Designing Polymer Electrolytes for Alkaline Anion

DESIGNING POLYMER ELECTROLYTES FOR ALKALINE ANION

EXCHANGE MEMBRANE FUEL CELLS

A DISSERTATION

SUBMITTED TO THE DEPARTMENT OF CHEMICAL ENGINEERING

AND THE COMMITTEE ON GRADUATE STUDIES

OF STANFORD UNIVERSITY

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

Steve Sidi He

May 2015

This dissertation is online at: http://purl.stanford.edu/wh023dj3361

ii I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Curtis Frank, Primary Adviser

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Thomas Jaramillo

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Andrew Spakowitz

Approved for the Stanford University Committee on Graduate Studies. Patricia J. Gumport, Vice Provost for Graduate Education

This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file in University Archives.

iii

Abstract

Increasing global demand and dependence on fossil fuels, coupled with environmental concerns arising from their use, have sparked interest in alternative energy sources.

Hydrogen-powered fuel cells are a promising solution, offering a clean, scalable method for energy production. The most prominent low-temperature fuel cell devices today operate under an acidic environment, using a semi-permeable proton exchange membrane (PEM) to separate the two electrodes. However, their caustic operating conditions present unique stability and activity issues for the metal catalysts and ultimately necessitates the use of platinum-group materials, severely limiting commercial viability.

A potential solution is to operate the fuel cell device under an alkaline environment using an anion exchange membrane (AEM), transporting hydroxide ions in lieu of protons. The basic environment opens the door for cheaper catalysts based on nickel and molybdenum, eliminating the cost barrier associated with PEM fuel cells. Unfortunately, typical AEMs exhibit poorer ionic conductivity and stability compared to traditional acidic membranes (e.g. Nafion), offsetting any potential cost advantage they may afford.

This dissertation discusses design rationales towards enhancing the macroscopic properties of AEMs. Specifically, I present two experimental design motifs for improving the device viability of AEMs. In the first case, I present a semi- interpenetrating network design where a linear AEM ionomer is stabilized by a crosslinked poly(styrene-co-divinylbenzene) matrix. The crosslinked network acts as a reinforcing scaffold, dramatically increasing dimensional stability while maintaining

excellent anion conductivity. Prototypical single-stack fuel cells with enhanced performance and stability have been fabricated from these materials, validating the design choices. In the second approach, I demonstrate the ability to increase hydroxide conductivity by tuning the nanostructure of the polymer electrolyte. Specifically, I show that tethering hydrophilic poly(ethylene glycol) grafts onto a benzyltrimethylammonium polysulfone benchmark AEM results in phase-separated, water-rich domains on the order of 5 to 10 nm. These domains serve as an ion transport pathway, facilitating the diffusion of hydroxide anions and consequently enhancing the efficiency of hydroxide conduction.

Finally, in order to better understand the phase behavior and structure-property relationships of typical AEM materials, we have developed coarse-grained simulations and fundamental polymer theory to elucidate the thermodynamic behavior of random copolymers. We find that both the stochastic distribution of monomers along the polymer backbone as well as the overall stiffness of the polymer chain heavily influences its phase behavior (i.e., morphology and critical point). The ultimate objective is to provide not only a theoretical basis for understanding and explaining structure-property relationships in existing AEM materials, but to provide a set of general design guidelines moving forward.

Acknowledgments

I would like to begin by thanking the all the people at Stanford who have supported me throughout my Ph.D. career, without whom I would not be in a position to write this very sentence. I am truly lucky to have found a research advisor, Prof. Curtis W.

Frank, whose encouragement and support has have had a highly positive influence on me during my time at Stanford. Curt’s flexibility and understanding has allowed me to pursue my own research ideas, allowing me to grow professionally in a way that I otherwise could not have.

I would also like to thank Prof. Andrew Spakowitz for providing me an opportunity to work on polymer theory and simulations, the nature of which is a far cry from the experimental studies I had become familiar with. This experience has opened me to new perspectives and understanding of polymer science, tremendously accelerating my scientific growth. Andy has been highly encouraging and has always been open and receptive to my questions, no matter how trivial they might seem.

Prof. Thomas Jaramillo has been a great resource for not only helping me understand electrochemistry and energy conversion technologies, but how they fit into the overall economic picture; bridging fundamental science with practical implementation has helped shaped me into a better engineer.

