours in oven for complete cross-linking reaction, Figure 19(e). The membranes are named as GO/EDA-X based on the pre-crosslinking time (X=1, 2, 6, 24). One control sample GO/EDA-0 was fabricated without pre-crosslinking. The amount and concentration of GO/EDA solution that using for control sample kept the same as others. After mixing GO with

EDA, the mixture was diluted and filtered on substrate immediately without any pre- crosslinking. At last, it was also heated to 80 °C for 2 hours in oven.

Figure 19. Schematic diagram of membrane fabrication process. (a) pure alumina substrate, (b) PDA coated alumina substrate, (c) vacuum filtration method and partial cross-linked GO/EDA dispersion, (d) structure of partial cross-linked GO/EDA, (e) GO/EDA membrane on PDA coated alumina substrate and (f) mixed stacking of GO/rGO sheets.

3.3.4. Characterization

The d-spacing of the supported membranes was characterized by X-ray diffraction (XRD, Bruker AXS D8 Discover diffractometer with General Area Detector

Diffraction System, 40 kV, 35mA) with scan rate of 1.0 degree/min within the range of

3-25 degree. The elemental composition of all samples was analyzed by X-ray

photoelectron spectroscopy (XPS, PHI VersaProbe II Scanning XPS Microprobe with

Al Kα line excitation source). The structure change of samples was measured by Raman spectroscopy (Horiba LabRam HR Micro Raman Spectrometer with a CCD camera detector). The surface morphology of alumina substrate before and after PDA coating was characterized by scanning electron microscope (SEM, JEOL-7404).

All membranes were stored in a desiccator at room temperature before gas permeation measurements. H2/CO2 binary gas permeation measurements were performed under room temperature and atmospheric pressure. The membranes were mounted in a stainless steel cell with the membrane surface facing the feed side, Figure

20. H2/CO2 were controlled by MFC at equal flowrate and mixed before feeding to the membrane separator. Argon at a flow rate of 20 cc/min was supplied to the permeate side as the sweeping gas. The composition of the permeate stream was analyzed by using an online gas chromatography (Shimadzu, GC-2014) equipped with a molecular sieve 13X column for the thermal conductivity detector (TCD).

The membrane permeance for gas component 푖 is defined as Equation (7):

푄푖 푃푚,푖 = (7) 퐴푚×∆푃푖

Where 푄푖 (mol/s) is the amount of the permeated gas through the membrane per

2 second; 퐴푚 (m ) is the active membrane area; ∆푃푖 (Pa) is the trans-membrane partial pressure difference of component 푖 between feed and permeate sides. The H2/CO2 separation factor for the binary mixture is defined as Equation (8):

(푦퐻2/푦퐶푂2)푝푒푟푚푒푎푡푒 훼퐻2/퐶푂2 = (8) (푦퐻2/푦퐶푂2)푓푒푒푑

where 푦퐻2 and 푦퐶푂2 are mole fractions of hydrogen and carbon dioxide, respectively.

Figure 20. Schematic diagram showing the apparatus for H2/CO2 separation.

The pervaporated desalination was performed via a lab-designed test apparatus.

The sea salt solution was prepared by dissolving 3.0 g of sea salt (Sigma-Aldrich) in

1.0 L of deionized water. The dish-supported GO and GO/EDA membranes were attached and sealed by an epoxy adhesive to a bell-shaped glass tube, and with the membrane side facing out of the bell-shaped tube. The membrane surface was then immersed into the sea salt solution with the other side of the tube connected to a vacuum pump, which created a trans-membrane pressure differential of approximately 1 atm across the membrane. The permeate water vapor was condensed and collected by the cold trap filled with liquid nitrogen.

The permeation flux 퐽 can be calculated by Equation (9):

푉 퐽 = (9) 퐴×푡 where 푉 (L) is the volume of permeate; 퐴 (m2) is the membrane area; 푡 (h) is the pervaporation time. The ion rejection rate for component 푖 is defined as Equation (10):

퐶 휂 = 1 − 푖 (10) 퐶0,푖 where 퐶0,푖 and 퐶푖 are the concentrations of solute 푖 in the feed and permeate side, respectively. The ion concentrations were measured by Inductively Coupled Plasma

spectrometers (Spectro, CirOs ICP Spectrometers) by Soil, Water and Forage

Analytical Laboratory (Oklahoma State University).

3.4. Results and Discussion

The degree of pre-crosslinking affects the dispersion of the GO sheets in the solution. Partial crosslinking helps to create an interconnected network of GO sheets, while over crosslinking will lead to undesired GO aggregation. Moreover, part of the crosslinking reactions would transform GO to reduced GO (rGO)[86]. Therefore, optimized crosslinking is critical to tune the final membrane structure and separation properties. Here, the crosslinking is performed with different durations from 1 to 24 hours at 85 oC. The corresponding digital images of GO/EDA solutions are shown in

Figure 21(a-d). After 1 h pre-crosslinking, the color of GO/EDA solution switched from brown to black, indicating the reduction reaction from GO to rGO (or partial reduction).

With reaction proceeds to the second hour, the good dispersion quality of GO/rGO can be well remained. After six-hour reaction, the aggregation of GO/rGO sheets becomes obvious, as evidenced by the adhered large pieces on the vial wall, Figure 21(c).

Extending the reaction time to 24 hours, severe aggregation was observed, Figure 21(d).

After the reaction, the solution was left at room temperature for 24 hours. The GO/rGO separates from the solution, indicating an over crosslinking that trigs severe aggregation of GO sheets and forms larger particles, Figure 21(e). All these results suggest that the degree of crosslinking can be controlled by extending the reaction time. Meanwhile, the reduction of GO to rGO seems occurred during the crosslinking reaction, which will be confirmed in later sections. Since aggregation is undesired for the quality control of

GO membranes, longer reaction time of 6 and 24 hours is not considered in following studies.

Figure 21. Digital images of GOEDA solution with pre-crosslinking for (a) 1, (b) 2, (c) 6, (d) 24 h and (e) solutions after 24 h settlement.

To better understand the crosslinking reaction between GO flakes and EDA, the elemental composition of pure GO, GO/EDA-0, GO/EDA-1, and GO/EDA-2 was analyzed by XPS, Figure 22. The slight amount of N element found in GO membrane could be contamination from synthetic chemicals. In general, the atomic percentage of

N element in the GO/EDA membranes increases greatly as evidenced by the intensified

N1s peak at 399.5 eV. Extending the crosslinking reaction time reduces the N content in the membrane, which is contradict to the general understanding of the crosslinking reaction. As summarized in Table 2, N element gradually decreases from 7.2%

(GO/EDA-0) to 6.9 and 6.0% in GO/EDA-1 and GO/EDA-2, respectively. The crosslinking reaction takes two steps: one is attachment of EDA to GO sheets with reaction at one terminal, which is a very fast process; following that, the other terminal will react with other available carboxylic groups from neighboring GO sheets to complete the crosslinking reaction (this step is relatively slow but more thermodynamically stable). A reaction scheme is proposed in Figure 23. At the beginning, EDA is in excess amount that maximizes the quantity of attached EDA molecules on GO sheets, Figure 23(b), and this is exactly the case of GO/EDA-0.

Driven by thermodynamic equilibrium, the other terminal -NH2 group in EDA seek to replace the attached EDA molecules and form a more stabilized structure, Figure 23(c).

The crosslinking reaction depletes the dangling EDA molecules and thus less EDA molecules in the final membrane. With longer reaction time, the crosslinking reaction proceeds further, Figure 23(d). Therefore, a gradual decrease of N element is observed in GO/EDA-1 and GO/EDA-2.

Figure 22. XPS wide scan of (a) pure GO, (b) GOEDA-0, (c) GOEDA-1and (d) GOEDA-2 membranes.

Figure 23. Schematic of reaction between GO and EDA. (a) pure GO; (b) GO/EDA-0; (c) GO/EDA-1; (d) GO/EDA-2.

Table 2. XPS elemental analyses of pure GO and GO with pre-crosslinking. Sample C % N % O % O/C

Pure GO 69.8 1.2 28.4 0.407

GOEDA-0 65.5 7.2 24.6 0.376

GOEDA-1 72.0 6.9 18.9 0.263

GOEDA-2 66.2 6.0 21.8 0.329

The specific reaction during the process can be analyzed by the deconvolution of C1s peak of XPS results, Figure 24. Pure GO membrane shows five signature peaks as shown in Figure 24(a), C=C (284.4 eV), C-O (285.7 eV), C-O-C (286.6 eV), C=O

(288.0 eV), and O-C=O (289.0 eV), which is consistent with literature reports[86, 126].

Simple mixing of GO and EDA (GO/EDA-0) leads to a few major changes: (1) vanish of O-C=O peak, (2) reduction of C-O-C peak intensity and (3) formation of new C-

O/C-N peak at 285.7 eV, Figure 24(b). The disappear of O-C=O peak and formation of new peak clearly suggest a fast amidation reaction between GO sheets and EDA. The decrease of C-O-C peak intensity suggests a reduction reaction that depletes the epoxide group from GO basal plane[86]. Extending reaction time further reduces the C-O-C peak intensity, Figure 24(b-d), indicating the gradual reduction of GO sheets with longer reaction time. This phenomenon is supported by XRD results which will be discussed later. It is found that the peak intensity of C-O/C-N does not change much.

Form the proposed reaction mechanism in Figure 23, the total number of C-N bonds does not change since it only depends on the number of carboxylic groups on GO surface. All these results confirm the inter-connection of GO sheets and GO reduction during the pre-crosslinking process.

Figure 24. XPS elemental analyses of (a) pure GO, (b) GOEDA-0, (c) GOEDA-1, and (d) GOEDA-2 membranes in C 1s.

The structural changes, especially the defects in pure GO, GO/EDA-0,

GO/EDA-1, and GO/EDA-2, were characterized by Raman spectroscopy in Figure 25.

Two signature peaks at 1350 1594 cm-1 were observed corresponding the D band and

G band, respectively[130, 131]. The relative difference of I(D)/I(G) ratio among the membranes can be used as an index to compare the quantity of defects in GO sheets.

Typically, a larger I(D)/I(G) ratio implies a higher defects level. The I(D)/I(G) ratio of the prepared membranes is listed in Figure 25. The lowest ratio of 0.91 is observed in pure GO membrane and it gradually increases to 1.08 (GO/EDA-0), 1.11(GO/EDA-1) and 1.15 (GO/EDA-2) with increasing crosslinking reaction time. The larger I(D)/I(G) ratio in crosslinked membranes indicates a higher level of structural defects in carbon lattice[132]. Similar phenomenon was also reported by other researchers[131]. During

the reduction process, some carbon atoms will be stripped from the lattice network as by-product like CO2. Such removal of carbon atoms leads to vacancies and topological defects on graphene nanosheets[133, 134]. After reduction reaction from GO to rGO, less functional groups, more vacancies and topological defects would present on GO surface that switches the hydrophilic surface to hydrophobic. Meanwhile, the elimination of functional groups facilitates a more compact packing of rGO sheets.

Such structure and property changes are expected to alter the permanent flow path within the slit channels of the GO/rGO hybrid membrane.

Figure 25. Raman spectra of (a) pure GO, (b) GO/EDA-0, (c) GO/EDA-1 and (d) GO/EDA-2.

Digital images of the prepared membrane are provided in Figure 26. After vacuum filtration, all these membranes were washed by DI water and heated at 80 °C for 2 hours. The surface of all the membranes is very smooth, and no defects or crackers

are observed. The color of membranes gradually changes from brown (GO, Figure 26a), dark brown (GO/EDA-0, Figure 26b) and to black (GO/EDA-1(2), Figure 26c&d). The color change confirms the formation of darker rGO in the crosslinked membranes. All the crosslinked membranes show excellent stability in liquid media while severe swelling occurs in pure GO membrane that damages its structural integrity. The structural stability in liquid media is critical since it provides the base of such membranes to be used in liquid separation applications. The PDA coating on top of porous alumina is essential in this work, since it not only provides a smooth surface for better patterning of GO membrane and offers sufficient reactive sites that enables a strong interface between support and GO membrane.

Figure 26. Digital images of (a) pure GO, (b) GO/EDA-0, (c) GO/EDA-1 and (d) GO/EDA-2.

Figure 27 shows the XRD profile of all the four membranes. Pure GO membrane shows a single peak at 2θ=10.7o, Figure 27(a). The d-spacing of GO membranes is then calculated as 8.2 Å by using Bragg’s equation, which is in good agreement with other published results[86]. The d-spacing of GO/EDA-0 increases to

11.0 Å due to the existence of inserted EDA molecules in between the GO layers, Figure

27(b). With pre-crosslinking, both GO/EDA-1 and GO-EDA-2 show two peaks indicating the formation of two different stacking pattern structures inside the

membrane. The two peaks at 2θ=6.8 and 11.7o in GO/EDA-1 correspond to two stacking structures with d-spacing values of 12.9 and 7.6 Å respectively. The larger d- spacing is from the cross-linked GO layers that is expanded due to the presence of reacted EDA molecules. The smaller d-spacing (small than pure GO) further confirms the reduction reaction of GO that removes the dangling surface groups and form a more compact layer structure. With reaction proceeds to 2 hours, the membrane structure becomes more compact in GO/EDA-2 as evidenced by the further reduced GO-GO d- spacing to 11.8 Å (peak 5). Meanwhile, further reduction reaction is not obvious since only slight decrease of d-spacing from 7.6 to 7.5 Å (peak 6) is observed. The reduced

4.3 Å in d-spacing is reasonable because it has been proved that the existence of epoxy and hydroxyl groups brings additional 0.44 nm of height to the carbon grid[134].

Figure 27. XRD profiles of (a) GO (b) GO/EDA-0, (c) GO/EDA-1 and (d) GO/EDA-2. Table summarizes the d-spacing value of each diffraction peak.

H2/CO2 binary gas separation was performed on pure GO, GO/EDA-0,

GO/EDA-1, and GO/EDA-2 membranes for quality evaluation. A control experiment was also carried out on PDA coated alumina support. All the membranes show apparent

H2/CO2 gas separation property indicating the satisfactory membrane quality without

pinholes or cracks. As shown in Table 3, GO/EDA-1 and GO/EDA-2 offers higher

H2/CO2 separation factors and lower H2 permeance values than pure GO membrane and

GO/EDA-0. The decline of H2 permeance under the same operating conditions is probably due to the presence of the small d-spacing structures in GO/EDA-1 and

GO/EDA-2 as evidenced by the XRD results (peak 4 and 6). When testing the PDA coated alumina support as a control, severe Ar back permeation was observed. Previous study by Lashkari et al.[135] revealed that back permeation of Argon would enhance the permeability of H2, resulting in false separation data. Therefore, the separation results for PDA coated support is not reported here. The permeation of CO2 molecules across the membrane is more restricted since its molecular size is larger size than H2.

As a result, the H2/CO2 separation factors of GO/EDA-1 and GO/EDA-2 become the largest among the tested membranes. Comparing the separation factor and d-spacing values, it is found that smaller d-spacing gives larger separation factor. Such correlation indicates a size exclusion mechanism for the separation. Moreover, longer reaction time seems beneficial for enhancing both H2 permeation and separation factor. It is worth mentioning that the simple mixing of GO and EDA (without pre-reaction) resulted in a worse membrane performance than pure GO membrane, which can be explained by the expanded layer spacing in the membrane. The gas separation property of all prepared membranes was listed in Table 3. The small deviation reveals the good batch-to-batch reproducibility of such membranes.

Table 3. H2/CO2 binary gas separation results with pure GO membrane, GO/EDA-0, GO/EDA-1, and GO/EDA-2, respectively.

-8 -2 -1 -1 Samples 훂퐇ퟐ/퐂퐎ퟐ 퐏퐦,퐇ퟐ (×10 mol.m s Pa ) Pure GO 17.32±1.25 4.27±0.23 GO/EDA-0 11.63±5.50 11.37±4.41 GO/EDA-1 19.95±2.07 2.23±0.12 GO/EDA-2 22.93±1.57 2.46±0.01

Based on H2/CO2 binary gas separation results, GO/EDA-1, GO/EDA-2 showed better membrane performance because of the exhibited permselectivity for H2 over CO2.

The H2 permeation is facilitated by a size exclusion mechanism through the slit of patterned GO sheets. Since the radii of hydrated ions (Table 4) is larger than CO2 molecule (0.33 nm), both GO/EDA-1 and GO/EDA-2 are expected to desalinate seawater ions efficiently. Pure GO membrane is not tested in this session since its fast swelling in aqueous environment weakens the membrane integrity within one hour.

Here, we used pervaporation method instead of conventional pressure-driven nanofiltration method on ions removal experiment because of the advantages of pervaporation method, such as higher flux and higher rejection rate. Table 5 listed the flux and ion removal results of GO/EDA-2 membrane on pervaporation and nanofiltration methods. The flux of the pervaporation was 7.8 times higher than the nanofiltration method with ion rejection rates close to 100%. Therefore, pervaporation method was used in this work.

Table 4. Radii of hydrated ions existing in seawater. Ions Hydrated ion radius Reference (nm) Na+ 0.358 [136] K+ 0.331 [136] Mg2+ 0.428 [136] Cl- 0.332 [136] 2- SO4 0.500 [137]

Table 5. Comparison of desalination with pervaporation and nanofiltration on GO/EDA-2. Desalination Flux Ion rejection rate (%) method (L/m2-h) + + 2+ - 2− Na K Mg Cl SO4 Pervaporation 1.15 98.86 100.00 99.13 99.79 99.48 Nanofiltration 0.13 12.42 15.19 85.21 28.17 84.39

The pervaporation desalination performance of GO/EDA-1 and GO/EDA-2 was also evaluated in this work. With pre-crosslinking, the membranes show excellent

+ + 2+ - 2− stability during the testing period. The rejection rates of Na , K , Mg , Cl , and SO4 ions of GO/EDA-1 and GO/EDA-2 membranes were summarized in Table 6. GO/EDA-

2 shows >99% rejection rate for all the ion species, which is better than GO/EDA-1

2+ - 2− with rejection rates of Mg , Cl and SO4 below 96.0%. Moreover, GO/EDA-2 shows larger water flux of 1.15 L/m2-h than that of GO/EDA-1 (0.69 L/m2-h). The synergistic enhancement of flux and ion rejection rate of GO/EDA-2 can be attributed to the unique hybrid GO/rGO structure with pre-crosslinking. Based on the size exclusive separation mechanism, the reduced d-spacing in GO/EDA-2 with longer crosslinking time is the major reason for the outstanding ion rejection. The enlarged water flux across

GO/EDA-2 is probably due to the introduced microdefects on the rGO sheets as evidenced by Raman results, which create extra water diffusion pathways in the membrane and thus larger water flux.

Table 6. Pervaporation desalination of GO/EDA-1 and GO/EDA-2. Ions Feed Permeate RR (%) Permeate RR (%) (ppm) (ppm) (ppm) GO/EDA-1 GO/EDA-2 Na+ 853.86 1.35 99.84 8.03 99.06 K+ 30.00

Membrane durability and the ability of regeneration were studied with 0.3 wt% sea salt solution. Experimental results do prove that RR remains above 99% after 30- hour continuous operation. In another set of experiment, the membrane was taken out and immersed in DI water for 10 minutes after each 10-hour operation. The RR is remained above 99% after three repeating cycles, while the flux is gradually decreased,

Table 7. The reduced flux is probably due to the ineffective regeneration process, where the small ions cannot be completely removed from the membrane channels. However, the membrane regeneration helps to improve the flux in general, increasing from 1.15 to 1.37 L/m2-h. The optimization of the regeneration process is still under exploration.

Table 7. Pervaporated ions removal results of GO/EDA-2 with membrane regeneration. Desalination Flux Ion rejection rate (%) time (h) (L/m2-h) + + 2+ - 2− Na K Mg Cl SO4 0-10 1.72 98.86 100.00 99.13 99.79 99.48 10-20 1.34 99.35 100.00 99.07 99.79 99.65 20-30 1.06 99.13 100.00 99.41 99.65 99.21 0-30* 1.37 99.34 100.00 99.20 99.74 99.45 0-30** 1.15 99.06 100.00 99.56 99.78 99.04 * A combined 30-hour ion removal with membrane regeneration every 10 hours. ** A 30-hour ion removal without membrane regeneration.