In addition, I would like to thank Prof. Do Y. Yoon for his countless advice, comments and critiques throughout the years. Do has been a great source of knowledge on general polymer science, and his extensive experience has provided unique viewpoints and interpretation of my data.

My friends and fellow graduate students have played a critical role not only in my graduate experience, but my research as well. I especially thank Elyse Coletta, who had worked on a similar project concerning proton exchange membranes, for all her advice and our countless discussions. Desmond Ng was a great resource for teaching me how to fabricate membrane electrode assemblies from my materials.

Shifan Mao has been an instrumental collaborator in the theory and simulation work and was always available to answer my questions and concerns. Moreover, I would like to thank the various undergraduate and graduate rotation students I had the pleasure of working with: Sumit Mitra, Nathaniel Morrison, Joseph Troderman, Leo

Shaw, Jeff Lopez and Alaina Strickler. I especially thank Alaina Strickler, who played a key role in the protocol development for the semi-interpenetrating network work.

I thank the administrative staff in the Department of Chemical Engineering at

Stanford for all their help. In particular, Jeanne Cosby, Jeannie Lewindowski, Pamela

Dixon and Pam Juanes guaranteed everything went smoothly, from setting up purchasing accounts to making room and equipment reservations.

Finally, this work would not have been possible without funding and support from the TomKat Center for Sustainable Energy and the Precourt Institute for Energy.

vii

Dedication

I would like to dedicate this thesis to my family. My parents immigrated to the

United States when I was still a toddler to seek out a better life; growing up here has opened up all sorts of opportunities for me that would not have been possible had we stayed in China. I thank them for not only their sacrifice, but also for their support, care and trust to let me carve out my own path in life. In addition, I want to thank my wife, Marina Kim. She believed in me even when I doubted myself, and has been a constant source of comfort, encouragement and support.

In retrospect, none of this would have been possible without Dr. Dena Leggett, my high school AP Chemistry teacher. Prior to her class, I had no idea what chemistry was or what a scientific career entailed. Indeed, the whole endeavor at the time seemed daunting after hearing horror stories about “stoichiometry” from upperclassman. Yet, my experience was totally the opposite – Dr. Leggett made chemistry intuitive, interesting and, above all, fun. I am eternally grateful to her for opening my eyes to the world of science and engineering; if not for her, I would not be where I am today.

viii

Table of Contents

Chapter 1: Introduction ...... 1 1.1 Background ...... 1 1.2 Motivation and Outline ...... 4

Chapter 2: Background ...... 7 2.1 Hydroxide Transport in Anion Exchange Membranes ...... 7 2.2 Major Challenges ...... 11 2.2.1 Low Ionic Conductivity ...... 11 2.2.2 Poor Chemical Stability ...... 13 2.2.2.1 Cation Stability ...... 13 2.2.2.2 Backbone Stability ...... 15 2.2.3 Carbon Dioxide Poisoning ...... 18 2.3 Design Approaches ...... 18

Chapter 3: Literature Review ...... 20 3.1 Homogeneous Membranes ...... 20 3.1.1 Quaternary Ammonium Aromatic Random Copolymers ...... 21 3.1.1.1 Polysulfone ...... 21 3.1.1.2 Poly(phenylene oxide) ...... 30 3.1.1.3 Poly(ether ether ketone) ...... 33 3.1.2 Aliphatic Copolymers ...... 35 3.1.2.1 Covalent Crosslinking ...... 36 3.1.2.1 Chitosan Modification ...... 37 3.1.2.3 Radiation Grafting ...... 38 3.1.2.4 Ring Opening Metathesis Polymerization (ROMP) ...... 39 3.1.3 Alternative Headgroups (Non-Quaternary Ammonium) ...... 40 3.1.3.1 Phosphonium ...... 41 3.1.3.2 Imidazolium ...... 44 3.1.3.3 Guanidinium ...... 51 3.1.3.4 Pyrrolidinium ...... 52 3.1.3.5 Sulfonium ...... 53 3.1.3.6 Ruthenium ...... 54 3.2 Composite Membranes ...... 55

3.2.1 Pore-Filled Composites ...... 56 3.2.2 Electrospun Composites ...... 59 3.2.3 Inorganic Composites ...... 60 3.3 Phase-Segregated Membranes ...... 62 3.3.1 Random Copolymer ...... 63 3.3.1.1 Linear Random Copolymers ...... 63 3.3.1.2 Graft Copolymers ...... 64 3.3.1.3 Multiblock ...... 73 3.3.2 Diblock and Triblock Copolymers ...... 77 3.3.3 Theory and Simulation of Phase Separation in Random Copolymers ...... 79

Chapter 4: Materials and Methods ...... 83 4.1 Materials ...... 83 4.2 Synthesis Protocols ...... 83 4.2.1 Chloromethylation of Polysulfone ...... 83 4.2.2 Quaternization of Chloromethylated Polysulfone ...... 84 4.3 Characterization Protocols ...... 85 4.3.1 1H NMR Characterization ...... 85 4.3.2 Ion Exchange Capacity (IEC) ...... 86 4.3.3 Water Uptake ...... 86 4.3.4 Small Angle X-ray Scattering (SAXS) ...... 87 4.3.5 Alkaline Stability ...... 88 4.3.6 Thermogravimetric Analysis ...... 88 4.3.7 Mechanical Testing ...... 88 4.3.8 Conductivity Measurements ...... 89 4.3.9 Fuel Cell Tests ...... 90

Chapter 5: A Semi-Interpenetrating Network Approach for Stabilizing Highly Charged Anion Exchange Membranes for Alkaline Fuel Cells ...... 91 5.1 Introduction ...... 91 5.2 Results and Discussion ...... 94 5.2.1 Synthesis and Characterization ...... 94 5.2.2 Monomer Uptake and Ion Exchange Capacity ...... 98 5.2.3 Water Uptake and Temperature Stability ...... 100

5.2.4 Swelling Kinetics ...... 105 5.2.5 Mechanical Properties ...... 108 5.2.6 Anion Conductivity ...... 111 5.2.7 Leaching and Alkaline Stability ...... 115 5.2.8 Membrane Electrode Assembly Performance ...... 117 5.3 Conclusions ...... 119

Chapter 6: Facilitating Hydroxide Transport in Alkaline Exchange Membranes via Hydrophilic Poly(ethylene glycol) Grafts ...... 121 6.1 Introduction ...... 121 6.2 Methods ...... 124 6.2.1 PEGylation of Chloromethylated Polysulfone ...... 124 6.2.2 Quaternization of PSf-g-PEG ...... 124 6.3 Results and Discussion ...... 125 6.3.1 Synthesis ...... 125 6.3.2 Morphology ...... 126 6.3.3 Scaling Analysis ...... 130 6.3.4 Hydroxide Conductivity ...... 132 6.3.5 Water Uptake ...... 133 6.3.6 Alkaline Stability ...... 135 6.3.7 Fuel Cell Performance ...... 138 6.4 Conclusions ...... 139

Chapter 7: Simulation and Theory of Random Copolymers Towards Understanding the Morphologies of Anion Exchange Membrane Materials ...... 141 7.1 Introduction ...... 141 7.2 Method ...... 144 7.3 Results and Discussion ...... 147 7.4 Experimental Implications ...... 154 7.5 Conclusion and Outlook ...... 156

Chapter 8: Conclusions ...... 158 8.1 Summary ...... 158

8.2 Outlook ...... 159

Appendix A: Morphology of Semi-Interpenetrating Network Anion Exchange Membranes ...... 161

Appendix B: Alternative Semi-IPN Synthesis Protocol with Thermal- Initiated Polymerization ...... 167

References ...... 169

xii

List of Tables

Table 5.1 Composition of monomer soaking solutions and sample nomenclature. .... 95

Table 5.2 Intrinsic rate constant for water uptake at 20°C...... 106

Table 5.3 Mechanical properties of dry membranes (20°C and 35% RH)...... 108

Table 5.4 Mechanical properties of hydrated membranes (20°C in water bath)...... 109

Table 5.5 Comparison between empirical and predicted elastic modulus of hydrated semi-IPN materials...... 110

Table 5.6 AEM conductivity at 20°C and 100% RH for different counterions...... 115