Facing the practical desalination, membrane separation with 3.5 wt% sea salt

+ + 2+ - 2− solution was carried out. The rejection rate of Na , K , Mg , Cl , and SO4 is all above

99%, Table 8. The water flux of 0.91 L/m2-h is smaller than that of 1.15 L/m2-h obtained

from 0.3 wt% solution, which can be explained by the faster accumulation of ions in the membrane as well as on top of the membrane. Overall, the excellent separation performance shows promising potential of such membranes in practical desalination applications.

Table 8. Desalination with 3.5 wt% sea salt solution by using GO/EDA-2. Ions Feed (ppm) Permeate (ppm) RR (%) GO/EDA-2 Na+ 8123.4 20.3 99.75 K+ 368.0

3.5. Conclusion

To sum up, this work presents a facile method to fabricate cross-linked GO/rGO hybrid membrane with outstanding permselective properties in both gas separation and desalination applications. The pre-cross-linking achieves two main goals, one is connecting GO sheets to improve stability in liquid media (restrict swelling) and the other is reducing interlayer d-spacing by GO reduction. The hybrid membranes show two different d-spacing values of ~12.0 Å and ~7.5 Å, respectively. The larger d- spacing corresponds to the expanded GO layer with inserted EDA cross-linker that enables the structural integrity of the membrane, the reduced d-spacing contributes to the excellent size exclusive separation. The hybrid membrane shows excellent permselective H2/CO2 separation with a separation factor of 22.93±1.57 and H2 permeance of 2.46±0.01× 10-8 mol.m-2s-1Pa-1. Extremely high ion (Na+, K+, Mg2+, Cl-,

2− and SO4 ) rejection rate of >99% is also obtained in a pervaporation desalination system. This work offers a new strategy to fabricate hybrid GO/rGO membrane with small d-spacing and excellent stability particularly in aqueous environment.

Considering the simple fabrication process, small d-spacing, excellent stability in liquid and outstanding separation property, the hybrid GO/rGO membrane with pre-crosslinking endows enormous potential in both gas and liquid separation fields. Considering the simple fabrication process, small d-spacing, excellent stability in liquid and outstanding separation property, the hybrid GO/rGO membrane with pre-crosslinking endows enormous potential in both gas and liquid separation fields.

CHAPTER IV

REDUCED WRINKLING IN GO MEMBRANE BY GRAFTING BASAL-PLANE

GROUPS FOR IMPROVED GAS AND LIQUID SEPARATIONS

4.1. Outline

Graphene oxide (GO) membrane assembling through edge-edge or plane-plane connection has huge influence on membrane microstructure and thus separation performance. In this work, plane-plane connection of GO sheets was promoted in supported GO membranes by grafting oxalic acid (OA) molecules in GO basal plane.

With subsequent crosslinking by ethylene diamine (EDA), better packing structure can be achieved with reduced wrinkling in GO membrane. With this unique approach, restricted swelling, tightly packed laminar structure, reduced wrinkling and improved separation property can be simultaneously achieved.

4.2. Introduction

The increasing demands of energy and clean water nowadays push for fast development of new membranes with advanced functionalities. Graphene oxide (GO) membranes have attracted great interests in recent years due to the intrinsic physicochemical properties of GO such as excellent thermal stability, chemical inertness and tunable surface functionality. With rich oxygenated groups present at both edge and basal plane, GO nanosheets can be conveniently modified for different purposes. Especially, its two-dimensional (2-D) sheet structure provides a great opportunity to fabricate membranes via stacking technologies[72, 84, 92, 109]. GO

membranes have been demonstrated excellent separation properties in both gas and liquid separation applications[52, 124-126, 138]. However, there are still a few key challenges need to be addressed before GO membranes can be practically used including but not limited to weak mechanical strength at nanoscale thickness, low membrane integrity due to swelling and limited approaches to tune d-spacing.[66, 102,

110].

Porous substrates are often used to support GO membrane and extend their application in wider range of operation conditions[41, 83, 104]. The d-spacing, interlayer distance of stacking GO nanosheets, is one of the most important factors that directly related to the separation performance of GO membrane[62, 80, 139]. The laminar structure formed by stacking GO nanosheets offers zigzag paths for molecule transportation. The size of path is determined by d-spacing. Thus, it is extremely important to control d-spacing when the size exclusion mechanism dominates in separation process. It has been reported that d-spacing of dried GO membrane and graphene membrane is ~0.80 and 0.34 nm, respectively. The steric hindrance exerted by oxygen functional groups on GO sheets is the main reason of larger d-spacing of GO membrane[80]. Meanwhile, the presence of such hydrophilic groups could further enlarge d-spacing to 6-7 nm in aqueous environment[80], which is almost ten-fold larger than the size of hydrated ions[136, 137]. Swelling not only weakens the separation performance, but also reduces mechanical strength of GO membranes. Thus, restricting GO swelling is another critical task to improve separation performance and maintain long term stability in liquid separation applications. Crosslinking is a widely adopted strategy to address swelling issue by bridging GO nanosheets via covalent bonds[6, 43, 86]. Besides crosslinking chemistry, the membrane fabrication method

also has huge influence on membrane microstructure and thus separation performance.

A recent study reveals that pressure-driven filtration can form more compacted laminar structure than the membrane prepared from vacuum filtration[65].

Recent development of modification of GO membrane for separation purpose mainly focuses on two directions, (1) reduced GO membrane[72, 133, 140] and (2) GO membrane with surface modification[141, 142]. First, by eliminating the functional groups, reduced GO nanosheets have weaker steric hindrance, which leads to a more compacted laminar structure and smaller d-spacing. However, reduced GO nanosheet sacrifices its hydrophilic property, which brings the difficulty of well-dispersion in water. Moreover, once large aggregation forms, it is hard to fabricate the membrane with good patterning. Second, it is general believed that, besides d-spacing, the surface charge of GO membrane also effects its separation performance, which can be explained by Gibbs-Donnan effect. By grafting high positively charged polyethyleneimine on surface of GO membrane, the rejections of cations, such as Mg2+,

Pb2+, Ni2+, Cd2+, and Zn2+ were improved[142]. However, such a GO modification only takes place at the surface of the membrane, the microstructure of GO membrane is changed. The area of modification on single GO nanosheet to control microstructure of

GO membrane has not been systematic studied yet.

During membrane fabrication, wrinkling phenomena of GO sheets has been found and confirmed by both experimental and simulation studies. It is widely accepted that wrinkles are often formed during GO membrane fabrication processes, such as drop-casting, spin-coating, spraying, and filtration[74, 143, 144]. The soft nature, small thickness and large 2-D dimension make it very easy to fold and form wrinkles. Once the folding occurred at early stage of membrane formation, the folding-induced

curvature will be amplified as membrane thickness increases. As a result, larger wrinkles will be formed on top of the GO membrane[74]. In a recent simulation study,

Kim et al. found that edge-edge connection and plane-plane connection of GO sheets has dramatic influence on the folding of GO sheets[145]. Edge-edge connection is more likely to form wrinkled structure and thus greatly influence the packing of GO sheets.

By reducing the wrinkles, improved Na2SO4 rejection rate from 21.3% to 85.8% was observed[74].

In this work, epoxide ring-opening reaction and subsequent modification with

OA were proceeded to enrich the functional groups on the basal plane of GO sheets. By facilitating the plane-plane connection of GO sheets with activated in-plane groups, better packing (reduced wrinkling) of GO membrane would be expected and thus superior separation performance. GO membranes were fabricated by pressure-assisted filtration method on porous polymer support. Membrane thickness was simply controlled by the amount of filtration solution. Cross-linking was introduced in the membrane to improve the mechanical strength as well as reduce swelling in aqueous media. The GO modification, cross-linking reaction, microstructure and wrinkle structure of GO membrane were systematically characterized. The prepared membranes were tested in various separation applications including H2/CO2 separation, organic dye separation from water, and desalination.

4.3. Experimental Procedures

4.3.1. Materials

The materials for synthesizing GO are showed in 3.3.1. Methylene blue (MB,

C16H18ClN3S.3H2O) were purchased from Fisher Scientific. Ethylenediamine (EDA, ≥

99.0%), dopamine hydrochloride, oxalic acid (C2H2O4, anhydrous, ≥ 99.0%),

hydrobromic acid (HBr, ≥ 48.0%), and Trizma base (≥ 99.9%) were all purchased from

Sigma Aldrich. Polyethersulfone (PES, 0.1 µm pore size) membrane filters was purchased from Sterlitech Corporation.

4.3.2. GO synthesis and modification

GO synthesis follows a few main steps including pre-oxidation, oxidation, washing and dialysis, which can be found in section 3.3.2. Then, GO was modified by oxalic acid following the procedure detailed as following. Firstly, the prepared GO solution (12.7 g/L after dialysis) was diluted in DI water to 0.4 g/L. Then 5.0 mL hydrobromic acid was added into 30 mL GO aqueous solution to open the in-plane epoxide ring and form hydroxyl groups. The mixture was stirred at ambient temperature for 12 hours. Then 1.5 g oxalic acid was added into the mixture and keep stirring for another 6 hours. The product was carefully filtered until the filtrate turned to neutral.

Finally, the oxalic acid modified GO (OAGO) was re-dispersed in DI water at concentration of 0.8 g/L for following membrane fabrication.

4.3.3. Membrane fabrication

To introduce crosslinking between the GO sheets during membrane formation,

EDA was used as a crosslinker in this work. Specifically, 0.8 g/L OAGO was mixed with 0.4 mol/L EDA with volume ratio of 1:1. The mixture was magnetically stirred and heated at 85 °C for 1 hour to initiate the reaction between OAGO and EDA. A gradual color change of the solution was observed from brown to black. The product was then allowed to cool down to ambient temperature. A control sample is also prepared following the same procedure by using GO instead of OAGO. Pure GO without adding EDA was also prepared for comparison. Figure 28 shows the synthetic procedure of GO/EDA and OAGO/EDA, respectively. It is well known that GO sheet

has rich -OH and -COOH groups at edge area and epoxide groups in basal plane[41, 56,

146]. The addition of EDA in GO solution makes the edge-edge connection of GO sheets as illustrated in Figure 28(a). Modified GO (OAGO) has rich –COOH groups at both edge and basal plane. Therefore, it is expected that the crosslinking would occur at both edge area and basal plane. With more bonding sites for such cross-linking, the assembled GO sheets could have better uniformity and tightly packed layer structure,

Figure 28(b).

Figure 28. Reactions involved in the membrane fabrication of (a) GO/EDA and (b) OAGO/EDA.

To ensure mechanical strength, GO membrane was supported on porous PES film with polydopamine (PDA) coating in between. Specifically, pure PES membrane was cut into 55 × 55 mm square shape and soaked into 50 mL solution contains 0.1 g dopamine and 0.06 g Trizma base. The polymerization was carried out in a shaker at ambient temperature for 24 hours, followed by 30 minutes drying in oven at 80 °C. The color of PES membrane changed from white to grey after the coating process, Figure

29(a). The dried PES/PDA membrane was cut into proper size and sealed by O-rings in the separation cell. The prepared GO/EDA (or OAGO/EDA) solution was then added

into the cell and 300 psi pressure was applied to filtrate the water out of the solution for

30 minutes and left the GO sheets on top of the PDA coated PES film, Figure 29(c). At last, the supported GO membrane was heated at 80 °C for 2 hours to complete the crosslinking, Figure 29(d). The membranes are named as GO/EDA-X and

OAGO/EDA-X based on the amount of 0.4 g/L solution used for membrane fabrication

(X=0.5, 1, and 2 represent the weight of 0.4 g/L GO or OAGO solution used in gram, respectively).

4.3.4. Characterization

The cross-section morphology of membrane was characterized by scanning electron microscope (SEM, JEOL-7404, 5kV) with sputter-coated silver layer on top of the sample surface. The d-spacing of the supported membranes was characterized by

X-ray diffraction (XRD, Bruker AXS D8 Discover diffractometer with General Area

Detector Diffraction System, 40 kV, 35mA) with scan rate of 1.0 degree/min within the range of 5-15 degree. The d-spacing is calculated by following Bragg’s law:

2푑 sin  = 푛

where d is the d-spacing value;  is the angle of incidence/scattering; n is the order of diffraction, which equals to 1; and  is wavelength of x-ray, which is 0.154 nm.

The elemental composition of all samples was analyzed by X-ray photoelectron spectroscopy (XPS, PHI VersaProbe II Scanning XPS Microprobe with Al Kα line excitation source). The surface morphologies were characterized by atomic force microscopy (AFM, Park System XE7). The details of H2/CO2 binary gas separation measurements are listed in 3.3.4

The dye separation test was performed in a dead-end separation cell. The prepared membrane was sealed in the separation cell being supported on a porous metal cylinder. 250 mL dye solution was then added into the cell for separation test.

Methylene blue was prepared at two different concentrations of 10 and 70 ppm. The dimension of the MB molecule is about 17.0 × 7.6 × 3.3 Å[147]. The separation was carried out at 100-300 psi. The permeate liquid was continuously collected for 6 hours.

The volume of liquid was measured for flux calculation, and UV-vis was used to evaluate rejection rate. The intensity values were obtained at peak wavelength of 663.6 nm for MB solution.

The permeation flux 퐽 can be calculated by Equation (11):

푉 퐽 = (11) 퐴×푡 where 푉 (L) is the volume of permeate of dye solution; 퐴 (m2) is the working membrane area; 푡 (h) is the operation time. And the rejection rate of dye molecule is calculated by Equation (12):

퐼 휂 = 1 − 푖 (12) 퐼0,푖 where 퐼0,푖 and 퐼푖 are the intensity of UV-vis of solution 푖 in the feed and permeate side, respectively. The details of desalination are listed in 3.3.4.

4.4. Results and Discussion

The thickness of membrane not only affects its microstructure, it is also one of the important parameters that influences separation. Thin membrane is preferred to facilitate large flux, but it generally results in lower selectivity. In this work, membranes with three different thicknesses were prepared on top of porous PES support with a thickness of about 100, 200 and 400 nm, respectively, Figure 30. Obviously, the increase of membrane thickness followed the same trend as increasing the amount of solutions used for membrane filtration even though small deviation was also observed.

Since part of the GO sheets could penetrate through the membrane pores or block the pores at the early stage of filtration, the small difference in membrane thickness is within our expectation. It is worth noticing that the GO sheets formed a well packed laminar structure, indicating the effectiveness of pressure-driven filtration method in producing quality membranes[65].

Figure 30. Cross section SEM images of (a) GO/EDA-0.5, (b) GO/EDA-1, (c) GO/EDA-2, (d) OAGO/EDA-0.5, (e) OAGO/EDA-1, and (f) OAGO/EDA-2.

To better understand the OA modification and crosslinking reaction between

GO (or OAGO) and EDA, the elemental composition of pure GO, GO/EDA-2 and

OAGO/EDA-2 membranes was analyzed by XPS, Figure 31. With the addition of EDA in the membrane, a new N 1s peak was observed at 399.5 eV in both GO/EDA-2 and

OAGO/EDA-2 membranes. The atomic percentage of N element in GO/EDA-2 and

OAGO/EDA-2 membranes is 4.8 and 6.7%, respectively, Figure 31(b-c). The larger fraction of N element in OAGO/EDA-2 can be attributed to a higher cross-linking density due to the presence of reactive –COOH groups in basal plane of GO after modification. The significant drop of O/C ratio from pure GO (0.407) to GO/EDA-2

(0.166) is attributed to the elimination of –OH from –COOH during the amidation reaction between GO and EDA. The O/C ratio in OAGO/EDA is about two times of

GO/EDA. Considering the molecular structure of OA itself with O/C ratio of 2 and the crosslinking reaction only consumes a quarter of oxygen in OA, it is reasonable to observe a higher O/C ratio in the OAGO/EDA membrane.

Figure 31. XPS wide scan of (a) pure GO, (b) GO/EDA-2, and (c) OAGO/EDA-2 membranes. The atomic percentages of each membrane (C, N and O) are at the top right corner.

In addition to that, high resolution XPS spectra was performed for all three membranes and C1s peak was carefully analyzed, Figure 32. The C1s peak can be deconvoluted into four peaks that are C=C at 284.4 eV, C-O at 285.9 eV, C-O-C at 286.6 eV, and C=O at 288.0 eV, Figure 32(a), which is highly consistent with other literature

reports [86, 126]. Direct mixing of GO and EDA (GO/EDA-2 membrane) led to two significant changes: (1) decrease of C-O-C peak at 286.6 eV and (2) increase of C-O/C-

N peak at 285.9 eV, Figure 32(b). The C-O-C peak represents the epoxide group in GO basal plane. The decrease of C-O-C and increase of C-O/C-N in GO/EDA-2 indicate the occurrence of amidation reaction between GO and EDA, which is also reported in previous literature[86]. The C-O-C peak is further reduced in OAGO/EDA-2, which reveals the successful ring-opening reaction by hydrobromic acid that promotes the surface grafting of OA and also subsequent crosslinking reaction. All these results confirm the successful surface grafting of OA molecules on GO basal plane and promoted crosslinking in the membrane.

Figure 32. XPS elemental analyses of (a) pure GO, (b) GO/EDA-2, and (c) OAGO/EDA-2 membranes in C1s.

Figure 33 summarized the XRD results from the studied membranes. GO membrane showed a peak at 2θ=11.6o, and the corresponding d-spacing was calculated as 0.76 nm from Bragg’s equation which is similar to the previously reported value [80].

The d-spacing of GO/EDA-2 membrane increased to 0.95 nm, confirming the increased steric hindrance between GO sheets by the inserted EDA molecules. After OA modification and crosslinking, the d-spacing of OAGO/EDA-2 membrane was surprisingly decreased to 0.88 nm. Comparing with the GO/EDA-2 membrane, OA molecules were inserted between the GO layers. This was expected to pose additional

steric hindrance and thus further expanded the inter-layer spacing of OAGO/EDA-2 membrane. On the other hand, the grafting of OA molecules on GO basal plane created the uniformity of reactive sites across the edge and basal plane areas. As a result, crosslinking reaction would occur across the GO plane and thus better packing structure of GO sheets can be achieved. The d-spacing results provided direct evidence of more compact structure of OAGO/EDA membrane than that of GO/EDA. A proposed packing structure was provided in Figure 34. The edge-edge connection of GO/EDA led to a wrinkled structure, while the plane-plane connection in OAGO/EDA was more likely to form a highly ordered laminar structure, Figure 34(a&b). A result of such different packing behavior can be seen from the membrane surface topography and roughness.

Figure 33. XRD results of (a) pure GO, (b) GO/EDA-2, and (c) OAGO/EDA-2 membranes.

Figure 34. Schematic of microstructure of (a) GO/EDA and (b) OAGO/EDA membrane.

AFM was then used to characterize the surface topography and roughness (Rq) of the GO membranes of different thicknesses. The surface topography and selected height profiles were shown in Figure 35 and 36, respectively. The thinnest GO/EDA-

0.5 (or OAGO/EDA-0.5) membrane has no obvious wrinkles on the surface, Figure

35(a&d). With the increase of the membrane thickness, wrinkles became obvious since wrinkles grow up gradually with the deposition of GO nanosheets[74], Figure 35(a-c, d-f). To further study the influence of OA grafting on membrane surface topography and roughness, the line profiles were analyzed for all membranes. The height profile along red lines and overall surface roughness were summarized in Figure 36. The largest height difference of each line was also measured and marked in the figure.

GO/EDA-0.5, GO/EDA-1, OAGO/EDA-0.5, and OAGO/EDA-1 exhibited similar value of Rq, which is about 60 nm or less, Figure 36(a, b, d, &e). When further increasing the membrane thickness, OAGO/EDA-2 remained similar surface roughness of 59.2 nm, as well as the largest height difference of 350 nm. As a comparison, the surface roughness of GO/EDA-2 increased to 113.6 nm, which is almost twice of other membranes. This can be attributed to the formation of large wrinkles. Meanwhile, the largest height difference of ~600 nm was observed, Figure 36(c1&c2). It can be concluded that the surface grafting of OA molecules helps to form a better laminated

GO membrane structure with reduced wrinkling.