Table 5.7 Leaching Stability of QA sIPN [72/20/8] in Cl- form...... 116

Table 6.1 Structural and Performance Data. (a) IEC [mmol OH- g-1] determined by NMR and back-titration (in parentheses). (b) OH- Conductivity [mS cm-1] at 60°C. (c) OH- Conductivity at 60°C normalized against titrated IEC values [mS g -1 -1 cm mmol ]. (d,e) Domain spacing d [nm] and correlation length ξ [nm] from Teubner-Strey fitting of the SAXS scattering profiles...... 128

Table 6.2 Water Uptake Data at 22°C. (a) Gravimetric water uptake. (b) λ [mol H2O mol-1 BTMA] as calculated from water uptake and titrated IEC. (c) λ-normalized - -1 -1 OH conductivity [mS cm mol H2O mol BTMA]...... 134

Table A.1 Phase composition of the DMA data as estimated by the Flory-Fox equation...... 164

xiii

List of Schemes

Scheme 2.1 Possible degradation pathways of benzyltrimethylammonium cation by hydroxide anion. Inset shows Hoffman degradation of an alkyltrimethylammonium cation via hydroxide attack of the β-hydrogen. Figure adapted from Varcoe et al.[20] ...... 14

Scheme 2.2 Proposed degradation pathways for benzyltrimethylammonium bisphenol A polysulfone. The electron withdrawing effects of the pendant cation is believed to render the (A) quaternary carbon and (B) ether groups susceptible to hydrolytic cleavage. Schematic adapted from Arges et al.[25] ...... 17

Scheme 3.3 Bisphenol A polysulfone with benzyltrimethylammonium sidechains (PAES-Q). Adapted from X. Li et al. [42] ...... 25

Scheme 3.4 Molecular designs with high concentration of ionic species. (A) PTMA PSf by Guiver et al., (B) Tetrafunctional PPSf by Choi et al., and (C) PAESK with tetrafunctional fluorenyl groups by Watanabe et al. Figure adapted from their respective sources...... 29

Scheme 3.5 Click reaction tethering 1,2,3-triazole to PPO. Figure adapted from Binder et al.[55] ...... 32

Scheme 3.6 Highly-crosslinked quaternary ammonium polypropylene- polyethyleneimine. Figure adapted from Herring et al. [60] ...... 37

Scheme 3.7 Quaternary ammoniun-functionalized chitosan (non-crosslinked). Adapted from Y. Wan et al.[62] ...... 37

Scheme 3.8 Ring Opening Metathesis Polymerization of tetraalkylammmonium- norbornene with dicyclopentadiene. Scheme adapted from Coates et al. [66] ...... 39

Scheme 3.9 Polyethylene functionalized with tetrakis(dialkylamino)phosphonium. Figure adapted from Coates et al.[79] ...... 43

Scheme 3.10 Alkaline stable anionic PBI with mesitylene group. Adapted from Holdcroft et al. [96] ...... 49

Scheme 3.11 Permutations of ionized/non-ionized PBI with and without imidazolium functionalities reported by L. Jheng et al. [97] ...... 50

Scheme 3.12 Pyrrolidinium-based random copolymer membrane reported by J. Qiao et al. [101] ...... 52

Scheme 3.13 Tertiary sulfonium with methoxyphenyl sidegroup tethered to PSf backbone. Image adapted from Y. Yan et al. [103] ...... 54

xiv

Scheme 3.14 Ruthenium-functionalized polynorbornene. Figure adapted from Hickner et al.[104] ...... 55

Scheme 3.15 Polyisoprene-co-poly(vinylbenzyltrimethylammonium chloride) random copolymer. Figure adapted from Tsai et al.[122] ...... 64

Scheme 3.16 Comb-shaped QA PPO membrane with aliphatic tail. Figure adapted from N. Li et al.[124] ...... 66

Scheme 3.17 Poly(arylene ether nitrile) random copolymer. Dark regions in TEM image indicate stained hydrophilic domains. Figure adapted from Q. Liu et al.[126] ...... 67

Scheme 3.18 Quaternary ammonium PPO with spacer chains reported by (A) Jannasch et al.[128] and (B) Bai et al.[129] Figures have been adapted from their respective sources...... 69

Scheme 3.19 PPO with ionic crafts synthesized via ATRP, adopting a rod-coil architecture. Figure adapted from T. Xu et al.[131] ...... 71

Scheme 3.20 Different molecular designs for quaternary ammonium polysulfone explored by L. Zhuang et al. (A) QA groups are directly tethered to the polysulfone backbone (e.g., BTMA PSf). (B) A hydrophobic aliphatic spacer is used to separate the QA group from the PSf backbone. (C) A hydrophobic aliphatic chain is used as ‘tail’ on the QA group. (D) Hydrophobic aliphatic chains are grafted alongside BTMA groups...... 72

List of Figures

Figure 1.1 Schematic of Proton Exchange Membrane (Left) and Alkaline (Right) Fuel Cells. Figure adapted from the Department of Energy.[1] ...... 1

Figure 1.2 Relative cost breakdown of major components in a PEMFC stack depending on annual production.[2] ...... 3

Figure 3.1 Molecular structure and stained TEM images of similarly charged (IEC ~ 1.20 mEq g-1) mono- (left) and di-quaternized (right) PEEK membranes with a DABCO headgroup. Dark regions indicate hydrophilic domains. Adapted from reference: [57] ...... 35

Figure 4.1 Representative 1H NMR spectra of a chloromethylated polysulfone. The degree of substitution here is approximately 1.09...... 85

Figure 5.1 Differential TGA plot of polymerized monomer soaking solutions...... 96

Figure 5.2 Differential TGA plot of QA PSf-226, QA sIPN [76/24/0] and QA sIPN [72/20/8] alkaline exchange membranes...... 97

Figure 5.3 Left Axis (Red Bars) - Gravimetric monomer uptake of the CMPSf films after soaking in styrene/divinylbenzene monomer solution for 24 hours. Right Axis (Blue Squares) – Theoretical IEC as calculated by the monomer mass uptake based on the maximum IEC of 2.99 mEq g-1 for unmodified QA PSf-299...... 99

Figure 5.4 Gravimetric water uptake as a function of temperature...... 100

Figure 5.5 Thickness swelling of membranes at equilibrium water uptake at different temperatures...... 101

Figure 5.6 Temperature response of water uptake as normalized against the room temperature hydrated mass...... 103

Figure 5.7 Water uptake kinetics for the AEMs...... 107

Figure 5.8 Water uptake kinetics for the AEMs plotted according to the Schott second-order kinetics model. The dashed lines represent linear fits, showing excellent agreement to the Schott model. The water uptake here is defined as a -1 ratio (gwater gpolymer ) instead of a percentage to facilitate calculation of the intrinsic rate constant...... 107

Figure 5.9 In-plane hydroxide conductivity as a function of temperature (100% RH). Dashed lines are for guiding the eye...... 112

xvi

Figure 5.10 Arrhenius plot of hydroxide conductivity at 100% RH. Dashed lines represent Arrhenius fit in the temperature range where Arrhenius scaling is observed...... 113

Figure 5.11 Alkaline stability of QA PSf-226 and QA sIPN [72/20/8] in 6 M KOH solution at 40°C...... 116

Figure 5.12 – Polarization (left axis, open markers) and power density (right axis, filled markers) for QA PSf-226 and QA sIPN [72/20/8] based MEAs at 35°C and 80°C at 100% RH. Back pressure set to 200 kPa absolute...... 118

Figure A.1 SAXS data for the semi-IPN anion exchange membranes...... 162

Figure A.2 Loss modulus (G’’) as a function of temperature for the semi-IPN membranes and QA PSf baseline. Peaks represent glass transition temperatures. Measurements performed on a TA Instruments Q800 Dynamic Mechanical Analyzer in tension mode at fixed frequency (1 Hz) and strain (1%) with a 10°C/min temperature ramp...... 164

Figure A.3 Phase contrast tapping mode AFM image of QA sIPN [90/7/3] membrane equilibrated under ambient conditions (22°C, ca. 35% RH). Image was recorded by Joseph Troderman...... 165

Figure B.1 Photographs of semi-IPN films synthesized via a thermal-initiated polyermization process. The membrane on the left is the result of polymerizing styrene and divinylbenzene in a benzyltrimethylammonium polysulfone/DMF solution and shows macroscopic heterogeneity due to poor compatibility between the polymer components. The membrane on the right is the result of polymerizing styrene and divinylbenzene in a chloromethylated polysulfone/DMF solution; the clear, homogeneous film suggests no large scale de-mixing...... 168

xvii Chapter 1

Chapter 1: Introduction

1.1 Background

Volatile pricing and rising environmental concerns associated with burning fossil fuels have raised the need for alternative energy solutions. While various renewable sources such as solar and wind are promising for large-scale, grid-level energy production, a more compact solution is needed for consumer-level devices (e.g., transportation, electronic devices, etc.). The high energy density of chemical fuels makes electrochemical energy conversion a favorable approach for tackling this problem. To this end, fuel cells, which convert chemical reduction-oxidation potentials into electricity, promise a clean and scalable alternative to traditional fossil fuels.