Figure 35. Surface AFM images of (a) GO/EDA-0.5, (b) GO/EDA-1, (c) GO/EDA-2, (d) OAGO/EDA-0.5, (e) OAGO/EDA-1, and (f) OAGO/EDA-2 membranes.

Figure 36. Height profiles of selected membrane area in Figure 35.

Membranes of different thicknesses were assembled in a dead-end membrane filtration system to separate MB from water. The separation tests were performed at different pressure conditions, i.e. 100, 200 and 300 psi. Due to the hydrophilic nature of pure GO membrane, severe swelling occurred during separation and membrane failure was observed after only 10 minutes. Both GO/EDA and OAGO/EDA showed

excellent stability. Figure 37 summarized the rejection rate and water flux results of

GO/EDA membrane under different pressure conditions by using two MB model solutions at 10 and 70 ppm. In general, smaller operation pressure and thicker membrane are beneficial to reach a higher rejection rate. Specifically, increasing pressure from 100 to 300 psi led to a big drop of MB rejection rate from 96.5 to 89.0%,

Figure 37(a). Similar drop was also observed with 70 ppm MB solution. Accompanied with the decrease of rejection rate, water flux is doubled by increasing the pressure from

100 to 300 psi. The tradeoff between selectivity and permeability is widely reported in almost all membranes[148-151]. Larger pressure drives faster diffusion rate of both water and MB molecules through the membrane channels, where permeant-membrane interaction becomes less dominant in the separation process. Similar tradeoff was also observed when varying the membrane thickness under 100 psi. The thicker membrane provided a longer diffusion path that restricted the molecular transport in a more efficient way. Therefore, smaller flux and larger rejection rate were observed as membrane thickness increases, Figure 37(c&d).

Figure 37. MB separation results of GO/EDA membranes as a function of pressure (a- b) and membrane thickness (c-d).

Similar separation test was performed on OAGO/EDA membranes. As shown in Figure 38, the water flux increased with increasing operation pressure, which was similar to the GO/EDA membrane. It is worth noticing that rejection rate seems not much dependent on applied pressure. The MB rejection rate kept as high as 99.8% in all the testing conditions, Figure 38(a&b). The rejection rate is also less dependent on the membrane thickness expect the one with the smallest thickness, Figure 38(c). The optimal separation results were achieved with rejection rate of 99.94% and water flux of 1.12 L/(hr*m2) (LMH) by OAGO/EDA membrane.

Figure 38. MB separation results of OAGO/EDA membranes as a function of (a-b) pressure and (c) membrane thickness.

Once confirmed the superior property in MB separation, pure GO, GO/EDA-2 and OAGO/EDA-2 membranes were further tested for equimolar H2/CO2 separation.

The separation results were summarized in Table 9, where a separation factor of 12.96,

10.97, and 16.14 was achieved in pure GO, GO/EDA-2, and OAGO/EDA-2 membranes, respectively. It has been proved by XRD that OAGO/EDA-2 owned a larger d-spacing than pure GO membrane. Following the size-exclusive mechanism, a higher separation factor can be expected from GO membrane. The unexpected higher separation factor from OAGO/EDA-2 clearly revealed that d-spacing was not the only factor influencing the H2/CO2 separation and other factors needed to be explored as well. When EDA was used as crosslinker in OAGO/EDA membrane, the amide group after crosslinking reaction as well as the possible unreacted amine group were expected to have strong affinity with CO2, thus restricted CO2 diffusion, and increased nonadsorbing H2 diffusion through the membrane[152-156]. With such unique features, OAGO/EDA-2 membrane significantly enhanced H2 mass transport efficiency and H2/CO2 selectivity.

Table 9. H2/CO2 binary gas separation results with pure GO, GO/EDA-2, OAGO/EDA- 2 membranes. -9 -2 -1 -1 Samples αH2/CO2 Pm,H2 (×10 mol.m s Pa ) GO 12.96 0.99 GO/EDA-2 10.97 8.16 OAGO/EDA-2 16.14 13.62

+ + 2+ - 2- It was reported that the diameters of hydrated Na , K , Mg , Cl , SO4 ions are

0.72, 0.66, 0.86, 0.66, and 1.00 nm, respectively[136, 137]. The rejection rates of all ions for GO/EDA-2 and OAGO/EDA-2 were summarized in Table 10. Both membranes showed excellent ion rejection especially for K+ where 100% rejection rate was achieved. GO/EDA-2 membrane exhibited excellent rejection rates of 99.25%,

+ 2+ - 2− 96.98%, 99.21%, and 99.78% for Na , Mg , Cl , and SO4 respectively. After OA

modification, the OAGO/EDA-2 membrane further improved the rejection rates to

99.61%, 98.05%, 99.62%, and 99.99%. In addition, OAGO/EDA-2 membrane also acquired a larger water flux of 22.41 LMH compared to 17.17 LMH of GO/EDA-2.

Though the simultaneous enhancement of ion rejection rate and water flux is still under exploration, the benefits of grafting oxalic acid on GO plane is obvious for fabricating better performed GO membranes.

Table 10. Pervaporation desalination of GO/EDA-2 and OAGO/EDA-2. Ions Feed Permeate RR (%) Permeate RR (%) (ppm) (ppm) (ppm) GO/EDA-2 (17.17 LMH) OAGO/EDA-2 (22.41 LMH) Na+ 815.50 6.11 99.25 3.18 99.61 K+ 36.00 < DL 100.00 < DL 100.00 Mg2+ 111.45 3.37 96.98 2.17 98.05 Cl- 1543.75 12.20 99.21 5.87 99.62 2- SO4 211.02 0.46 99.78 0.02 99.99 *DL: detection limit.

4.5. Conclusion

To sum up, this work provided a novel modification method to fabricate OAGO membranes with excellent permselectivity in gas separation, dye removal, and desalination applications. The OA modification introduced active functional groups on the basal plane of GO nanosheets, and subsequent crosslinking with EDA improves membrane stability in liquid environment. After modification and crosslinking, strong plane-plane connection was constructed with tightly ordered laminar structure. The

OAGO/EDA-2 membrane showed good permselective H2/CO2 separation with a

-9 -2 -1 -1 separation factor of 16.14 and H2 permeance of 13.6 × 10 mol.m s Pa . In MB separation tests, the rejection rates of 70 ppm MB solution of OAGO/EDA-1 and

OAGO/EDA-2 membranes achieved nearly 100% under 300 psi. The overall high ions

+ + - 2- 2+ (Na , K , Cl , and SO4 and Mg ) rejection rates of > 98.1% were also obtained in a

pervaporation desalination system. This work offers a new strategy to fabricate modified GO membrane with strong plane-plane interaction and better ordered microstructure. Considering the simple fabrication method, controlled d-spacing, excellent stability, and flexible choice of grafting molecules, the modified GO membrane has enormous potential to be developed in the future and to be applied in both gas and liquid separation fields.

CHAPTER V

GRAPHITE OXIDE/BORON NITRIDE HYBRID MEMBRANE: THE ROLE OF

CROSS-PLANE LAMINAR BONDING FOR A DURABLE MEMBRANE WITH

LARGE WATER FLUX AND HIGH REJECTION RATE

5.1. Outline

Swelling of graphene oxide (GO) membrane greatly restricts its use in liquid separation applications. Cross-linking is widely used to stabilize GO membrane but often results in enlarged d-spacing. This work presents a durable hybrid GO/boron nitride membrane with both physical confinement and chemical bonding, which simultaneously achieves large water flux and excellent dye rejection rate. By alternating

GO nanosheets and amino-modified boron nitride (a-BN) nanosheets in the hybrid membrane (a-BNGO) structure, plane-plane conformation and edge-edge covalent bonding facilitate the formation of tightly packed laminar structure, which is demonstrated efficient to restrict membrane swelling, improve water flux and methylene blue (MB) separation performance.

5.2. Introduction

Fresh water supply is becoming a worldwide challenge nowadays. The increasing demands of clean water accelerate the development of clean water technologies in industry. Among current practicing technologies including low temperature distillation, freezing method, electrodialysis, and dewvaporation method, membrane separation is considered one of the most efficient and economic processes.

Both ionic and molecular separations can be achieved by membranes such as salt[121,

157, 158], organic dye[159-161] and heavy metals[41, 142, 162, 163]. Although polymeric membranes can be used in a wide range of separation processes, they also suffer from a few limitations, such as low operation temperature, swelling in organic solvents, and fast aging rate in harsh environment[157, 164]. To maintain excellent separation performance and satisfactory stability in crucial conditions, GO-based materials are attracting increased attention in recent years due to its excellent chemical and thermal stability[33, 60, 92, 122, 123, 165]. More importantly, the unique 2-D structure and rich surface group of GO nanosheets offer a set of unique features for facile membrane fabrication via simple techniques such as filtration and pressurization[165]. Previous literatures reported that GO membranes have been fabricated and showed promising properties in liquid separation applications[72, 128,

166, 167]. However, a few disadvantages of GO membranes have been identified including but not limit to weak mechanical strength at nanoscale thickness, poor long- term stability in liquid separation due to the significant swelling, and expanded d- spacing after introducing cross-plane linkers.

Porous polymer membranes are often used as mechanical support to expand the operational window of GO membranes[43, 142, 165]. GO membrane is formed by stacking multiple layers of GO nanosheets, which results in a sub-nanometer interspacing between the sheets that is often called d-spacing. The d-spacing has been demonstrated as one of the most important parameters that affect the separation performance of GO-based membranes[62, 168]. The d-spacing offers transport channels for small molecules while screening out large molecules, which can be described as size exclusion mechanism. It has been clearly stated that the d-spacing of

dry graphene and GO membrane are about 0.34 nm and 0.8 nm respectively[80]. The larger d-spacing of GO membrane can be explained by steric hindrance formed by oxygen functional groups on GO nanosheets. Meanwhile, the hydrophilic functional groups prefer to attract water molecules in aqueous environment, which leads to a huge expanding of d-spacing of GO membrane. The d-spacing of GO membrane can reach to 6-7 nm due to swelling, which is much larger than most of the hydrated dye molecules, not to mention the smaller salt ions[2, 5, 80, 114, 136, 137]. To enable long- term durability of GO membranes in liquid separations, swelling suppression is the first and most important issue needs to be addressed. The most common strategy to address swelling is to cross-link GO nanosheets via covalent bonds or form strong interaction via electrostatic attraction[6, 86, 168]. Moreover, the fabrication method of GO membrane also has a huge influence on microstructure and the resulting separation performance. A recent work demonstrated that pressurization filtration method could lead to a more compacted stacking laminar microstructure compared to simple vacuum filtration method, and enhanced separation performance was observed[65].

Recent working on using cross-linkers to bridge GO membrane for improved stability purpose in liquid separation mainly focuses on two connection (1) electrostatic attraction[45, 169] and (2) stronger covalent bonds[6, 47, 112, 124, 125]. First, the abundant negatively charged functional groups on GO nanosheets offer an opportunity to electrostatic attraction by introducing positively charged species. Such an electrostatic attraction could help to withstand the d-spacing expanding caused by trapped water molecules. However, the electrostatic attraction is not strong enough to ensure the long-term stability of GO membrane, especially under harsh conditions in practical liquid separation applications. Second, the rich oxygen functional groups at

the edge of GO flakes can be utilized as active reaction sites for covalent connection with designed cross-linkers, like two-amine species[86]. The covalent bonds offer great potential for GO membrane to stay stable in liquid separation tests since they are much stronger than electrostatic attraction. However, the introduction of cross-linkers could enlarge the d-spacing even under dry condition[166]. It is generally believed that cross- linkers will be placed between GO nanosheets. For example, Hung et al. fabricated GO membranes cross-linked by three diamine cross-linker, ethylenediamine (EDA), butylenediamine (BDA), and p-phenylenediamine (PPD). The d-spacing of prepared membranes increased with increasing of lengths of cross-linkers[86]. From previous work of our group, we found out that the d-spacing of EDA cross-linked GO membrane was enlarged to about 0.95 nm, while the value of pure GO membrane is only 0.76 nm[166].

The laminar boron nitride (h-BN), which shares similar structure with graphene, is promising 2-D material for membrane fabrication[170]. Recent work reported that h-

BN can be exfoliated to few layers and modified with amino groups[171]. Thin BN flakes have advantages to form laminar structure by stacking process, while amino groups offer active reaction sites for cross-linking. It makes a-BN an excellent candidate to bridge GO nanosheets and form compacted laminar structure together with

GO. What’s more, unlike the short-chain cross-linker, a-BN nanosheets are not cross- linker only, but also the part of layered material of hybrid membrane.

In this work, a-BN nanosheets were prepared from bulk BN by ball milling method with urea as agent. By cross-linking GO nanosheets with a-BN nanosheets, the a-BNGO hybrid membrane with narrow d-spacing and long-term stability in liquid separation applications would be expected and thus advanced methylene blue (MB)

separation performance. The a-BNGO hybrid membranes were fabricated by pressurization filtration method on porous polyethersulfone (PES) substrate with polydopamine (PDA) coating. Membrane thickness can be simply controlled by the amount of GO and a-BN nanosheets. A two-step cross-linking process was processed to ensure the covalent bonds connection, one was applied in a-BNGO mixture solution before membrane fabrication and another was applied after membrane was formed. The a-BN modification, cross-link reaction, microstructure and surface morphology and properties were characterized systematically. The prepared a-BNGO membranes were tested in MB separation from water under different pH conditions.

5.3. Experimental Procedures

5.3.1. Materials

The materials for synthesizing GO can be found in 3.3.1. Methylene blue (MB,

C16H18ClN3S·3H2O) were purchased from Fisher Scientific. Boron nitride (powder,

~1µm, 98%), and urea (CH4N2O, powder) were purchased from Sigma Aldrich. The polyethersulfone (PES, 0.1 µm pore size) membrane filters were purchased from

Sterlitech Corporation.

5.3.2. Preparation of GO and a-BN nanosheets

The GO was synthesized by modified Hummer method. Four main steps are involved including pre-oxidation, oxidation, washing and dialysis. The experimental details of synthesis refer to section 3.3.1. After the four steps, GO aqueous solution of

12.7 g/L can be obtained. The preparation of a-BN is detailed as follow. Firstly, BN powder was dry milled with urea in ball milling, urea is used as modification agent to graft amino groups on BN nanosheets. The BN:urea mass ratio was controlled at 1:10 and the milling time is 24 hours at 500 rpm. Thereafter, the obtained powder was

washed with water, separated via centrifuge and re-dispersed in water until the supernatant reaches pH=7. After final dispersion, a-BN solution was centrifuged at

5000 rpm for 30 mins to remove large aggregates. The concentration of a-BN solution was determined to be 1.94 g/L. Both GO and a-BN aqueous solutions were diluted to

0.4 g/L for following membrane fabrication. The size of both GO and BN sheets was measured by AFM and the corresponding topography images and line profiles were summarized in Figure 39. It is worth mentioning that size distribution is present in the sample, and larger pieces of samples were analyzed in this work. Specifically, a-BN nanosheets are about 1 μm in lateral size and 150 nm in thickness, while GO nanosheets have lateral size of 30-50 μm and thickness of 10-20 nm.

Figure 39. AFM images of (a) BN nanosheets and (b) GO nanosheets, and (a1,a2,b1,b2) their corresponding height profiles in nm.

5.3.3. Membrane fabrication

The a-BN nanosheets were used as laminar cross-linker to provide interfacial bonding within the GO-a-BN-GO hybrid membrane structure. Specifically, 2.0 g GO solution was mixed with certain amount (0.5, 1.0, 2.0, 3.0 and 4.0 g) of a-BN solution to prepare the membrane precursor solution. Both solutions are at concentration of 0.4

g/L. The resulting membranes were named as a-BN#GO (# means the mass ratio of a-

BN solution/GO solution, which is 0.25, 0.5, 1.0, 1.5 and 2.0 respectively). Following on the mixing, the mixture solution was further heated at 85 °C for 2 hours with the aim to pre-cross-link GO and a-BN nanosheets before membrane fabrication. During this process, the color of mixtures gradually changed from yellow to black. After that, the mixtures were cooled down to room temperature, which remains a homogeneous solution without obvious aggregates or precipitation. It is generally accepted that GO nanosheets have rich -OH and -COOH groups at the edge, as illustrated in Figure 40(a).

Meanwhile, a recent work[172] demonstrates that bulk BN can be exfoliated and modified with amino groups at the edge by using a ball milling process, and the representative structure is shown in Figure 40(b). During pre-crosslinking, it can be expected that part of the edge groups from a-BN and GO will be reacted to form an edge-edge connection with covalent bonding, Figure 40(c). Once processed into membrane, a laminar hybrid GO-a-BN-GO structure with cross-plane covalent bonding could be expected, Figure 40(d). It is also worth mentioning that GO has been partially reduced during this process as evidenced by the color change of solution as well as the corresponding characterization.

Figure 40. The schematic structure of (a) GO, (b) a-BN nanosheets, (c) crosslinking reaction between GO and a-BN, and (d) scheme of compact laminar structure of a- BNGO membrane.

To improve the mechanical strength of membrane, porous PES coated with polydopamine (PDA) (PDA/PES) was used as supporting substrate. The details of PDA coating can be found in 4.3.3. The color change from white to grey indicates the successful polymerization of dopamine monomer. For membrane preparation,

PDA/PES substrate was firstly fixed in a cylindrical cell, Figure 41(b). The pre-cross- linked a-BN-GO solution was diluted to 20 mL and transferred to the cell. The cell was then sealed and pressurized at 300 psi. The water can be separated out of the mixture solution in about 30 minutes and the filtrate forms a uniform a-BNGO membrane on top of the PDA/PES substrate, Figure 41(c). Finally, the supported membrane was transferred to a convection oven and heated at 80 °C for 2 hours. The final a-BNGO membrane shows very smooth surface without visible defects or cracks, Figure 41(d).

For comparison, pure GO membrane was also prepared following the same procedure without adding BN.

Figure 41. Schematic process of membrane fabrication. (a) PES membrane before and after PDA coating, (b) separation cell with PDA/PES between two O-rings, (c) sealed PDA/PES and prepared a-BNGO solution in separation cell under 300 psi pressure, and (d) final a-BNGO membrane.

5.3.4. Characterization

FT-IR spectrum was characterized by Nicolet iS10 FT-IR Spectrometer. The cross-section morphology of membrane was characterized by scanning electron microscope (SEM, JEOL-7404, 5kV) with sputter-coated silver layer on top of the sample surface. The elemental composition of all samples was analyzed by X-ray photoelectron spectroscopy (XPS, PHI VersaProbe II Scanning XPS Microprobe with

Al Kα line excitation source). The size of GO and a-BN nanosheets and surface morphology were characterized by atomic force microscopy (AFM, Park System XE7).

The contact angle values were characterized by NRL C.A. goniometer, model 100-00

(Rame-hart, Inc). The d-spacing of the supported membranes was characterized by X- ray diffraction (XRD, Bruker AXS D8 Discover diffractometer with General Area

Detector Diffraction System, 40 kV, 35mA) with scan rate of 1.0 degree/min within the range of 5-15 degree. The d-spacing is calculated by Bragg’s law, its equation is listed in 4.3.4. The details of MB separation tests are all showed in 4.3.4.

5.4. Results and Discussion

FT-IR characterization was used to confirm the functional groups in GO, a-BN nanosheets and the bonding interaction in between. Figure 42(a) compares the FT-IR spectrum of BN before and after modification. Two intense peaks at about 750-800 and

1320-1370 cm-1 were assigned to the in-plane B-N stretching vibration and out-plane

B-N-B bending vibration, respectively[171]. The enlarged spectrum in the range of

3200-3700 cm-1 shows new peak formation, which corresponds to the N-H stretching vibration of amino-group on the a-BN nanosheets. The a-BN nanosheets remain excellent dispersion in water for several weeks. Figure 42(b) shows the FT-IR results of GO, a-BN, and a-BNGO after the cross-linking. For GO, the peaks at 1720, 1620,

1364, 1055 and 3250 cm-1 can be assigned to the C=O, C=C, C-OH, C-O, and O-H respectively, which are consistent with other literatures[173, 174]. The presence of these oxygen-containing functional groups allows excellent dispersion GO nanosheets in water. A new peak appeared at 1267 cm-1, which confirms the presence of C-N bonding arising from the cross-linking reaction between GO (-COOH) and a-BN (-

NH2).

Figure 42. The FT-IR spectrum of (a) BN and a-BN, and (b) a-BN, GO and a-BNGO.

To further confirm the amide bonds between GO and BN, the elemental composition of pure GO and a-BN2.0GO membranes were comparatively analyzed by

XPS, Figure 43(a). A new N 1s peak was observed at 398.0 eV in a-BN2.0GO membrane, indicating the presence of BN in the membrane. High resolution XPS spectra was then performed for pure GO and a-BN2.0GO membranes, and both N 1s and C 1s were carefully analyzed, Figure 43(b-d). The new N 1s peak of a-BN2.0GO can be deconvoluted to two peaks: B-N at 398.1 eV from a-BN, and amide C-N at 399.7 eV formed by reaction between a-BN and GO, Figure 43(b). The C 1s peak of pure GO membrane can be deconvoluted into five peaks those are C=C at 284.8 eV, C-O at 286.3 eV, C-O-C at 287.0 eV, C=O at 288.4 eV, and O-C=O at 289.4 eV, Figure 43(c), these results are highly consistent with literatures[86, 175]. Two significant changes were noticed after reaction: (1) decrease of C-O-C peak at 287.0 eV and (2) increase of C-

O/C-N peak at 286.3 eV. The C-O-C peak represents the epoxide group in GO basal plane, the decrease of this peak can be a result of GO reduction during the pre-crosslinking process. The prominent increase of C-O/C-N peak indicated the successful amidation reaction between a-BN and GO, which is also reported in previous literature[86]. All these results confirm the successful reaction of a-BN and GO, and promoted cross-linking in the membranes.

Figure 43. (a) XPS wide scan of pure GO and a-BN2.0GO. The XPS elemental analyses of (b) a-BN2.0GO in N 1s, (c) pure GO in C 1s, and (d) a-BN2.0GO in C 1s.

Due to the hydrophilic nature of GO sheets, pure GO membrane tends to swell in water and results in significantly enlarged d-spacing. Therefore, GO membrane has been rarely used in liquid separations. Having a similar 2-D sheet structure as GO, a-

BN could be processed into membrane as well on top of the PDA/PES substrate, Figure

44(a). Noticing that the membrane surface is smooth and uniform, and as flexible as

GO membrane. By running a separation test on a-BN membrane with MB aqueous solution, slight color change was observed indicating a poor separation performance,

Figure 44(b). Further investigation on the microstructure of membrane by SEM, it is found that a-BN sheets are loosely packed with obvious stacking voids, Figure 44(c).

Such voids provide wide-open channels for quick passing through of the solution without filtering, resulting in poor separation. Indeed, both GO and a-BN membranes have their limitations in separation, either due to the surface chemistry or structural

defects. However, the combination of GO and a-BN into a hybrid membrane could provide a solution by addressing the swelling and structural defects issues.

Figure 44. The digital images of (a) pure a-BN membrane and (b) 70 ppm MB solution before and after separation. The surface SEM image of pure a-BN membrane.

In this work, pre-cross-linking method was used before membrane fabrication to create partial inter-connection between GO and a-BN nanosheets. By pre-crosslinking process, partially connected network of GO and a-BN nanosheets can be formed and the stability can be ensured in water separation applications. The pre-cross-linking process changes reaction solution from yellow to black at all different mass ratios, as example of a-BN1.0GO and a-BN2.0GO, Figure 45(a&b). The color change of the reaction solution is due to the elimination of functional groups on GO nanosheets in two different ways: one is GO reduction; the other is amide reaction occurred between a-BN and GO[131, 140]. The final a-BNGO membranes were filtrated on PDA coated

PES substrates with good surface quality, Figure 45(c). No crackers or defects can be found on these a-BNGO membranes, and they show excellent stability in water while pure GO membrane suffered with severe swelling issue and structural integrity damage.

It is worth mentioning that the PDA coating on PES substrate is also essential in this work, since it provides reaction sites to enable strong covalent connection between a-

BNGO and PES substrate and also better stability of a-BNGO membrane[166].

Figure 45. The digital images of a-BN1.0GO and a-BN2.0GO solutions (a) before and (b) after pre-cross-linking and (c) the corresponding membranes.

The thickness of pure GO membrane is about 100 nm, Figure 46(a). With the addition of a-BN, the membrane thickness increases to a wide range of 195 to 1210 nm as the a-BN/GO ratio increases from 0.25 to 2, Figure 46(b-f). Although the increase of the membrane thickness does not follow a strict proportional relationship, only small deviation was observed probably due to the non-uniform size of both GO and a-BN

nanosheets that generate certain randomness of packing. It is worth noticing that, unlike the pure a-BN membrane, all the hybrid membranes including a-BN2.0GO show well- oriented laminar structure, clearly implying a synergistic effect between GO and a-BN in facilitating ordered-packing of membrane structure.

Figure 46. Cross-section SEM images of (a) pure GO, (b) a-BN0.25GO, (c) a-BN0.5GO, (d) a-BN1.0GO, (e) a-BN1.5GO, and (f) a-BN2.0GO.

Figure 47 summarizes the XRD results of GO and a-BNGO membranes. Pure

GO membrane shows a signature peak at 2θ=11.8o corresponding to a d-spacing value of 0.82 nm, which is consistent with previous report[86]. The d-spacing values of the hybrid membranes are calculated as 0.82, 0.81, 0.78, and 0.80 nm for a-BN0.5GO, a-

BN1.0GO, a-BN1.5GO and a-BN2.0GO, which reveals that the addition of a-BN nanosheets in the hybrid membrane does not increase the d-spacing, instead, slightly decrease. The reduced d-spacing of hybrid membranes can be explained from following perspectives: firstly, compared to GO sheets, the in-plane area of a-BN nanosheets is relatively clean without dangling groups. Thus, less steric hindrance is expected between the nanosheets; secondly, with reduced steric hindrance, a tightly packed plane-plane conformation can be expected, which adds stiffness to the hybrid membrane; thirdly, edge-edge connection between GO and a-BN nanosheets allows

flexible adjustment of nanosheets positioning during membrane formation and thus a more compact structure. The d-spacing results provide direct evidence of the well- ordered nanosheet packing of a-BNGO membranes. It can be expected that such unique interlayer structure would help to achieve exceptional separation properties based on a size exclusion mechanism.

Figure 47. XRD results of (a) pure GO, (b) a-BN0.5GO, (c) a-BN1.0GO, (d) a- BN1.5GO, and (e) a-BN2.0GO.

To better understand the micro-structure and stacking pattern, AFM was used to characterize the surface topography and roughness (Rq) of the a-BNGO membranes,

Figure 48. All the a-BNGO membranes show wrinkles on the surface. The wrinkles became more obvious with the increase of a-BN amount and membrane thickness, since the wrinkles accumulate with continuous stacking of nanosheets. To further study the surface topography of a-BNGO membranes, two lines were selected, and height profiles were analyzed for each membrane. The a-BN0.5GO showed the smallest height

difference of ~300 nm and roughness (Rq) of 43.5±6.2 nm. With increasing a-BN content, the surface roughness of a-BN1.0GO and a-BN1.5GO increased to 72.0±7.4 and 115.3±12.8 nm respectively, and corresponding height difference increased to ~500 and 650 nm. This can be explained by wrinkle accumulation and inclusion of stiff a-

BN nanosheet. For GO-based membranes, it has been reported that wrinkles formed because of accumulation on initial overlapping or folding of GO nanosheets. Small wrinkles grow into larger ones as the nanosheets keep stacking up[166]. On the other hand, ordering packing of pure a-BN membrane is impossible as seen in Figure 48, and thus lead to the larger surface roughness and height difference. Though significant difference in height difference and roughness was observed, d-spacing of all the a-

BNGO membranes was about the same, indicating a negligible correlation between d- spacing and surface roughness. This phenomenon is different from pure GO membranes since earlier study demonstrated that larger wrinkles in GO membranes led to larger d- spacing and thus worse separation performance[166].

Figure 48. AFM morphology images and line profiles of (a) a-BN0.5GO, (b) a- BN1.0GO, (c) a-BN1.5GO, and (d) a-BN2.0GO.

For GO-based membrane, the separation performance is majorly dominated by d-spacing and surface properties. The d-spacing servers as screen that allows

permeation of smaller sized ions or molecules, while filtering out larger ones. In liquid separation, surface charge and hydrophilicity also affect the rejection rate and water flux[176, 177]. For the hybrid a-BNGO membrane, the surface properties are determined by the functional groups on both GO and a-BN nanosheets as well as the covalent bonding in between. Thus, to better understand interaction of membrane and separation liquid, contacted angle was measured on all a-BNGO membranes and the results are summarized in Figure 49. As a control, the contact angle of pure GO membrane without cross-linker was measured as 58.6°. These results suggest that the contact angle neither differ much from each other (<10) nor follow a certain trend as increasing a-BN content. This is not surprising since contact angle can be affected by many factors such as surface energy, roughness, etc. Besides, gradient distribution of heavier component (a-BN) in membrane thickness direction could be possible, which needs to be explored further in a separate study.

Figure 49. Contact angle result of pure GO, a-BN0.25GO, a-BN0.5GO, a-BN1.0GO, a-BN1.5GO, and a-BN2.0GO.

The separation tests were performed to remove MB from water at 300 psi. The pure GO membrane cannot be tested due to the severe swelling. All the a-BNGO membranes survived after a 9-hour separation test in water, which indicated their excellent stability in aqueous media after cross-linking. Figure 50 summarized the rejection rate (%) and water flux (LMH) results of 70 ppm MB solution separation for a-BN0.25GO to a-BN2.0GO membranes. During the 9-hour test, the permeated solutions were collected every 3 hours to study the durability of each sample. To be specific, a-BN0.5GO to a-BN2.0GO membranes all showed excellent MB rejection rates, higher than 99.5% (red dot line), over the 9-hour test period, Figure 50. The high rejection rates implied that the narrow inter-channels in a-BNGO membranes could effectively block MB permeation, while water molecules can pass through freely. It is worth mentioning that a-BN0.25GO membrane showed slight drop of rejection rate

from 99.67% in the first 3-hour to 98.82% in 6-9 hour. One of the possible reasons is that the amount of a-BN was insufficient to cross-link GO nanosheets, thus swelling issue was not completely addressed for long-term durability. Both a-BN1.5GO and a-

BN2.0GO exhibited the most stable performance and the rejection rates kept above ~99.

8% within the testing period. Generally, the thickness of GO-based membrane affects the water flux greatly since thicker membrane provides a longer zig-zag path that restricts the molecule transport. Surprisingly, the water flux increased with increase of the membrane thickness in this work. The thickest a-BN2.0GO membrane showed the largest water fluxes, which were 4.15, 2.83, and 2.42 LMH for first, second, and third

3-hour period respectively, Figure 50. For a-BNGO membranes, the increased thickness was contributed by a-BN nanosheets only since the amount of GO in each membrane was kept the same. Compared to GO, a-BN nanosheets exert weaker dragging force to water molecules since no hydrophilic functional groups presented at in-plane area[171]. Thus, though thicker a-BNGO membrane had a longer diffusion path, the diffusion resistance for water molecules is much smaller. As a result, the overall water flux become larger in the hybrid membrane. A similar pattern was also observed in all the membranes where the water flux decreased continuously with separation time. As separation proceeds, MB molecules were filtered and accumulated surrounding the entry point of diffusion channels, which imposed extra diffusion resistance for water molecules and thus decreased water flux.

3 6 9 3 6 3 3 6 7 100 3 9 6 9 9 6 6 99 9 5 3 98 4 3 6 3 97 3 6 9 3 3 6 6 9 9 2

6 9 9 Water flux (LMH) flux Water

Rejection rate (%) rate Rejection 96 1 95 0 0.25 0.5 1.0 1.5 2.0

a-BN/GO ratio Figure 50. Rejection rate (black) and water flux (blue) of 70 ppm MB separation tests of hybrid a-BNGO membranes. The numbers of 3, 6, and 9 represent tine period of 0- 3 hours, 3-6 hours, and 6-9 hours.

It is well known that besides the d-spacing, surface charge also plays an important role in membrane separation especially when charged molecules are involved.

In this work, separation tests were performed on a-BN1.0GO membrane at different pH conditions to understand the role of charge-charge interaction in MB separation. As shown in Figure 51, a-BN1.0GO membrane exhibited excellent and stable separation performance with rejection rate higher than 99.7%, within pH range of 4.0-10.0. The water flux at three different pH conditions was quite similar, Figure 51. It can be concluded that the size exclusion mechanism dominates the separation in a-BN1.0GO membrane, while electrostatic repulsion or attraction has negligible influence on separation.

8 3 3 3 6 9 7 100 6 9 6 9 6 99 5

98 3 4 3 6 3 97 3 9 6 6

9 9 2 Water flux (LMH) flux Water

Rejection rate (%) rate Rejection 96 1 95 0 pH=4 DI water pH=10 Figure 51. Rejection rate (black) and water flux (blue) of 70 ppm MB separation tests of a-BN1.0GO membrane. The numbers of 3, 6, and 9 represent tine period of 0-3 hours, 3-6 hours, and 6-9 hours.

5.5. Conclusion

To sum up, this work used a two-dimensional a-BN nanosheets as cross-linker to fabricate a-BNGO hybrid membrane that shows high separation efficiency as well as excellent durability in liquid separation. The bulk BN was exfoliated and modified with amino groups to provide bonding sites for following cross-linking with GO nanosheets.

The a-BNGO membranes exhibit compact laminar structure and narrow d-spacing of about 0.8 nm. Inserting a-BN nanosheets in GO membrane not only improves membrane durability by restricted swelling, but also enlarges water flux through reducing resistance in diffusion channels. The a-BN2.0GO membrane acquires the best separation performance with rejection rate of 99.98% and water flux of 4.15 LMH. The outstanding separation performance can be maintained within a wide pH range of 4-10.

Different from conventional studies where organic linkers were widely used, this work offers a promising approach that uses inorganic two-dimensional BN nanosheets as cross-linker in GO membrane. Both physical confinement and chemical bonding can be achieved simultaneously which results in a durable hybrid membrane with enhanced

rejection rate and water flux. Considering the simple fabrication process, separation efficiency and durability, such hybrid membranes have great potential to be practically useful in liquid separation field.

CHAPTER VI

CROSS-LINKED GO MEMBRANES ASSEMBLED WITH GO NANOSHEETS OF

DIFFERENTLY SIZED LATERAL DIMENSIONS FOR ORGANIC DYE AND

CHROMIUM SEPARATION

6.1. Outline

Membrane microstructure and separation performance largely depend on the property of building units and its subsequent patterning. In this work, three differently sized graphene oxide (GO) nanosheets were prepared by a simple centrifugation method. Supported GO membranes were then fabricated based on these separated GO nanosheets. GO nanosheets of different lateral sizes show varied degree of oxidation and distribution of oxygenated groups at both edge and in-plane areas, which leads to different membrane microstructure and separation performance. The swelling and long- term stability of GO membrane were addressed by a pre-crosslinking process with ethylene diamine (EDA). The ultimate separation performance of these GO/EDA membranes was affected by several factors such as cross-linking density, d-spacing, and microstructure like wrinkles all relating to the lateral size of GO nanosheets.

6.2. Introduction

Graphene oxide (GO), a 2-dimensional material, has attracted great attention in membrane research in the past decades due to its intrinsic physicochemical properties, such as the tunable surface functionality, excellent chemical inertness, and thermal stability[41, 43, 61, 91, 126, 178]. With the presence of rich oxygenated groups at both edge and basal plane, GO nanosheets can be easily modified and tuned for different

applications[84, 168, 179]. Besides, these oxygenated groups enable excellent dispersibility of GO nanosheets in water, which further promotes processability of GO sheets into membranes via different methods such as vacuum filtration, spin-coating, and layer-by-layer assembling[72, 92, 109]. GO-based membranes have been demonstrated excellent gas and liquid separation performance due to its compact laminar structure and controllable inter-spacing distance (also named d-spacing)[139,

180]. Unfortunately, free-standing GO membrane still suffers from its poor mechanical strength especially at nanoscale thickness and low integrity in water due to severe swelling[80, 168]. Thus, Porous substrates are often used to strengthen the mechanical property of GO membrane and expand the operational window of practical separation conditions [41, 83, 104]. It is well accepted that d-spacing (the distance between neighboring GO nanosheets) of GO membrane has strong correlation with its separation performance[62, 66, 181]. By stacking GO nanosheets, the laminar structure of GO membrane provides zigzag paths for transporting small molecules. Thus, if d-spacing can be controlled precisely, GO membranes can be designed for separating different targets, especially when the size exclusion mechanism dominates in the separation process[27, 80, 83, 139]. Although d-spacing of the graphene membrane has been reported as 0.34 nm, d-spacing of GO membrane at dry condition would expand to about 0.80 nm due to the steric hindrance of oxygenated functional groups on GO nanosheet[80]. In aqueous environment, the d-spacing could further expand to 6-7 nm due to the existence of rich hydrophilic groups and thus severe swelling[179].

Consequently, the GO membrane completely loses its separation capability since the d- spacing is about ten-fold larger than the separation molecules/ions[136, 137]. Moreover, the severe swell also leads to poor mechanical strength of the GO membrane. Thus,

addressing the swelling issue of GO membrane is critically important to improve the separation performance as well as long-term stability in aqueous separation applications.

For example, cross-linking via covalent bonds has been adopted to restrict swelling of

GO membranes[28, 41, 125, 180]. Besides the bonding chemistry, membrane fabrication technique has been demonstrated as an impactful factor that influences membrane microstructure and thus separation performance. A recent work reported out that a pressure-driven filtration method led to a better compacted and uniform laminar structure in GO membrane than the one fabricated by vacuum filtration method[65].

The difference in GO nanosheet stacking led to varied microstructures, reflecting in the unique pattern of surface wrinkling. From different fabrication processes, like pressure- driven filtration, vacuum filtration, drop-casting, and spraying, wrinkles are always formed[74, 143, 144]. The soft and 2-D GO nanosheets are very likely to fold or overlap at the early stage of stacking. Once the folding or overlapping occurred, wrinkles started to form and became sharper as increasing membrane thickness[74]. So far, it is still not clear how the winkles are structurally different in a membrane and how the separation performance would be affected by them.

The property of building units (GO nanosheet) such as the lateral size and density/distribution of functional groups could dramatically impact the membrane microstructure and separation[182]. Simply speaking, the larger lateral size of GO nanosheets tend to form longer transportation channels in GO membrane, which is beneficial to improve the separation performance. However, more likely, larger sized

GO sheets may form more surface wrinkles which are not considered advantages of an effective membrane. Besides, due to the size difference in lateral dimension, the species and density of oxygenated functional groups on GO nanosheets could be different.

Therefore, a unique patterning behavior of GO membranes with differently sized GO nanosheets could be expected and microstructure-separation property relationship can be studied and established.

In this work, GO nanosheets with three different sizes were separated by a simple centrifugation method. By cross-linking with ethylenediamine (EDA) and filtrating on polydopamine (PDA) coated porous substrate, the issues of weak mechanical strength and swelling in water were addressed simultaneously. The thickness of each membrane can be easily controlled by the amount of filtration solution.

The size selection, cross-linking reaction, microstructure, and wrinkle structure of GO membrane were systematically characterized. The prepared membranes were tested in various liquid separation applications including methylene blue separation, methyl orange separation, and chromium removal from water. Each type of separation was proceeded under three different pH conditions (pH=4, 7 and 10).