Figure 1.1 Schematic of Proton Exchange Membrane (Left) and Alkaline (Right) Fuel Cells. Figure adapted from the Department of Energy.[1]

1 Chapter 1

The hydrogen-powered proton exchange membrane fuel cell (PEMFCs) is the most popular low-temperature fuel cell technology on the market today, having seen use in buses, forklifts, and even commercially available consumer vehicles such as the

Toyota Mirai and the Hyundai ix35. These PEMFC devices operate by feeding in hydrogen gas, which is oxidized at the anode into protons, and oxygen gas, which is reduced in the presence of protons at the cathode to produce water (Figure 1.1). The overall net reaction, then, is hydrogen and oxygen combining to form water and electricity. A semipermeable polymer electrolyte membrane is used to separate the two electrodes and provides a means of transporting protons from the anode to the cathode. Although this technology is fairly mature, as evidenced by its industrial use and upcoming incorporation in consumer vehicles, the low pH environment resulting from the high proton concentrations severely limits materials options for device fabrication. For example, the bipolar plates must be resistant to corrosion under caustic operating conditions. More importantly, the acidic environment limits the choice of catalyst materials -- only platinum-based metals exhibit suitable stability and activity under these conditions. This dependence on precious metals presents a major financial barrier for large-scale commercialization; a recent market report by the U.S.

Department of Energy estimates that the platinum cost would represent half the total cost of a fuel cell stack at a production rate of 500,000 stacks per year (Figure 1.2).[2]

2 Chapter 1

Figure 1.2 Relative cost breakdown of major components in a PEMFC stack depending on annual production.[2]

One possible solution to this limitation is to switch from an acidic operating environment to an alkaline operating environment that enhances ORR catalysis, enabling the use of cheaper catalysts based on nickel, molybdenum, etc.[3,4] Alkaline fuel cells (AFCs) are not a new technology and have served as the primary power sources aboard several of NASA’s vehicles, including those from the Apollo and

Space Shuttle programs.[5] However, these early AFCs employed an aqueous potassium hydroxide electrolyte, which occupies more volume than the solid-state polymer electrolytes found in PEMFCs. More importantly, the dissolved potassium cations are prone to forming insoluble carbonate precipitates upon contact with CO2.

These AFCs are therefore highly susceptible to CO2 poisoning, requiring purified fuels and operation in an inert atmosphere. In light of these issues, there has been a strong push to develop Alkaline Exchange Membrane Fuel Cells (AEMFCs) which use a polymeric, solid-state Alkaline Exchange Membrane (AEM). Compared to a liquid electrolyte, these polymer membranes occupy less space and have pendant basic

3 Chapter 1 ionizable groups that cannot precipitate out given their covalent attachment to the polymer chain.

Hydrogen-powered AEMFCs operate under similar principals as their PEMFC counterparts, with the same overall chemistry albeit with different half-reactions. As a result, the primary charge carrier shifts from protons to hydroxide ions. Accordingly, this requires the development of Alkaline Exchange Membranes (AEMs) materials that (1) transport hydroxide ions with low resistive loss; (2) are mechanically and chemically robust under extended operation in alkaline conditions; (3) are electrically insulating to prevent short-circuiting; and (4) inhibit fuel crossover between the electrodes. Meeting these requirements has proven difficult, especially when compared against PEMs; current AEMs offer less performance and lower stability, resulting in AEMFCs devices that are subpar compared to the PEMFCs they are trying to replace. The proceeding section outlines the critical materials challenges that must be resolved before commercial realization of AEMFCs.

1.2 Motivation and Outline

This thesis aims to address some of the performance and stability challenges afflicting current alkaline exchange membrane materials. Chapter 2 provides background on the fundamental challenges associated with alkaline exchange membranes and the general design approaches that have been employed in addressing these challenges, and

Chapter 3 is a comprehensive literature review of AEM materials. Chapter 4 offers an overview of general synthesis protocols and characterization methods.

Chapter 5 describes a method that we developed towards dimensionally stabilizing highly-charged alkaline exchange membranes. While higher charge

4 Chapter 1 contents can result in enhanced conductivity and performance, it also leads to an increase in osmotic pressure. As a result, highly charged membranes often suffer from excessive water uptake and mechanical weakening. The approach outlined in this chapter sees the use of a poly(styrene-co-divinylbenzene) scaffold to dimensionally stabilize a linear benzyltrimethylammonium polysulfone ionomer, creating in effect a semi-interpenetrating network. The synthesis that was developed takes advantage of the similar solubility parameter between the styrene/divinylbenzene monomers and the aromatic nature of polysulfone. Hence, it can foreseeably be extended to other polyaromatic backbones.

Chapter 6 details the design, synthesis and characterization of benzyltrimethylammonium-functionalized polysulfone-graft-poly(ethylene glycol) copolymer. The introduction of hydrophilic poly(ethylene glycol) sidechains in a benzyltrimethylammonium polysulfone random copolymer induces a phase- segregated morphology characterized by hydrophilic hydroxide transport channels.

Concomitant with this structure formation is an increase in the hydroxide transport efficiency, manifested in higher in-plane conductivity and enhanced device performance.

Chapter 7 discusses collaborative efforts towards developing a fundamental understanding of phase-segregation in random copolymers via theoretically-informed coarse-grained simulations and mean-field theory development, which provide a glimpse into the compositionally-dependent morphology of these materials.

Specifically, we focus on how the stochastic arrangement of chemical identities and chain rigidity influence both the order-disorder transition and phase-separated

5 Chapter 1 structures, and discuss the implication of these results in both interpreting current empirical results and designing new materials.

6 Chapter 2

Chapter 2: Background

2.1 Hydroxide Transport in Anion Exchange Membranes

Ion transport in an electrolyte generally occurs through a combination of the following modes: (1) en masse diffusion driven by a concentration gradient of ions in the electrolyte; (2) convection from a net flow of the electrolyte; (3) or migration driven by an electrical potential gradient across the electrolyte. However, these universal modes of ion transport do not adequately capture the anomalously high mobilities of protons and hydroxide. Instead, these particular ions are uniquely transported through a structural diffusion process. Protons are known to undergo Grotthus transport in aqueous solution, wherein proton migration occurs as a result of interconversion

+ + between tri-coordinated H3O (H2O)3 and H5O2 via hydrogen bond breaking and formation at the second hydration shell (Figure 2.1). The net result can be qualitatively described as protons rapidly “hopping” across a hydrogen-bonded network of water molecules.

Figure 2.1 Illustration of Grotthus transport of a proton. A proton from hydronium molecule a is transferred upon cleavage of a hydrogen bond at the second hydration shell between water molecules b and c. Figure adapted from Agmon.[6]

7 Chapter 2

Hydroxide anions are suspected to undergo a similar Grotthus-like transport mechanism, although the fundamental mechanistic details are not fully understood.

The traditional explanation for hydroxide transport invokes direct analogies to the

Grotthus diffusion of protons, treating the process as a migration of a “proton hole”

- that results in interconversion of the hydroxide between tri-coordinated OH (H2O)3

- and H7O4 via hydrogen bond cleavage and formation at the second hydration shell.

However, this treatment, in its most basic version, does not adequately explain why hydroxide mobility is empirically found to be only 57% of proton mobility. One possible explanation offered by Agmon suggests that there is an additional O-O bond contraction during the hydroxide transfer process, contributing an additional 0.5 kcal mol-1 to the activation energy.[6]

Recent reports, however, contend that this simple explanation for hydroxide transport via symmetry-invoking analogy to proton migration is fundamentally incorrect. Instead, ab initio calculations reveal that the dominant form of the

- [7,8] hydroxide anion is a hyper-coordinated OH (H2O)4 complex. Conversion of this

- planar complex into OH (H2O)3 via hydrogen bond cleavage at the first hydration shell

-1 - (1.16 kcal mol EA) results in a weak hydrogen bond between the OH hydrogen and bulk water; this results in a bond contraction between the OH- and one of its ligating