6.3. Materials and Methods

6.3.1. Materials

Potassium permanganate (KMnO4, ≥99.0%), phosphorus pentoxide (P2O5,

≥98.0%), sulfuric acid (H2SO4, 95.0-98.0%), hydrogen peroxide (H2O2, 30 wt% in

H2O), dopamine hydrochloride, Trizma base (≥ 99.9%), methyl orange (MO), phosphoric acid (≥ 85.0 wt%), ethylenediamine (EDA, ≥ 99.0%), were all purchased from Sigma Aldrich. Graphite powder (SP-1) was purchased from Bay Carbon Inc,

USA. Hydrochloric acid (HCl) was purchased from EMD Millipore Corporation.

Potassium persulfate (K2S2O8, ≥ 99.0%), methylene blue (MB, C16H18ClN3S·3H2O),

1,5-diphenylcarbazide (DPC, C13H14N4O) were purchased from Fisher Scientific.

Potassium dichromate (K2Cr2O7, 99.0%) was purchased from Alfa Aesar. The

polyethersulfone (PES, 0.1 µm pore size) membrane filters were purchased from

Sterlitech Corporation. The dialysis membrane (Spectra/Por 4, Molecular weight cut off: 12–14 kD) was purchased from Spectrum Laboratories, Inc.

6.3.2. Preparation of GO with different lateral sizes

The GO was synthesized by a modified Hummer method. The experimental details refer to our previous work[107]. The final concentration of GO aqueous solution was determined to be 12.7 g/L, then it was diluted to 0.4 g/L. Centrifugation was used to separate GO nanosheets of different lateral sizes. Specifically, 60 mL of diluted GO solution was centrifugated at 500 rpm for 10 mins, and large GO nanosheets (GO-L) were settled at the bottom of the vial, Figure 52(a). Then, the supernatant solution was transferred to another vial with another cycle of centrifugation at 2000 rpm for 20 mins.

After this step, medium sized GO nanosheets (GO-M) can be collected at the bottom of the vial, Figure 52(b). After transferring the supernatant solution to another vial, a final cycle of centrifugation was applied at 5000 rpm for 30 mins and small sized GO nanosheets (GO-S) were collected, Figure 52(c). After each cycle, the supernatant solution becomes lighter colored, indicating the continuous separation of GO sheets from the solution. After collecting the separated solids at the bottom of separation vials, each sample was diluted to 20 mL with DI water and uniform suspension can be obtained, Figure 52(d-f).

100

Figure 52. The digital images of GO solutions after centrifugation (a) 500 rpm for 10 mins, (b) 2000 rpm for 20 mins, and (c) 5000 rpm for 30mins. And correspond solutions after dispersing in DI water correspond (d-f).

6.3.3. Membrane fabrication

EDA was used as cross-linker to strengthen the GO-GO interaction. Take GO-

L solution as an example, 0.2 g EDA was firstly added into 20 mL GO-L solution and the mixture was heated at 85 °C for 2 hours with vigorous stirring. It was observed that the color of GO solution turned from light brown to black, implying a partial reduction of GO nanosheets to graphene. Afterwards, the mixture was cooled down to room temperature and kept for following membrane fabrication process. The same crosslinking process was applied to both GO-M and GO-S solutions.

The preparation of GO/EDA membrane is illustrated in Figure 53. Realizing the weak mechanical strength of pure GO membrane, the pre-cross-linked GO/EDA solutions were filtrated on polydopamine (PDA) coated PES substrates (PDA/PES).

Specifically, PES substrate (Figure 53a) with 2-inch diameter was soaked in 50 mL aqueous solution containing 0.1 g dopamine and 0.06 g tris base. The PDA polymerization was processed in a shaker for 24 hours at room temperature. After polymerization, the PDA/PES membrane was dried at 80 °C for 1 hour (Figure 53b) before it was assembled in a separation cell, then the prepared GO/EDA solution was placed in the cell and completely sealed. After that, 300 psi pressure was applied to push out the liquid and left over a solid GO/EDA membrane on top of PDA/PES

101

substrate, Figure 53(c). To complete the cross-linking of GO membrane and strengthen the GO/substrate interfacial bonding, supported membranes were further heated at

80 °C for 2 hours in an oven. The final membranes were named GO/EDA-S, GO/EDA-

M and GO/EDA-L, respectively.

6.3.4. Characterization

The cross-section morphology of membrane was characterized by scanning electron microscope (SEM, JEOL-7404, 5kV) with sputter-coated platinum layer on top of the sample surface. The d-spacing of the supported membranes was characterized by X-ray diffraction (XRD, Bruker AXS D8 Discover diffractometer with General Area

Detector Diffraction System, 40 kV, 35mA) with a scan rate of 1.0 degree/min of 5-15 degree. The d-spacing is calculated by Bragg’s law, Equation (13):

2푑 sin  = 푛 (13) where d is the d-spacing value,  is the angle of incidence/scattering, n is the order of diffraction, (n=1 in this case), and  is the wavelength of X-ray ( =0.154 nm). The

102

elemental composition of all samples was analyzed by X-ray photoelectron spectroscopy (XPS, PHI VersaProbe II Scanning XPS Microprobe with Al K α line excitation source). FT-IR results were characterized by Nicolet iS10 FT-IR

Spectrometer. The surface topography of the membranes was characterized by atomic force microscopy (AFM, Park System XE7).

All the MB, MO and Cr (VI) separation tests follow the same procedure in a dead-end separation cell. The prepared membrane was sealed in the separation cell being supported on a porous metal disk. For MB separation, 250 mL MB solution at the concentration of 70 ppm was added into the cell. The separation was carried out at

300 psi pressure with N2 purging gas. The permeated liquid was continuously collected for 9 hours. The volume of liquid was measured for flux calculation, and UV-vis was used to determine the MB concentration in the permeate liquid. Peak intensity at 663.6 nm wavelength was recorded and analyzed to calculate MB rejection rate. The permeation flux J can be calculated by Equation (14):

푉 퐽 = (14) 퐴×푡 where 푉 (L) is the volume of permeate; 퐴 (m2) is the working membrane area; 푡 (h) is the operation time. And the rejection rate of the dye molecule can be calculated by

Equation (15):

퐼 휂 = 1 − 푖 (15) 퐼0,푖 where 퐼0,푖 and 퐼푖 are the intensity of UV-vis peak at 663.6 nm from the feed and permeate solutions, respectively.

The concentration of the initial MO solution was fixed to 70 ppm as well. The peak intensity of UV-vis was obtained at 463 nm. The concentration of initial Cr (VI) was 1 ppm. Before UV-vis absorbance measurement, color development of Cr (VI) was

103

applied. To be specific, 5 mL of Cr (VI) target solution was placed into the vial, and

0.5 mL of o-phosphoric acid (4.5 M) and 0.25 mL of DPC (5 g/L in acetone) were added into vial respectively. After 30 mins of color development at room temperature, the solutions were collected for UV-vis measurement. Peak intensity at 540 nm was measured to quantify the Cr (VI) concentration.

6.4. Results and Discussion

AFM was used to characterize the lateral size of GO nanosheets. The topography images of GO nanosheets were summarized in Figure 54. Realizing the fact that three major groups of GO nanosheets were divided based on size, size distribution still presents in each group. Therefore, measurements were taken on larger pieces of samples in each group to better represent the difference of GO size across groups. To assure consistency of the results, at least five samples with similar lateral size were captured and one is reported here. The GO-S nanosheet shows a lateral dimension of about 15*20 µm, Figure 54(a). Figure 54(b) clearly exhibits the multi-pieces of medium sized nanosheets, with typical size of 20*40 µm. The brighter curves distributed on top of GO sheets are signature wrinkle structures, which were commonly observed on the surface of GO nanosheets[74, 145]. For the group of large GO nanosheets, no obvious borders were found within the largest scan scale (45*45 µm), Figure 54(c). Thus, the size of large GO nanosheets is expected to be larger than 45*45 µm. More wrinkles were observed in large sized GO nanosheets, since GO nanosheets with larger dimensions are more likely to form wrinkles due to overlapping or folding. The largest length in basal dimension of 15 pieces GO nanosheets of each size were quantitatively analyzed and summarized in Figure 54(d). The largest dimension of GO-S, GO-M, and

GO-L was in the range of 12-26 μm, 28-40 μm, and >45 μm. The thickness of GO

104

nanosheets was also measured by AFM, which is about 9, 13.5 and 42 nm for GO-S,

GO-M and GO-L, respectively. All these results confirmed that GO nanosheets of different dimensions and thicknesses were effectively separated into three main groups.

Figure 55 compared the FT-IR spectrum of pure GO nanosheets of different lateral sizes. The peaks at 1720, 1620, 1362, 1230, 1065, and 3120 cm-1 can be assigned to the C=O, C=C, C-OH, C-O-C, C-O, and O-H, respectively, which are consistent with other literature[183]. The peak positions of three GO nanosheets were identical, indicating a similar oxidation process regardless of the dimensional size. The existence of these oxygen functional groups in GO is responsible for their excellent dispersion in water.

105

Figure 55. The FT-IR spectrum of (a) GO-S, (b) GO-M, and (c) GO-L.

To quantify the different functional groups, high-resolution XPS was used to characterize all three groups of GO nanosheets. The C 1s peak of GO can be deconvoluted into four peaks that are C=C at 284.8 eV, C-O at 286.3 eV, C-O-C at

287.0 eV and C=O at 288.4 eV[183], Figure 56. One apparent difference among these samples is that the GO-S shows strong C=C peak (Figure 56a), while GO-M and GO-

L exhibit additional C-O-C, Figure 56(b&c). These results indicate that larger GO sheets are more likely to be oxidized into C-O-C epoxide ring, which is mostly distributed at the basal plane area per earlier reports[56]. Also, comparing to GO-S, the

GO-M and GO-L exhibited larger C-O and C=O peaks representing the carboxyl groups.

The carboxyl groups are commonly served as reactive sites for cross-linking reaction.

Thus, the more carboxyl groups, the higher cross-linking density in the following membrane fabrication process. Also, part of epoxy rings on basal plane of GO could also be reacted with amine groups, which offered another type of reactive sites for

106

cross-linking. Here, to quantify the oxidation degree, the percentage of oxygenated carbon (oxy-C%) is calculated by the following Equation (16), where A stands for the area of each deconvoluted peak.

퐴 표푥푦 − 퐶% = (1 − 퐶=퐶 ) ∗ 100% (16) 퐴퐶=퐶+퐴퐶−푂+퐴퐶−푂−퐶+퐴퐶=푂

The calculated oxy-C% of GO-S, GO-M, and GO-L are 33.6%, 63.9%, and 59.0% respectively. The larger oxy-C% implies a higher degree of oxidation of GO nanosheets.

The presence of oxygen-containing groups on GO nanosheets creates steric hindrance during membrane formation, which poses a potential challenge to tune down the d- spacing. Besides, higher oxy-C% leads to a more hydrophilic membrane, thus membrane stability in aqueous separation could be an issue. However, these groups are considered reactive sites for cross-linking reaction. The balance between the oxygen- containing group density and cross-linking density seems critical to control the membrane separation property and long-term stability. Especially, the GO nanosheets featuring different lateral sizes and oxy-C% in this work present a unique control on the property of membrane building blocks towards high performance GO membranes.

Figure 56. High resolution XPS C 1s analysis of pristine GO nanosheets. (a) GO-S, (b) GO-M, and (c) GO-L. The four deconvoluted peaks are C=O, C-O-C, C-O, and C=C.

To confirm the cross-linking reaction after adding EDA cross-linker, the elemental composition of GO-S, M, L and GO/EDA-S, M, L were comparatively analyzed, Figure

57. A new N 1s peak appeared at 398.0 eV in all three cross-linked membranes,

107

indicating the presence of EDA in the membranes. The atomic percentage of C, O and

N elements was listed in the figure. Without cross-linking, GO-S exhibited the lowest oxygen composition of 23.4%, Figure 57(a). Relatively higher O% of 33.2 and 32.0% was observed in GO-M and GO-L respectively, Figure 57(b&c). After cross-linking, the O% of GO/EDA-S dropped down to 21.2%, while a more significant drop to 26.9 and 24.2% was observed in GO/EDA-M and GO/EDA-L. This can be also attributed to the successful cross-linking reaction, that could occur between -NH2 and -COOH as well as -NH2 and epoxy rings. Secondly, the GO/EDA-S, GO/EDA-M, and GO/EDA-

L showed gradually increased atomic N% of 8.9, 9.9, and 13.8%, respectively. The higher N% means more reacted EDA in the GO membrane. The highest N% observed in GO/EDA-L signifies the highest cross-linking density in the GO/EDA-L membrane.

As revealed in Figure 56(c), GO-L exhibits the largest proportion of C-O in C 1s, which correlates to the highest density of -COOH groups and thus cross-linking density. It is worth noticing that GO/EDA-S has a slightly lower N% of 8.9%, although O% of GO-

S is significantly lower compared to the other two groups of GO. Such phenomena can be caused by distribution of oxygenated groups, more likely a higher density of reactive edge -COOH groups and lower density of non-reactive in-plane C-O-C groups, which is consistent with the analysis in Figure 56.

108

To further study how functional groups affected cross-linking, the high- resolution XPS N 1s was also processed to all three membranes. The N 1s peaks can be deconvoluted into 4 peaks that are N-H at 398.3 eV, C-N at 399.4 eV, N-C=O at 400.5 eV, and NH3+-C at 401.4 eV[184], Figure 58. Firstly, the peak of N-C=O refers to the amide bond between -COOH and -NH2 after the reaction, that is directly related to the cross-linking process with EDA. The GO/EDA-L had the highest cross-linking density due to the largest N-C(O) peak, Figure 58(c), while the GO/EDA-S and GO/EDA-M shared the similar peak area, Figure 58(a&b). To quantify the extent of cross-linking reaction, the percentage of N in amide groups (amide-N%) was calculated by following

Equation (17), where A stands for the area of each deconvoluted peak.

퐴 푎푚푖푑푒 − 푁% = ( 푁−퐶(푂) ) ∗ 100% (17) 퐴퐶−푁퐻 +퐴푁−퐶(푂)+퐴 + 2 퐶−푁퐻3

109

The calculated amide-N% of GO/EDA-S, GO/EDA-M, and GO/EDA-L were 25.9%,

30.4%, and 34.6%, respectively. The higher cross-linking density of GO/EDA-L can be attributed to the existence of more reactive carboxyl groups. Secondly, the strong

H2N-C peak appeared to all three samples, which indicated the existence of unreacted dangling -NH2 groups due to the excess amount of EDA added during membrane fabrication.

Figure 58. High resolution XPS N 1s analysis of (a) GO/EDA-S, (b) GO/EDA-M, and 3+ (c) GO/EDA-L. The four deconvoluted peaks are NH -C, N-C(O), H2N-C, and N-H.

The preparation of GO/EDA membranes was carried out in a dead-end filtration system and the cross-section images of each membrane were captured by SEM, Figure

59. All three membranes show well-oriented and compact laminar structure and uniform thickness. The thickness of GO/EDA-S, GO/EDA-M, and GO/EDA-L was about the same which falls in the range of 450-600 nm.

Figure 59. Cross section SEM images of (a) GO/EDA-S, (b) GO/EDA-M, and (c) GO/EDA-L. Thickness was controlled to be about 500 nm by adjusting the amount of GO/EDA solution.

It is well known that the stacking modes of GO nanosheets could significantly influence membrane microstructure and separation ability[179]. In this work, AFM was

110

used to study the surface topography of the GO/EDA membranes prepared by differently sized GO nanosheets, Figure 60. All three membranes showed obvious wrinkles on the surface (bright area) with unique distribution patterns and height profiles. To quantitatively compare the membrane surface structure, surface roughness

(Rq) was introduced which can be used to index the surface uniformity of each membrane. The Rq values of GO/EDA-S, GO/EDA-M, and GO/EDA-L were 115.0,

133.0, and 138.5 nm, respectively, indicating that smaller GO nanosheets tend to form a more uniform membrane probably due to the reduced wrinkling during the membrane fabrication process. It was reported that GO nanosheets with smaller dimensions have less chance to generate initial wrinkles formed by edge overlapping or self-folding

[145]. Two line profiles across the brightest area were analyzed for each membrane.

The largest height difference values were also measured and marked in Figure

60(a1&a2, b1&b2, c1&c2). Relatively smaller height difference of ~620 nm was observed in GO/EDA-S membrane, while GO/EDA-M and GO/EDA-L showed their height difference in a higher range of ~1400 and ~970 nm, respectively. It is worth noticing that GO/EDA-M and GO/EDA-L show similar Rq value, but large difference in height profile. Such difference in membrane surface morphology could be attributed to the unique stacking behavior of GO sheets with different lateral sizes.

111

Figure 61 summarized the XRD results of three GO/EDA membranes that characterized at both dry and wet conditions at varied pH environments. At dry

112

condition, the d-spacing of three membranes were calculated as 0.99, 0.93, and 0.95 nm, respectively, Figure 61(a). These results are slightly larger than d-spacing of GO membrane reported in the literature[180], probably due to the increased steric hindrance after inserting EDA cross-linker in GO membranes. Similar testing was carried out on wet membranes, which were soaked in DI water, HCl solution (pH=4) and KOH solution (pH=10) for 24 hours before measurement. It is obvious that the d-spacing of

GO membranes was enlarged to ~1.2 nm after soaking in DI water and KOH solution,

Figure 61(b&d). However, the d-spacing of GO/EDA-S and GO/EDA-M membranes were unexpectedly reduced to 0.87 and 0.85 nm after soaking in HCl solution, Figure

61(c). Though the mechanism of reduced d-spacing in acidic environment is still not clear, it is believed that the presence of proton in the membrane could offer additional sheet-sheet interactions such as hydrogen bonding and result in reduced d-spacing[185].

In GO/EDA-L, the large density of cross-linking bonds dominated the inter-sheet interaction, and thus the d-spacing was not affected in acidic environment. It was noticed that the d-spacing of the membranes all increased to ~1.2 nm in wet conditions.

As membrane exposed to liquid media, swelling occurred that increased the d-spacing, meanwhile, the covalent bonding by EDA restricted the expansion of d-spacing. In theory, the d-spacing should match the fully stretched molecular length of the EDA chain, which is ~1.25 nm as seen in Figure 62. In a word, the swelling issue is greatly restricted in all three tested membranes since the d-spacing of pristine GO membrane could expand to 6-7 nm in water[80].

113

Figure 61. XRD results of three GO/EDA membranes (a) at dry state, and after 24 hours soaking in (b) pH=7, (c) HCl solution, pH=4, and (d) KOH solution, pH=10. The corresponding d-spacing values were marked in plot, which was calculated by Bragg’s Law.

114

Figure 62. Scheme of GO/EDA membrane with (a) compacted laminar structure at dry state and (b) enlarged d-spacing at aqueous environment. (c) Scheme of length of EDA.

The GO/EDA membranes were firstly tested to remove the positively charged

MB dye (70 ppm) from water under different pH conditions at 300 psi pressure. The

GO membrane without crosslinking was tested as a control sample, severe swelling and membrane failure occurred within 30 minutes which is consistent with other reports[80,

107]. All the GO/EDA membranes survived the 9-hour separation test except the

GO/EDA-S membrane under pH=10. Figure 63 summarized the rejection rate (%) and water flux results of GO/EDA membranes under different pH conditions. The corresponding digital images of the MB solution before and after separation were also provided in Figure 64&65. The permeated solutions were collected every 3 hours during the 9-hour separation test to calculate the flux and evaluate the membrane durability. In general, all three membranes showed outstanding rejection rate of higher than 96.0% in DI water, Figure 63(a). Specifically, the rejection rate reached up to 99.98%

115

for GO/EDA-M membrane, which kept stable within the entire testing period. However, the gradual decrease of rejection rate with separation time was observed in GO/EDA-S and GO/EDA-L membranes. To evaluate the membrane stability in different pH environments, additional tests were performed at pH=4 and pH=10. At pH=4, the rejection rate was still higher than 96.0% for all three membranes, Figure 63(b). The

GO/EDA-S membrane showed the most stable rejection rate probably due to the reduced d-spacing that was revealed by XRD results, Figure 61. When separation is performed under pH=10, the GO/EDA-S lost its separation function and rejection rate was only 36.84% during the first 3-hour period. The rejection rates of second and third

3-hour periods were only 28.65% and 23.63%. The water fluxes were all slightly higher than GO/EDA-M and GO/EDA-L, indicating the integrity of the GO/EDA-S membrane while the reason for the poor separation is still under exploration. For GO/EDA-L membrane, rejection rate as high as 99.0% was achieved and maintained in the entire testing period. Under such basic environment, MB attraction in the membrane is expected and thus reduced water flow was observed, Figure 63(c).

116

Figure 64. Digital images of MB solutions before and after (3, 6, and 9 hours) separation tests under (a) pH=7, (b) pH=4, and (c) pH=10 conditions.

117

Figure 65. Digital image of MB solution before (70 ppm) and after (6 h and 9 h) separation test.

To evaluate the long-term stability of the GO/EDA-S, GO/EDA-M, and

GO/EDA-L membranes, each membrane was tested for 24 hours based on its best performance among three pH conditions. The rejection rate and water flux were summarized in Figure 66. All three membranes showed excellent stability during 24 hours testing period with stable rejection rates of > 99.0% (GO/EDA-S&M) and > 98.0%

(GO/EDA-L).

Based on the satisfactory separation of positively charged MB dye, separation on a negatively charged MO dye (70 ppm) was also tested in this work. The rejection rate and water flux were summarized in Figure 67, and their digital images were exhibited in Figure 68. MO separation in DI water showed a rejection rate of higher than 96.0% for all three membranes even after the 9-hour testing period, while water flux dropped from 3 to 2 LMH due to the narrowed diffusion channel by MO

118

accumulation, Figure 67(a). At pH=4, the testing started with a low rejection rate and then ramped up gradually to 96.0, 94.7, and 99.0% at the end of test for GO/EDA-S,

GO/EDA-M and GO/EDA-L, respectively. Such improved rejection rates with time can be attributed to the gradually narrowed d-spacing due to the MO accumulation, which was evidenced by the decreased water flux with increasing separation time, Figure

67(b). The GO/EDA-L exhibited the best performance when pH=4, but its MB separation performance was the worst at the same condition, Figure 67(b). Such dramatic difference of separation can be attributed to the opposite charges of MB and

MO. The MO separation was relatively poor in basic environment (pH=10) while water flux was obviously larger especially for GO/EDA-M and GO/EDA-L membranes,

Figure 67(c). The highest rejection rate of 93.3% was observed in the GO/EDA-M membrane. With the existence of a large number of negative hydroxide ions and MO molecules, the repulsion through the membrane tends to enlarge the d-spacing, and thus sacrifices the separation performance.

119

Figure 68. Digital images of MO solutions before and after (3, 6, and 9 hours) separation tests under (a) pH=7, (b) pH=4, and (c) pH=10 conditions.

More data on MB and MO separation with GO-based membranes were summarized in Table 11. The literature data indicated that GO-based membranes showed good separation for positively charged MB molecules, our work further

120

revealed that size of GO nanosheets could affect the functional groups, the stacking mode, micro-structure, and thus varied MB separation performances were observed.

Compared to MB, the efficiency of MO removal was reported to be relatively lower. In a word, the efficiency of MB and MO removal in this work is top-rated among the listed results. The concentration (70 ppm) of original dye solution in this work is relatively higher than most of reported data, indicating a good potential of GO/EDA membrane for separating concentrated dye solutions.

Table 11. Comparison of organic dye separation efficiency with GO based membranes in this work and literatures. Membrane name Method Target Efficiency Ref. (Concentration) GO/MXene Filtration MB (10 ppm) 99.5% [186] GOQD NF Dead-end filtration MO (50 ppm) 97% [187] MB (50 ppm) 71% HP-COF-TpPa/GO Dead-end filtration MB (10 ppm) 97.05% [188] BPEI/GO_TU Filtration MB (10 ppm) ~99% [189] MO (10 ppm) 99.6% GO/APTF Dead-end filtration MB (7.5 ppm) 99.7% [190] GO-PMMAhyd Dead-end filtration MB (50 ppm) ~58% [191] MO (50 ppm) ~87% 44-GO-0.5BA-T Dead-end filtration MO (10 ppm) ~82% [192] GO/NH2-Fe3O4-8 Cross-flow MB (100 ppm) 70.0% [193] circulation MO (100 ppm) 75.0% G-CNTms Dead-end filtration MO (50 ppm) >96% [194] GO/EDA-M Dead-end filtration MB (70 PPM) 99.98% This GO/EDA-L Dead-end filtration MO (70 ppm) 99.0% work

Besides the oppositely changed organic dye molecules, inorganic heavy metal ion Cr (VI) separation was also performed in this work. At neutral condition, the separation performance was rather poor, as can be seen from the very low rejection rates ranging from 35.2 to 62.6%, Figure 69(a). Considering the much smaller hydrated

2- diameter of CrO4 ion (0.45 nm) compared to the d-spacing of 1.19-1.22 nm for all

121

three membranes, it is not surprising to observe the poor separation performance[195].

The slight decrease of water flux with separation time can be attributed to the ion accumulation in the diffusion channel of membrane. Surprisingly high rejection rates were observed under the condition of pH=4. The GO/EDA-S and GO-EDA-M reached almost 100% rejection in 2nd and 3rd 3-hour periods, while GO/EDA-L also achieved higher than 99.0% in the 2nd period, Figure 69(b). The significantly improved rejection could be attributed to the narrowed d-spacing of GO/EDA-S and GO/EDA-M

2− membranes, as well as the enlarged ionic size of Cr2O7 (0.60 nm) in acidic condition[195], as seen following.

2− + 2− 2퐶푟푂4 + 2퐻 ↔ 퐶푟2푂7 + 퐻2푂

When the pH was tuned to 10, the overall rejection rates reduced to 85%, Figure 69(c).

2− At pH=10, the balance moves to small sized CrO4 ion, together with the enlarged d- spacing leads to poor separation. All the digital images of colored Cr (VI) solutions, both before and after separation, were provided in Figure 70. Overall, the GO/EDA-S exhibited the best performance under each condition.

122

Figure 69. Rejection rate (black) and water flux (blue) of Cr (VI) removal in dead-end filtration system at room temperature under different pH conditions. (a) pH=7, (b) pH=4, and (c) pH=10. The pH was adjusted by HCl and KOH. Initial concentration of Cr (VI) solution is 1 ppm. Digital images of colored Cr (VI) solutions were shown in Figure 70.

Figure 70. Digital images of colored Cr (VI) solutions before and after (3, 6, and 9 hours) separation tests under (a) pH=7, (b) pH=4, and (c) pH=10 conditions.

123

In this work, the GO/EDA-S membrane showed the smallest wrinkles and the lowest cross-linking density; while GO/EDA-L showed the largest wrinkles and the highest cross-linking density, as evidenced by XPS and AFM results. Though simulation study indicated that membrane fabricated by large GO nanosheets had better separation performance because of longer diffusion path for molecular transportation[182], the practical separation performance was also affected by wrinkling, swelling, and charge of membranes. Therefore, the lateral size of GO nanosheets not only affect the d-spacing and length of diffusion path through the membrane, the stacking of GO sheet and its corresponding surface morphology, microstructure and surface charge are all related to determine the membrane separation properties.

6.5. Conclusion

To sum up, this work investigated the effect of GO nanosheet lateral size on microstructure and liquid separation performance of EDA cross-linked membranes.

Three groups of GO nanosheets with different lateral sizes were separated by centrifugation, which were then used for membrane fabrication with the involvement of cross-linker EDA. The differently sized GO nanosheets form unique stacking patterns resulting in varied separation properties, especially in different pH environments. All the membranes showed excellent long-term (24 hours) stability in neutral (pH=7), acidic (pH=4) and basic (pH=10) conditions. MB separation performance has been found closely related to GO lateral size at specific pH environments. Specifically, the highest rejection rate in GO/EDA-S, GO/EDA-M, and

GO/EDA-L membranes has been found at pH=4, 7 and 10 conditions, respectively. For

MO, the separation is less effective relatively as compared to MB, while rejection rate

124

of > 96.0% can be still achieved on all three membranes. Rejection rate of > 99.0% for

Cr (VI) was accomplished on GO/EDA-S and GO/EDA-M membranes taking advantage of narrowed d-spacing and enlarged ion size in acidic environment. This work presents the relationship between differently sized GO nanosheets and membrane microstructure in relating to the membrane performance in liquid separation applications. Considering the simple method of separating different size GO nanosheets, simple membrane fabrication method, and excellent separation performance on MB,

MO and chromium ions, this work can be used to guide the selection of GO nanosheets for membrane fabrication towards various separating targets.

125

REFERNECES

[1] http://www.who.int/mediacentre/factsheets/fs391/en/.

[2] A.D. Khawaji, I.K. Kutubkhanah, J.M. Wie, Advances in Seawater Desalination Technologies, Desalination, 221 (2008) 47-69.

[3] M. Al-Shammiri, M. Safar, Multi-Effect Distillation Plants: State of The Art, Desalination, 126 (1999) 45-59.

[4] A.R. Glueck, Desalination by An Ion Exchange-Precipitation-Complex Process, Desalination, 4 (1968) 32-37.

[5] X.F. Sun, J. Qin, P.F. Xia, B.B. Guo, C.M. Yang, C. Song, S.G. Wang, Graphene Oxide-Silver Nanoparticle Membrane for Biofouling Control and Water Purification, Chem. Eng. J., 281 (2015) 53-59.

[6] B. Feng, K. Xu, A.S. Huang, Covalent Synthesis of Three-Dimensional Graphene Oxide Framework (Gof) Membrane for Seawater Desalination, Desalination, 394 (2016) 123-130.

[7] M.S. Rahman, M. Ahmed, X.D. Chen, Freezing-melting Process and Desalination: I. Review of the State-of-the-art, Sep. Purif. Rev., 35 (2006) 59-96.

[8] P.M. Williams, M. Ahmad, B.S. Connolly, Freeze Desalination: An Assessment of An Ice Maker Machine for Desalting Brines, Desalination, 308 (2013) 219-224.

[9] H.M. Qiblawey, F. Banat, Solar Thermal Desalination Technologies, Desalination, 220 (2008) 633-644.

[10] C.N. Li, Y. Goswami, E. Stefanakos, Solar Assisted Sea Water Desalination: A Review, Renew. Sust. Energ. Rev., 19 (2013) 136-163.

[11] C.S. Zhao, S.Q. Nie, M. Tang, S.D. Sun, Polymeric PH-Sensitive Membranes-A Review, Prog. Polym. Sci., 36 (2011) 1499-1520.

[12] M. Ulbricht, Advanced Functional Polymer Membranes, Polymer, 47 (2006) 2217- 2262.

126

[13] P. Shao, R.Y.M. Huang, Polymeric Membrane Pervaporation, J. Membr. Sci., 287 (2007) 162-179.

[14] C. Laberty-Robert, K. Valle, F. Pereira, C. Sanchez, Design and Properties of Functional Hybrid Organic-Inorganic Membranes for Fuel Cells, Chem. Soc. Rev., 40 (2011) 961-1005.

[15] Y. Yampolskii, Polymeric Gas Separation Membranes, Macromolecules, 45 (2012) 3298-3311.

[16] R.M. de Vos, H. Verweij, High-Selectivity, High-Flux Silica Membranes for Gas Separation, Science, 279 (1998) 1710-1711.

[17] H. Verweij, Inorganic Membranes, Curr. Opin. Chem. Eng., 1 (2012) 156-162.

[18] L. Cot, A. Ayral, J. Durand, C. Guizard, N. Hovnanian, A. Julbe, A. Larbot, Inorganic Membranes and Solid State Sciences, Solid State Sci., 2 (2000) 313-334.

[19] Z.Y. Yeo, T.L. Chew, P.W. Zhu, A.R. Mohamed, S.P. Chai, Synthesis and Performance of Microporous Inorganic Membranes for CO2 Separation: A Review, J. Porous Mater., 20 (2013) 1457-1475.

[20] A. Iulianelli, A. Basile, Hydrogen Production from Ethanol Via Inorganic Membrane Reactors Technology: A Review, Catal. Sci. Technol., 1 (2011) 366-379.

[21] F. Zhang, W.B. Zhang, Z. Shi, D. Wang, J. Jin, L. Jiang, Nanowire-Haired Inorganic Membranes with Superhydrophilicity And Underwater Ultralow Adhesive Superoleophobicity for High-Efficiency Oil/Water Separation, Adv. Mater., 25 (2013) 4192-4198.

[22] Z. Wang, C.Q. Hong, X.H. Zhang, X. Sun, J.C. Han, Microstructure and Thermal Shock Behavior of Zrb2-Sic-Graphite Composite, Mater. Chem. Phys., 113 (2009) 338- 341.

[23] K.C. Kim, S.G. Kang, D.S. Sholl, Predictions of Sulfur Resistance in Metal Membranes for H-2 Purification Using First-Principles Calculations, Ind. Eng. Chem. Res., 51 (2012) 301-309.

[24] C.R. Reid, K.M. Thomas, Adsorption of Gases on A Carbon Molecular Sieve Used for Air Separation: Linear Adsorptives as Probes for Kinetic Selectivity, Langmuir, 15

127

(1999) 3206-3218.

[25] C.A. Grande, S. Cavenati, F.A. Da Silva, A.E. Rodrigues, Carbon Molecular Sieves for Hydrocarbon Separations by Adsorption, Ind. Eng. Chem. Res., 44 (2005) 7218- 7227.

[26] M. Rungta, G.B. Wenz, C. Zhang, L.R. Xu, W.L. Qiu, J.S. Adams, W.J. Koros, Carbon Molecular Sieve Structure Development and Membrane Performance Relationships, Carbon, 115 (2017) 237-248.

[27] B. Daniel, F. Karel, P. Kryštof, V. Ondřej, L. Marek, F. Kristián, P. Martin, S. David, L. Jan, S. Zdeněk, Thin, High‐Flux, Self‐Standing, Graphene Oxide Membranes for Efficient Hydrogen Separation from Gas Mixtures, Chem. Eur. J., 23 (2017) 11416- 11422.

[28] Z.Q. Jia, W.X. Shi, Tailoring Permeation Channels of Graphene Oxide Membranes for Precise Ion Separation, Carbon, 101 (2016) 290-295.

[29] A.B. Fuertes, T.A. Centeno, Preparation of Supported Carbon Molecular Sieve Membranes, Carbon, 37 (1999) 679-684.

[30] Q. Yang, Y. Su, C. Chi, C.T. Cherian, K. Huang, V.G. Kravets, F.C. Wang, J.C. Zhang, A. Pratt, A.N. Grigorenko, F. Guinea, A.K. Geim, R.R. Nair, Ultrathin Graphene-Based Membrane with Precise Molecular Sieving and Ultrafast Solvent Permeation, Nat. Mater., 16 (2017) 1198-1202.

[31] A. Morelos-Gomez, R. Cruz-Silva, H. Muramatsu, J. Ortiz-Medina, T. Araki, T. Fukuyo, S. Tejima, K. Takeuchi, T. Hayashi, M. Terrones, M. Endo, Effective NaCl And Dye Rejection of Hybrid Graphene Oxide/Graphene Layered Membranes, Nat. Nanotechnol., 12 (2017) 1083-1088.

[32] S. Homaeigohar, M. Elbahri, Graphene Membranes for Water Desalination, NPG Asia Mater., 9 (2017) e427.

[33] R.R. Nair, H.A. Wu, P.N. Jayaram, I.V. Grigorieva, A.K. Geim, Unimpeded Permeation of Water Through Helium-Leak-Tight Graphene-Based Membranes, Science, 335 (2012) 442-444.

[34] S. Gahlot, P.P. Sharma, H. Gupta, V. Kulshrestha, P.K. Jha, Preparation of

128

Graphene Oxide Nano-Composite Ion-Exchange Membranes for Desalination Application, RSC Adv., 4 (2014) 24662-24670.

[35] L.J. Yang, B.B. Tang, P.Y. Wu, UF Membrane With Highly Improved Flux by Hydrophilic Network Between Graphene Oxide and Brominated Poly(2,6-Dimethyl- 1,4-Phenylene Oxide), J. Mater. Chem. A, 2 (2014) 18562-18573.

[36] S. Zinadini, A.A. Zinatizadeh, M. Rahimi, V. Vatanpour, H. Zangeneh, Preparation of A Novel Antifouling Mixed Matrix Pes Membrane by Embedding Graphene Oxide Nanoplates, J. Membr. Sci., 453 (2014) 292-301.

[37] T. Hwang, J.S. Oh, W. Yim, J.D. Nam, C. Bae, H.I. Kim, K.J. Kim, Ultrafiltration Using Graphene Oxide Surface-Embedded Polysulfone Membranes, Sep. Purif. Technol., 166 (2016) 41-47.

[38] D. Kim, A. Coskun, Graphene Oxide-Templated Preferential Growth of Continuous Mof Thin Films, Crystengcomm, 18 (2016) 4013-4017.

[39] P. Sheath, M. Majumder, Flux Accentuation and Improved Rejection in Graphene- Based Filtration Membranes Produced by Capillary-Force-Assisted Self-Assembly, Philos. T. R. Soc. A, 374 (2016) 2060.

[40] X.J. Song, Q.Z. Zhou, T. Zhang, H.B. Xu, Z.N. Wang, Pressure-Assisted Preparation of Graphene Oxide Quantum Dot-Incorporated Reverse Osmosis Membranes: Antifouling and Chlorine Resistance Potentials, J. Mater. Chem. A, 4 (2016) 16896-16905.

[41] Y.P. Ying, D.H. Liu, J. Ma, M.M. Tong, W.X. Zhang, H.L. Huang, Q.Y. Yang, C.L. Zhong, A Go-Assisted Method for The Preparation of Ultrathin Covalent Organic Framework Membranes for Gas Separation, J. Mater. Chem. A, 4 (2016) 13444-13449.

[42] W. Choi, J. Choi, J. Bang, J.H. Lee, Layer-by-Layer Assembly of Graphene Oxide Nanosheets on Polyamide Membranes for Durable Reverse-Osmosis Applications, ACS Appl. Mater. Interfaces, 5 (2013) 12510-12519.

[43] M. Hu, B.X. Mi, Enabling Graphene Oxide Nanosheets as Water Separation Membranes, Environ. Sci. Technol., 47 (2013) 3715-3723.

[44] N.X. Wang, S.L. Ji, G.J. Zhang, J. Li, L. Wang, Self-Assembly of Graphene Oxide

129

and Polyelectrolyte Complex Nanohybrid Membranes for Nanofiltration And Pervaporation, Chem. Eng. J., 213 (2012) 318-329.

[45] M. Hu, B.X. Mi, Layer-By-Layer Assembly of Graphene Oxide Membranes via Electrostatic Interaction, J. Membr. Sci., 469 (2014) 80-87.

[46] K.L. Goh, L. Setiawan, L. Wei, R.M. Si, A.G. Fane, R. Wang, Y. Chen, Graphene Oxide as Effective Selective Barriers on A Hollow Fiber Membrane for Water Treatment Process, J. Membr. Sci., 474 (2015) 244-253.

[47] J.L. Han, X. Xia, Y. Tao, H. Yun, Y.N. Hou, C.W. Zhao, Q. Luo, H.Y. Cheng, A.J. Wang, Shielding Membrane Surface Carboxyl Groups by Covalent-Binding Graphene Oxide to Improve Anti-Fouling Property and The Simultaneous Promotion of Flux, Water Res., 102 (2016) 619-628.

[48] Q. Nan, P. Li, B. Cao, Fabrication of Positively Charged Nanofiltration Membrane Via the Layer-By-Layer Assembly of Graphene Oxide and Polyethylenimine for Desalination, Appl. Surf. Sci., 387 (2016) 521-528.

[49] L. Wang, N.X. Wang, J. Li, J.W. Li, W.W. Bian, S.L. Ji, Layer-By-Layer Self- Assembly of Polycation/Go Nanofiltration Membrane with Enhanced Stability and Fouling Resistance, Sep. Purif. Technol., 160 (2016) 123-131.

[50] Y. Zhang, S. Zhang, J. Gao, T.S. Chung, Layer-By-Layer Construction of Graphene Oxide (Go) Framework Composite Membranes for Highly Efficient Heavy Metal Removal, J. Membr. Sci., 515 (2016) 230-237.

[51] L.L. Zhao, B.B. Yuan, Y.R. Geng, C.N. Yu, N.H. Kim, J.H. Lee, P. Li, Fabrication of Ultrahigh Hydrogen Barrier Polyethyleneimine/Graphene Oxide Films by Lbl Assembly Fine-Tuned with Electric Field Application, Compos. Part A Appl. Sci. Manuf., 78 (2015) 60-69.

[52] C.L. Chi, X.R. Wang, Y.W. Peng, Y.H. Qian, Z.G. Hu, J.Q. Dong, D. Zhao, Facile Preparation of Graphene Oxide Membranes for Gas Separation, Chem. Mater., 28 (2016) 2921-2927.

[53] S.F. Pei, H.M. Cheng, The Reduction of Graphene Oxide, Carbon, 50 (2012) 3210- 3228.

130

[54] W.S. Hummers, R.E. Offeman, Preparation of Graphitic Oxide, J. Am. Chem. Soc., 80 (1958) 1339-1339.

[55] D.C. Marcano, D.V. Kosynkin, J.M. Berlin, A. Sinitskii, Z.Z. Sun, A. Slesarev, L.B. Alemany, W. Lu, J.M. Tour, Improved Synthesis of Graphene Oxide, ACS Nano, 4 (2010) 4806-4814.

[56] D.R. Dreyer, S. Park, C.W. Bielawski, R.S. Ruoff, The Chemistry of Graphene Oxide, Chem. Soc. Rev., 39 (2010) 228-240.

[57] A. Lerf, H.Y. He, M. Forster, J. Klinowski, Structure of Graphite Oxide Revisited, J. Phys. Chem. B, 102 (1998) 4477-4482.

[58] G. Eda, M. Chhowalla, Chemically Derived Graphene Oxide: Towards Large-Area Thin-Film Electronics and Optoelectronics, Adv. Mater., 22 (2010) 2392-2415.

[59] X.L. Li, G.Y. Zhang, X.D. Bai, X.M. Sun, X.R. Wang, E. Wang, H.J. Dai, Highly Conducting Graphene Sheets and Langmuir-Blodgett Films, Nat. Nanotechnol., 3 (2008) 538-542.

[60] S. Quan, S.W. Li, Y.C. Xiao, L. Shao, CO2-Selective Mixed Matrix Membranes (MMMs) Containing Graphene Oxide (GO) for Enhancing Sustainable CO2 Capture, Int. J. Greenh. Gas Control., 56 (2017) 22-29.

[61] J. Zhao, Y.W. Zhu, G.W. He, R.S. Xing, F.S. Pan, Z.Y. Jiang, P. Zhang, X.Z. Cao, B.Y. Wang, Incorporating Zwitterionic Graphene Oxides into Sodium Alginate Membrane for Efficient Water/Alcohol Separation, ACS Appl. Mater. Interfaces, 8 (2016) 2097-2103.

[62] B.X. Mi, Graphene Oxide Membranes for Ionic and Molecular Sieving, Science, 343 (2014) 740-742.

[63] W.S. Hung, Q.F. An, M. De Guzman, H.Y. Lin, S.H. Huang, W.R. Liu, C.C. Hu, K.R. Lee, J.Y. Lai, Pressure-Assisted Self-Assembly Technique for Fabricating Composite Membranes Consisting of Highly Ordered Selective Laminate Layers of Amphiphilic Graphene Oxide, Carbon, 68 (2014) 670-677.

[64] R. Ding, H.Q. Zhang, Y.F. Li, J.T. Wang, B.B. Shi, H. Mao, J.C. Dang, J.D. Liu, Graphene Oxide-Embedded Nanocomposite Membrane for Solvent Resistant

131

Nanofiltration With Enhanced Rejection Ability, Chem. Eng. Sci., 138 (2015) 227-238.

[65] C.H. Tsou, Q.F. An, S.C. Lo, M. De Guzman, W.S. Hung, C.C. Hu, K.R. Lee, J.Y. Lai, Effect of Microstructure of Graphene Oxide Fabricated Through Different Self- Assembly Techniques On 1-Butanol Dehydration, J. Membr. Sci., 477 (2015) 93-100.

[66] S.J. Xia, M. Ni, T.R. Zhu, Y. Zhao, N.N. Li, Ultrathin Graphene Oxide Nanosheet Membranes with Various D-Spacing Assembled Using the Pressure-Assisted Filtration Method for Removing Natural Organic Matter, Desalination, 371 (2015) 78-87.

[67] S.A. Kiran, Y.L. Thuyavan, G. Arthanareeswaran, T. Matsuura, A.F. Ismail, Impact of Graphene Oxide Embedded Polyethersulfone Membranes for The Effective Treatment of Distillery Effluent, Chem. Eng. J., 286 (2016) 528-537.

[68] J.A. Prince, S. Bhuvana, V. Anbharasi, N. Ayyanar, K.V.K. Boodhoo, G. Singh, Ultra-Wetting Graphene-Based Pes Ultrafiltration Membrane - A Novel Approach for Successful Oil-Water Separation, Water Res., 103 (2016) 311-318.

[69] K. Huang, G.P. Liu, Y.Y. Lou, Z.Y. Dong, J. Shen, W.Q. Jin, A Graphene Oxide Membrane with Highly Selective Molecular Separation of Aqueous Organic Solution, Angew. Chem. Int. Ed., 53 (2014) 6929-6932.

[70] S. Kim, J. Nham, Y.S. Jeong, C.S. Lee, S.H. Ha, H.B. Park, Y.J. Lee, Biomimetic Selective Ion Transport through Graphene Oxide Membranes Functionalized with Ion Recognizing Peptides, Chem. Mater., 27 (2015) 1255-1261.

[71] A. Gimenez-Perez, S.K. Bikkarolla, J. Benson, C. Bengoa, F. Stuber, A. Fortuny, A. Fabregat, J. Font, P. Papakonstantinou, Synthesis Of N-Doped and Non-Doped Partially Oxidised Graphene Membranes Supported Over Ceramic Materials, J. Mater. Sci., 51 (2016) 8346-8360.

[72] B. Liang, P. Zhang, J.Q. Wang, J. Qu, L.F. Wang, X.X. Wang, C.F. Guan, K. Pan, Membranes with Selective Laminar Nanochannels Of Modified Reduced Graphene Oxide for Water Purification, Carbon, 103 (2016) 94-100.

[73] S. Xia, M. Ni, T. Zhu, Y. Zhao, N. Li, Ultrathin Graphene Oxide Nanosheet Membranes with Various D-Spacing Assembled Using the Pressure-Assisted Filtration Method for Removing Natural Organic Matter, Desalination, 371 (2015) 78-87.

132

[74] Y. Wei, Y.S. Zhang, X.L. Gao, Y.Q. Yuan, B.W. Su, C.J. Gao, Declining Flux and Narrowing Nanochannels Under Wrinkles of Compacted Graphene Oxide Nanofiltration Membranes, Carbon, 108 (2016) 568-575.

[75] Y.Y. Lou, G.P. Liu, S.N. Liu, J. Shen, W.Q. Jin, A Facile Way to Prepare Ceramic- Supported Graphene Oxide Composite Membrane Via Silane-Graft Modification, Appl. Surf. Sci., 307 (2014) 631-637.

[76] J. Zhao, Y.W. Zhu, F.S. Pan, G.W. He, C.H. Fang, K.T. Cao, R.S. Xing, Z.Y. Jiang, Fabricating Graphene Oxide-Based Ultrathin Hybrid Membrane for Pervaporation Dehydration Via Layer-By-Layer Self-Assembly Driven by Multiple Interactions, J. Membr. Sci., 487 (2015) 162-172.

[77] T. Wang, J.R. Lu, L.L. Mao, Z.N. Wang, Electric Field Assisted Layer-By-Layer Assembly of Graphene Oxide Containing Nanofiltration Membrane, J. Membr. Sci., 515 (2016) 125-133.

[78] S. Park, K.S. Lee, G. Bozoklu, W. Cai, S.T. Nguyen, R.S. Ruoff, Graphene Oxide Papers Modified by Divalent Ions - Enhancing Mechanical Properties via Chemical Cross-Linking, ACS Nano, 2 (2008) 572-578.

[79] Z.T. Luo, Y. Lu, L.A. Somers, A.T.C. Johnson, High Yield Preparation of Macroscopic Graphene Oxide Membranes, J. Am. Chem. Soc., 131 (2009) 898-899.

[80] S.X. Zheng, Q.S. Tu, J.J. Urban, S.F. Li, B.X. Mi, Swelling of Graphene Oxide Membranes in Aqueous Solution: Characterization of Interlayer Spacing and Insight into Water Transport Mechanisms, ACS Nano, 11 (2017) 6440-6450.

[81] D.A. Dikin, S. Stankovich, E.J. Zimney, R.D. Piner, G.H.B. Dommett, G. Evmenenko, S.T. Nguyen, R.S. Ruoff, Preparation and Characterization of Graphene Oxide Paper, Nature, 448 (2007) 457-460.

[82] R.K. Joshi, P. Carbone, F.C. Wang, V.G. Kravets, Y. Su, I.V. Grigorieva, H.A. Wu, A.K. Geim, R.R. Nair, Precise and Ultrafast Molecular Sieving Through Graphene Oxide Membranes, Science, 343 (2014) 752-754.

[83] L. Chen, G. Shi, J. Shen, B. Peng, B. Zhang, Y. Wang, F. Bian, J. Wang, D. Li, Z. Qian, G. Xu, G. Liu, J. Zeng, L. Zhang, Y. Yang, G. Zhou, M. Wu, W. Jin, J. Li, H. Fang, Ion sieving in graphene oxide membranes via cationic control of interlayer spacing,

133

Nature, 550 (2017) 380-383.

[84] Z.Q. Jia, Y. Wang, W.X. Shi, J.L. Wang, Diamines Cross-Linked Graphene Oxide Free-Standing Membranes for Ion Dialysis Separation, J. Membr. Sci., 520 (2016) 139- 144.

[85] S.F. Wang, Y.Z. Wu, N. Zhang, G.W. He, Q.P. Xin, X.Y. Wu, H. Wu, X.Z. Cao, M.D. Guiver, Z.Y. Jiang, A Highly Permeable Graphene Oxide Membrane with Fast and Selective Transport Nanochannels for Efficient Carbon Capture, Energy Environ. Sci., 9 (2016) 3107-3112.

[86] W.S. Hung, C.H. Tsou, M. De Guzman, Q.F. An, Y.L. Liu, Y.M. Zhang, C.C. Hu, K.R. Lee, J.Y. Lai, Cross-Linking with Diamine Monomers to Prepare Composite Graphene Oxide-Framework Membranes with Varying d-spacing, Chem. Mater., 26 (2014) 2983-2990.

[87] P.Z. Sun, K.L. Wang, H.W. Zhu, Recent Developments in Graphene-Based Membranes: Structure, Mass-Transport Mechanism and Potential Applications, Adv. Mater., 28 (2016) 2287-2310.

[88] J. Abraham, K.S. Vasu, C.D. Williams, K. Gopinadhan, Y. Su, C.T. Cherian, J. Dix, E. Prestat, S.J. Haigh, I.V. Grigorieva, P. Carbone, A.K. Geim, R.R. Nair, Tunable Sieving of Ions Using Graphene Oxide Membranes, Nat. Nanotechnol., 12 (2017) 546- 550.

[89] J. Shen, G.P. Liu, K. Huang, Z.Y. Chu, W.Q. Jin, N.P. Xu, Subnanometer Two- Dimensional Graphene Oxide Channels for Ultrafast Gas Sieving, ACS Nano, 10 (2016) 3398-3409.

[90] J. Zhu, C.M. Andres, J.D. Xu, A. Ramamoorthy, T. Tsotsis, N.A. Kotov, Pseudonegative Thermal Expansion and the State of Water in Graphene Oxide Layered Assemblies, ACS Nano, 6 (2012) 8357-8365.

[91] G.Y. Dong, Y.T. Zhang, J.W. Hou, J.N. Shen, V. Chen, Graphene Oxide Nanosheets Based Novel Facilitated Transport Membranes for Efficient CO2 Capture, Ind. Eng. Chem. Res., 55 (2016) 5403-5414.

[92] M. Karunakaran, L.F. Villalobos, M. Kumar, R. Shevate, F.H. Akhtar, K.V. Peinemann, Graphene Oxide Doped Ionic Liquid Ultrathin Composite Membranes for

134

Efficient CO2 Capture, J. Mater. Chem. A, 5 (2017) 649-656.

[93] Y.X. Hu, J. Wei, Y. Liang, H.C. Zhang, X.W. Zhang, W. Shen, H.T. Wang, Zeolitic Imidazolate Framework/Graphene Oxide Hybrid Nanosheets as Seeds for the Growth of Ultrathin Molecular Sieving Membranes, Angew. Chem. Int. Ed., 55 (2016) 2048- 2052.

[94] H.L. Guo, G.S. Zhu, H. Li, X.Q. Zou, X.J. Yin, W.S. Yang, S.L. Qiu, R. Xu, Hierarchical Growth of Large-Scale Ordered Zeolite Silicalite-1 Membranes with High Permeability and Selectivity for Recycling CO2, Angew. Chem. Int. Ed., 45 (2006) 7053-7056.

[95] S. Li, C.Q. Fan, High-Flux SAPO-34 Membrane for CO2/N2 Separation, Ind. Eng. Chem. Res., 49 (2010) 4399-4404.

[96] A.L. Ahmad, Z.A. Jawad, S.C. Low, S.H.S. Zein, A Cellulose Acetate/Multi- Walled Carbon Nanotube Mixed Matrix Membrane for CO2/N2 Separation, J. Membr. Sci., 451 (2014) 55-66.

[97] Y. Dai, J.R. Johnson, O. Karvan, D.S. Sholl, W.J. Koros, Ultem®/ZIF-8 Mixed Matrix Hollow Fiber Membranes for CO2/N2 Separations, J. Membr. Sci., 401-402 (2012) 76-82.

[98] S.G. Li, J.L. Falconer, R.D. Noble, Improved SAPO-34 Membranes for CO2/CH4 Separations, Adv. Mater., 18 (2006) 2601-2603.

[99] D.D. Iarikov, P. Hacarlioglu, S.T. Oyama, Supported Room Temperature Ionic Liquid Membranes for CO2/CH4 Separation, Chem. Eng. J., 166 (2011) 401-406.

[100] K. Zahri, P. Goh, A. Ismail, The Incorporation of Graphene Oxide into Polysulfone Mixed Matrix Membrane for CO2/CH4 Separation, IOP Conf. Ser.: Earth Environ. Sci, 36 (2016) 012007.

[101] Y. Zhao, B.T. Jung, L. Ansaloni, W.S.W. Ho, Multiwalled Carbon Nanotube Mixed Matrix Membranes Containing Amines for High Pressure CO2/H2 Separation, J. Membr. Sci., 459 (2014) 233-243.

[102] H. Li, Z.N. Song, X.J. Zhang, Y. Huang, S.G. Li, Y.T. Mao, H.J. Ploehn, Y. Bao, M. Yu, Ultrathin, Molecular-Sieving Graphene Oxide Membranes for Selective

135

Hydrogen Separation, Science, 342 (2013) 95-98.

[103] D.E. Jiang, V.R. Cooper, S. Dai, Porous Graphene as the Ultimate Membrane for Gas Separation, Nano Lett., 9 (2009) 4019-4024.

[104] K.C. Guan, J. Shen, G.P. Liu, J. Zhao, H.L. Zhou, W.Q. Jin, Spray-Evaporation Assembled Graphene Oxide Membranes for Selective Hydrogen Transport, Sep. Purif. Technol., 174 (2017) 126-135.

[105] L.M. Robeson, The Upper Bound Revisited, J. Membr. Sci., 320 (2008) 390-400.

[106] D. Cohen-Tanugi, J.C. Grossman, Water Desalination across Nanoporous Graphene, Nano Lett., 12 (2012) 3602-3608.

[107] H. Lin, N. Mehra, Y. Li, J. Zhu, Graphite Oxide/Boron Nitride Hybrid Membranes: The Role of Cross-Plane Laminar Bonding for A Durable Membrane with Large Water Flux and High Rejection Rate, J. Membr. Sci., 593 (2020) 117401.

[108] S. Safaei, R. Tavakoli, On the Design of Graphene Oxide Nanosheets Membranes for Water Desalination, Desalination, 422 (2017) 83-90.

[109] Y.T. Nam, J. Choi, K.M. Kang, D.W. Kim, H.T. Jung, Enhanced Stability of Laminated Graphene Oxide Membranes for Nanofiltration via Interstitial Amide Bonding, ACS Appl. Mater. Interfaces, 8 (2016) 27376-27382.

[110] Y. Han, Z. Xu, C. Gao, Ultrathin Graphene Nanofiltration Membrane for Water Purification, Adv. Funct. Mater., 23 (2013) 3693-3700.

[111] M. Coleman, X.W. Tang, Diffusive Transport of Two Charge Equivalent and Structurally Similar Ruthenium Complex Ions Through Graphene Oxide Membranes, Nano Res., 8 (2015) 1128-1138.

[112] Z.Q. Jia, Y. Wang, Covalently Crosslinked Graphene Oxide Membranes by Esterification Reactions for Ions Separation, J. Mater. Chem. A, 3 (2015) 4405-4412.

[113] M.C. Zhang, K.C. Guan, Y.F. Ji, G.P. Liu, W.Q. Jin, N.P. Xu, Controllable Ion Transport by Surface-Charged Graphene Oxide Membrane, Nat. Commun., 10 (2019) 1253.

[114] H.M. Hegab, L.D. Zou, Graphene Oxide-Assisted Membranes: Fabrication and Potential Applications in Desalination and Water Purification, J. Membr. Sci., 484

136

(2015) 95-106.

[115] D.J. Miller, D.R. Dreyer, C.W. Bielawski, D.R. Paul, B.D. Freeman, Surface Modification of Water Purification Membranes, Angew. Chem. Int. Ed., 56 (2017) 4662-4711.

[116] J.H. Jang, J.Y. Woo, J. Lee, C.S. Han, Ambivalent Effect of Thermal Reduction in Mass Rejection through Graphene Oxide Membrane, Environ. Sci. Technol., 50 (2016) 10024-10030.

[117] F. Perreault, M.E. Tousley, M. Elimelech, Thin-Film Composite Polyamide Membranes Functionalized with Biocidal Graphene Oxide Nanosheets, Environ. Sci. Tech. Let., 1 (2014) 71-76.

[118] H.M. Hegab, Y. Wimalasiri, M. Ginic-Markovic, L. Zou, Improving the Fouling Resistance of Brackish Water Membranes via Surface Modification with Graphene Oxide Functionalized Chitosan, Desalination, 365 (2015) 99-107.

[119] C.A. Scholes, K.H. Smith, S.E. Kentish, G.W. Stevens, CO2 Capture from Pre- Combustion Processes-Strategies for Membrane Gas Separation, Int. J. Greenh. Gas Con., 4 (2010) 739-755.

[120] G.M. Geise, H.S. Lee, D.J. Miller, B.D. Freeman, J.E. McGrath, D.R. Paul, Water Purification by Membranes: The Role of Polymer Science, J Polym Sci B Polym Phys, 48 (2010) 1685-1718.

[121] K.Y. Wang, S.W. Foo, T.S. Chung, Mixed Matrix PVDF Hollow Fiber Membranes with Nanoscale Pores for Desalination through Direct Contact Membrane Distillation, Ind. Eng. Chem. Res., 48 (2009) 4474-4483.

[122] L. Huang, M. Zhang, C. Li, G.Q. Shi, Graphene-Based Membranes for Molecular Separation, J. Phys. Chem. Lett., 6 (2015) 2806-2815.

[123] Z. Liu, W. Wang, X.J. Ju, R. Xie, L.Y. Chu, Graphene-Based Membranes for Molecular and Ionic Separations in Aqueous Environments, Chin. J. Chem. Eng., 25 (2017) 1598-1605.

[124] S. Kim, X.C. Lin, R.W. Ou, H.Y. Liu, X.W. Zhang, G.P. Simon, C.D. Easton, H.T. Wang, Highly Crosslinked, Chlorine Tolerant Polymer Network Entwined Graphene

137

Oxide Membrane for Water Desalination, J. Mater. Chem. A, 5 (2017) 1533-1540.

[125] L.M. Jin, Z.Y. Wang, S.X. Zheng, B.X. Mi, Polyamide-Crosslinked Graphene Oxide Membrane for Forward Osmosis, J. Membr. Sci., 545 (2018) 11-18.

[126] F.L. Zhou, H.N. Tien, W.W.L. Xu, J.T. Chen, Q.L. Liu, E. Hicks, M. Fathizadeh, S.G. Li, M. Yu, Ultrathin Graphene Oxide-Based Hollow Fiber Membranes with Brush- Like Co2-Philic Agent for Highly Efficient CO2 Capture, Nat. Commun., 8 (2017).

[127] Q.Z. Xue, X.L. Pan, X.F. Li, J.Q. Zhang, Q.K. Guo, Effective Enhancement of Gas Separation Performance in Mixed Matrix Membranes Using Core/Shell Structured Multi-Walled Carbon Nanotube/Graphene Oxide Nanoribbons, Nanotechnology, 28 (2017) 6.

[128] R.R. Nair, Graphene-Based Membranes with Limited Swelling Sieve Common Salts Out of Salty Water, Membr. Technol., 2017 (2017) 7.

[129] H.G. Wei, J.H. Zhu, S.J. Wu, S.Y. Wei, Z.H. Guo, Electrochromic Polyaniline/Graphite Oxide Nanocomposites with Endured Electrochemical Energy Storage, Polymer, 54 (2013) 1820-1831.

[130] A.C. Ferrari, D.M. Basko, Raman Spectroscopy as A Versatile Tool for Studying the Properties of Graphene, Nat. Nanotechnol., 8 (2013) 235.

[131] I.K. Moon, J. Lee, R.S. Ruoff, H. Lee, Reduced graphene oxide by chemical graphitization, Nat. Commun., 1 (2010) 73.

[132] N. Wu, X.L. She, D.J. Yang, X.F. Wu, F.B. Su, Y.F. Chen, Synthesis of Network Reduced Graphene Oxide in Polystyrene Matrix by A Two-Step Reduction Method for Superior Conductivity of the Composite, J. Mater. Chem., 22 (2012) 17254-17261.

[133] L.C. Lin, J.C. Grossman, Atomistic Understandings of Reduced Graphene Oxide as An Ultrathin-Film Nanoporous Membrane for Separations, Nat. Commun., 6 (2015) 8335.

[134] H.C. Schniepp, J.L. Li, M.J. McAllister, H. Sai, M. Herrera-Alonso, D.H. Adamson, R.K. Prud'homme, R. Car, D.A. Saville, I.A. Aksay, Functionalized Single Graphene Sheets Derived from Splitting Graphite Oxide, J. Phys. Chem. B, 110 (2006) 8535-8539.

138

[135] S. Lashkari, A. Tran, B. Kruczek, Effect of Back Diffusion and Back Permeation of Air on Membrane Characterization in Constant Pressure System, J. Membr. Sci., 324 (2008) 162-172.

[136] A.G. Volkov, S. Paula, D.W. Deamer, Two Mechanisms of Permeation of Small Neutral Molecules and Hydrated Ions Across Phospholipid Bilayers, Bioelectrochem. Bioenerg., 42 (1997) 153-160.

[137] J. Kielland, Individual Activity Coefficients of Ions in Aqueous Solutions, J. Am. Chem. Soc., 59 (1937) 1675-1678.

[138] X.B. Hu, Y. Yu, J.E. Zhou, Y.Q. Wang, J. Liang, X.Z. Zhang, Q.B. Chang, L.X. Song, The Improved Oil/Water Separation Performance of Graphene Oxide Modified Al2O3 Microfiltration Membrane, J. Membr. Sci., 476 (2015) 200-204.

[139] R.L.G. Lecaros, G.E.J. Mendoza, W.S. Hung, Q.F. An, A.R. Caparanga, H.A. Tsai, C.C. Hu, K.R. Lee, J.Y. Lai, Tunable Interlayer Spacing of Composite Graphene Oxide- Framework Membrane for Acetic Acid Dehydration, Carbon, 123 (2017) 660-667.

[140] L. Huang, J. Chen, T.T. Gao, M. Zhang, Y.R. Li, L.M. Dai, L.T. Qu, G.Q. Shi, Reduced Graphene Oxide Membranes for Ultrafast Organic Solvent Nanofi ltration, Adv. Mater., 28 (2016) 8669-8674.

[141] L.G. Wu, C.H. Yang, T. Wang, X.Y. Zhang, Enhanced the Performance of Graphene Oxide/Polyimide Hybrid Membrane for Co2 Separation by Surface Modification of Graphene Oxide Using Polyethylene Glycol, Appl. Surf. Sci., 440 (2018) 1063-1072.

[142] Y. Zhang, S. Zhang, T.S. Chung, Nanometric Graphene Oxide Framework Membranes with Enhanced Heavy Metal Removal via Nanofiltration, Environ. Sci. Technol., 49 (2015) 10235-10242.

[143] F. Kim, L.J. Cote, J.X. Huang, Graphene Oxide: Surface Activity and Two- Dimensional Assembly, Adv. Mater., 22 (2010) 1954-1958.

[144] L. Chen, L.L. Huang, J.H. Zhu, Stitching Graphene Oxide Sheets into A Membrane at A Liquid/Liquid Interface, Chem. Commun., 50 (2014) 15944-15947.

[145] X. Shen, X.Y. Lin, N. Yousefi, J.J. Jia, J.K. Kim, Wrinkling in Graphene Sheets

139

and Graphene Oxide Papers, Carbon, 66 (2014) 84-92.

[146] I. Rodriguez-Pastor, G. Ramos-Fernandez, H. Varela-Rizo, M. Terrones, I. Martin-Gullon, Towards the Understanding of The Graphene Oxide Structure: How to Control the Formation of Humic- And Fulvic-Like Oxidized Debris, Carbon, 84 (2015) 299-309.

[147] M. Arias, E. Lopez, A. Nunez, D. Rubinos, B. Soto, M.T. Barral, F. Diaz-Fierros, Adsorption of Methylene Blue by Red Mud, An Oxide-Rich Byproduct of Bauxite Refining, Effect of Mineral-Organic-Microorganism Interaction on Soil and Freshwater Environments, (1999) 361-365.

[148] H.B. Park, J. Kamcev, L.M. Robeson, M. Elimelech, B.D. Freeman, Maximizing the Right Stuff: The Trade-Off Between Membrane Permeability and Selectivity, Science, 356 (2017) 6343.

[149] T.H. Bae, J.S. Lee, W.L. Qiu, W.J. Koros, C.W. Jones, S. Nair, A High- Performance Gas-Separation Membrane Containing Submicrometer-Sized Metal- Organic Framework Crystals, Angew. Chem. Int. Ed., 49 (2010) 9863-9866.

[150] G.M. Geise, H.B. Park, A.C. Sagle, B.D. Freeman, J.E. McGrath, Water Permeability and Water/Salt Selectivity Tradeoff in Polymers for Desalination, J. Membr. Sci., 369 (2011) 130-138.

[151] H.W. Yoon, Y.H. Cho, H.B. Park, Graphene-Based Membranes: Status and Prospects, Phil. Trans. R, Soc. A, 374 (2016) 20150024.

[152] G.T. Rochelle, Amine Scrubbing for CO2 Capture, Science, 325 (2009) 1652- 1654.

[153] A.B. Rao, E.S. Rubin, A Technical, Economic, And Environmental Assessment of Amine-Based CO2 Capture Technology for Power Plant Greenhouse Gas Control, Environ. Sci. Technol., 36 (2002) 4467-4475.

[154] R. Vaidhyanathan, S.S. Iremonger, G.K.H. Shimizu, P.G. Boyd, S. Alavi, T.K. Woo, Direct Observation and Quantification of CO2 Binding Within an Amine- Functionalized Nanoporous Solid, Science, 330 (2010) 650-653.

[155] M.S. Walters, Y.J. Lin, D.J. Sachde, T.F. Edgar, G.T. Rochelle, Control Relevant

140

Model of Amine Scrubbing for CO2 Capture from Power Plants, Ind. Eng. Chem. Res., 55 (2016) 1690-1700.

[156] K. Kim, O. Buyukcakir, A. Coskun, Diazapyrenium-Based Porous Cationic Polymers for Colorimetric Amine Sensing and Capture from CO2 Scrubbing Conditionsl, RSC Adv., 6 (2016) 77406-77409.

[157] K.P. Lee, T.C. Arnot, D. Mattia, A Review of Reverse Osmosis Membrane Materials for Desalination-Development to Date and Future Potential, J. Membr. Sci., 370 (2011) 1-22.

[158] E. Drioli, A. Criscuoli, E. Curcio, Integrated Membrane Operations for Seawater Desalination, Desalination, 147 (2002) 77-81.

[159] G. Crini, Non-Conventional Low-Cost Adsorbents for Dye Removal: A Review, Bioresour. Technol., 97 (2006) 1061-1085.

[160] S. Yu, M. Liu, M. Ma, M. Qi, Z. Lü, C. Gao, Impacts of Membrane Properties on Reactive Dye Removal from Dye/Salt Mixtures by Asymmetric Cellulose Acetate and Composite Polyamide Nanofiltration Membranes, J. Membr. Sci., 350 (2010) 83-91.

[161] I. Koyuncu, Reactive Dye Removal in Dye/Salt Mixtures by Nanofiltration Membranes Containing Vinylsulphone Dyes: Effects of Feed Concentration and Cross Flow Velocity, Desalination, 143 (2002) 243-253.

[162] C. Blöcher, J. Dorda, V. Mavrov, H. Chmiel, N.K. Lazaridis, K.A. Matis, Hybrid Flotation—Membrane Filtration Process for The Removal of Heavy Metal Ions from Wastewater, Water Res., 37 (2003) 4018-4026.

[163] H. Abu Qdais, H. Moussa, Removal of Heavy Metals from Wastewater by Membrane Processes: A Comparative Study, Desalination, 164 (2004) 105-110.

[164] R.K. Joshi, S. Alwarappan, M. Yoshimura, V. Sahajwalla, Y. Nishina, Graphene Oxide: The New Membrane Material, Appl. Mater. Today, 1 (2015) 1-12.

[165] C. Athanasekou, M. Pedrosa, T. Tsoufis, L.M. Pastrana-Martínez, G. Romanos, E. Favvas, F. Katsaros, A. Mitropoulos, V. Psycharis, A.M.T. Silva, Comparison of Self- Standing and Supported Graphene Oxide Membranes Prepared by Simple Filtration: Gas and Vapor Separation, Pore Structure and Stability, J. Membr. Sci., 522 (2017) 303-

141

315.

[166] H. Lin, S. Dangwal, R. Liu, S.-J. Kim, Y. Li, J. Zhu, Reduced Wrinkling in GO Membrane by Grafting Basal-plane Groups for Improved Gas and Liquid Separations, J. Membr. Sci., 563 (2018) 336-344.

[167] R. Liu, G. Arabale, J. Kim, K. Sun, Y. Lee, C. Ryu, C. Lee, Graphene Oxide Membrane for Liquid Phase Organic Molecular Separation, Carbon, 77 (2014) 933-938.

[168] H. Lin, R. Liu, S. Dangwal, S.-J. Kim, N. Mehra, Y. Li, J. Zhu, Permselective H2/CO2 Separation and Desalination of Hybrid GO/rGO Membranes with Controlled Pre-cross-linking, ACS Appl. Mater. Interfaces, 10 (2018) 28166-28175.

[169] C.N. Yeh, K. Raidongia, J.J. Shao, Q.H. Yang, J.X. Huang, On the Origin of The Stability of Graphene Oxide Membranes in Water, Nat. Chem., 7 (2015) 166-170.

[170] Y. Lin, J.W. Connell, Advances in 2d Boron Nitride Nanostructures: Nanosheets, Nanoribbons, Nanomeshes, and Hybrids with Graphene, Nanoscale, 4 (2012) 6908- 6939.

[171] C. Chen, J.M. Wang, D. Liu, C. Yang, Y.C. Liu, R.S. Ruoff, W.W. Lei, Functionalized Boron Nitride Membranes with Ultrafast Solvent Transport Performance for Molecular Separation, Nat. Commun., 9 (2018) 1902.

[172] W.W. Lei, V.N. Mochalin, D. Liu, S. Qin, Y. Gogotsi, Y. Chen, Boron Nitride Colloidal Solutions, Ultralight Aerogels and Freestanding Membranes Through One- Step Exfoliation and Functionalization, Nat. Commun., 6 (2015) 8849.

[173] B.D. Ossonon, D. Belanger, Synthesis and Characterization of Sulfophenyl- Functionalized Reduced Graphene Oxide Sheets, RSC Adv., 7 (2017) 27224-27234.

[174] M. Naebe, J. Wang, A. Amini, H. Khayyam, N. Hameed, L.H. Li, Y. Chen, B. Fox, Mechanical Property and Structure of Covalent Functionalised Graphene/Epoxy Nanocomposites, Sci. Rep., 4 (2014) 4375.

[175] F. Zhou, H.N. Tien, W.L. Xu, J.-T. Chen, Q. Liu, E. Hicks, M. Fathizadeh, S. Li, M. Yu, Ultrathin Graphene Oxide-Based Hollow Fiber Membranes with Brush-Like CO2-Philic Agent for Highly Efficient CO2 Capture, Nat. Commun., 8 (2017) 2107.

[176] J. Zhu, M. Tian, J. Hou, J. Wang, J. Lin, Y. Zhang, J. Liu, B. Van der Bruggen,

142

Surface Zwitterionic Functionalized Graphene Oxide for A Novel Loose Nanofiltration Membrane, J. Mater. Chem. A, 4 (2016) 1980-1990.

[177] Y. Oh, D.L. Armstrong, C. Finnerty, S.X. Zheng, M. Hu, A. Torrents, B.X. Mi, Understanding the pH-Responsive Behavior of Graphene Oxide Membrane in Removing Ions and Organic Micropollulants, J. Membr. Sci., 541 (2017) 235-243.

[178] K. Zahri, P.S. Goh, A.F. Ismail, The Incorporation of Graphene Oxide into Polysulfone Mixed Matrix Membrane for CO2 /CH4 Separation, IOP Conference Series: Earth and Environmental Science, 36 (2016) 012007.

[179] H. Lin, S. Dangwal, R.C. Liu, S.J. Kim, Y.F. Li, J.H. Zhu, Reduced Wrinkling in Go Membrane by Grafting Basal-Plane Groups for Improved Gas and Liquid Separations, J. Membr. Sci., 563 (2018) 336-344.

[180] W.-S. Hung, C.-H. Tsou, M. De Guzman, Q.-F. An, Y.-L. Liu, Y.-M. Zhang, C.- C. Hu, K.-R. Lee, J.-Y. Lai, Cross-Linking with Diamine Monomers to Prepare Composite Graphene Oxide-Framework Membranes with Varying d-Spacing, Chem. Mater., 26 (2014) 2983-2990.

[181] M. Ceppatelli, D. Scelta, G. Tuci, G. Giambastiani, M. Hanfland, R. Bini, High- Pressure Chemistry of Graphene Oxide in the Presence of Ar, N-2, and NH3, J. Phys. Chem. C, 120 (2016) 5174-5187.

[182] A. Gogoi, T.J. Konch, K. Raidongia, K.A. Reddy, Water and Salt Dynamics in Multilayer Graphene Oxide (Go) Membrane: Role of Lateral Sheet Dimensions, J. Membr. Sci., 563 (2018) 785-793.

[183] T. Rattana, S. Chaiyakun, N. Witit-anun, N. Nuntawong, P. Chindaudom, S. Oaew, C. Kedkeaw, P. Limsuwan, Preparation and Characterization of Graphene Oxide Nanosheets, Procedia Eng., 32 (2012) 759-764.

[184] O.C. Compton, D.A. Dikin, K.W. Putz, L.C. Brinson, S.T. Nguyen, Electrically Conductive “Alkylated” Graphene Paper via Chemical Reduction of Amine- Functionalized Graphene Oxide Paper, Adv. Mater. 22 (2010) 892-896.

[185] N.V. Medhekar, A. Ramasubramaniam, R.S. Ruoff, V.B. Shenoy, Hydrogen Bond Networks in Graphene Oxide Composite Paper: Structure and Mechanical Properties, ACS Nano, 4 (2010) 2300-2306.

143

[186] T. Liu, X.Y. Liu, N. Graham, W.Z. Yu, K.N. Sun, Two-Dimensional MXene Incorporated Graphene Oxide Composite Membrane with Enhanced Water Purification Performance, J. Membr. Sci., 593 (2020) 117431.

[187] G.K. Zhao, R.R. Hu, Y.J. He, H.W. Zhu, Physically Coating Nanofiltration Membranes with Graphene Oxide Quantum Dots for Simultaneously Improved Water Permeability and Salt/Dye Rejection, Adv. Mater. Interfaces, 6 (2019) 1801742.

[188] G.D. Kong, J. Pang, Y.C. Tang, L.L. Fan, H.X. Sun, R.M. Wang, S. Feng, Y. Feng, W.D. Fan, W.P. Kang, H.L. Guo, Z.X. Kang, D.F. Sun, Efficient Dye Nanofiltration of A Graphene Oxide Membrane via Combination with A Covalent Organic Framework by Hot Pressing, J. Mater. Chem. A, 7 (2019) 24301-24310.

[189] S. Rajesh, A.B. Bose, Development of Graphene Oxide Framework Membranes via the "from" and "to" Cross-Linking Approach for Ion-Selective Separations, ACS Appl. Mater. Interfaces, 11 (2019) 27706-27716.

[190] C.Y. Wang, W.J. Zeng, T.T. Jiang, X. Chen, X.L. Zhang, Incorporating Attapulgite Nanorods Into Graphene Oxide Nanofiltration Membranes for Efficient Dyes Wastewater Treatment, Sep. Purif. Technol., 214 (2019) 21-30.

[191] M. Mahmoudian, M.G. Kochameshki, M. Hosseinzadeh, Modification of Graphene Oxide by Atrp: A pH-Responsive Additive in Membrane for Separation of Salts, Dyes and Heavy Metals, J. Environ. Chem. Eng., 6 (2018) 3122-3134.

[192] T.T. Gao, H.B. Wu, L. Tao, L.T. Qu, C. Li, Enhanced Stability and Separation Efficiency of Graphene Oxide Membranes in Organic Solvent Nanofiltration, J. Mater. Chem. A, 6 (2018) 19563-19569.

[193] L. Dong, M. Li, S. Zhang, X. Si, Y. Bai, C. Zhang, NH2-Fe3O4-Regulated Graphene Oxide Membranes with Well-Defined Laminar Nanochannels for Desalination of Dye Solutions, Desalination, 476 (2020) 114227.

[194] Y. Han, Y.Q. Jiang, C. Gao, High-Flux Graphene Oxide Nanofiltration Membrane Intercalated by Carbon Nanotubes, ACS Appl. Mater. Interfaces, 7 (2015) 8147-8155.

[195] H.K. Roobottom, H.D.B. Jenkins, J. Passmore, L. Glasser, Thermochemical Radii of Complex Ions, J. Chem. Educ., 76 (1999) 1570-1573.

144