- H2O molecules, consequently enabling proton transfer from the H2O to the OH and a net migration of the hydroxide anion. This proposed mechanism is fundamentally different than that governing proton transport, and the rate limiting step of converting the hyper-coordinated hydroxide complex into an active tri-coordinated complex offers an explanation as to the lower mobility of the hydroxide anion. Many aspects

8 Chapter 2 proposed by these calculations have been supported by experimental evidence. The hyper-coordinated OH- complex has been observed in neutron and X-ray scattering experiments [9,10], while FTIR spectroscopy data supports the existence of the weak

- hydrogen bond between the hydrogen on OH and the oxygen on H2O.

However, still other studies report that, although hyper-coordinated OH- is prominent, its population is not as dominant as believed and there exists a substantial fraction of tri-coordinated OH-. Indeed, there have been other proposed mechanisms and the exact details of hydroxide transport remain contentious.[11,12] Interestingly, a recent experimental study showed that hydroxide transport in amorphous ice no longer undergoes the anomalous transport behavior observed in liquid water, instead migrating through simple Brownian diffusion; by contrast, protons continue to diffuse through the Grotthus mechanism.[13] While the detailed transport mechanisms are not fully understood, it is generally agreed that water solvation is critical to efficient migration of the hydroxide anion.

Of course, the preceding discussion has focused exclusively on ion transport in electrolyte solution. Nevertheless, the same principles apply in polymer electrolyte membranes, which uptake a notable amount of water from dissociation and solvation of the pendant salt moieties. Hydration and swelling results in water-rich regions where bulk transport can occur in a manner analogous to that in dilute solution. In addition, ions can also be transported via direct transfer of protons or hydroxide ions between the pendant counterion headgroups. Because structural diffusion is significantly faster than other modes of ion transport, the overall conductivity of a polymer electrolyte membrane is highly dependent on its water content and relative

9 Chapter 2 humidity. However, the polymer system also imposes several constraints not present in dilute solution, including tortuosity of the water-rich regions, Coloumbic interaction between solvated ions and pendant counterions, or even electrostatic condensation of the free ions, all of which can mitigate the overall effectiveness of ion transport.[14–16]

Thus, the ionic conductivity of a polymer electrolyte material is contingent on several factors, including degree of ionization, relative humidity and morphology.

Macroscopic Conductivity

The empirical, macroscopic ionic conductivity can be described as a function of ion mobility, µ, concentration, C, and charge number, z:

� = �|�|�� where the mobility of the ion can be related to an effective diffusion coefficient, D, via the Nernst-Einstein relation:

� � � = � � = � �!!!/!" � �� ! ��

The diffusivity and activation energy represent holistic parameters that take into all modes of ion transport within the system. The conductivity, then, can be expressed as:

� �! � = � ��!!!/!" ≈ � �!!!/!" ! �� !

Within a narrow temperature window, this conductivity relationship is essentially

Arrhenius. Thus, it is possible to calculate an empirical, effective activation energy that can provide qualitative insight on ion transport efficiency. Note that this description of conductivity assumes that the polymer itself plays no significant role in the charge transport mechanics. If, however, the thermal motion of the polymer chains

10 Chapter 2 significantly influences ion transport, as is often the case for non-aqueuous ion transport, then the conductivity-temperature data typically follows a Vogel-Fulcher-

Tammann or William-Landau-Ferry relationship.[17]

2.2 Major Challenges

Current AEM materials are afflicted by two major limitations. First, hydroxide conductivities in typical AEMs are notably lower than the proton conductivities in the

PEMs they are vying to replace, leading to large resistive losses in a fuel cell device.

Second, many AEMs exhibit poor chemical stability under alkaline conditions, resulting in insufficient operating lifespans for AEMFCs.

2.2.1 Low Ionic Conductivity

The lower conductivity of AEMs compared to PEMs can be traced to two underlying reasons. First, for reasons described in the previous section, the hydroxide anion has a dilute solution mobility that is only 57% of that of the proton, leading to a fundamental kinetic limitation. Thus, to first approximation, the charge density of an

AEM must be nearly twice that of a comparable PEM material for similar overall conductivity. The ion exchange capacity (IEC), defined as the milliequivalents of charge per gram of polymer, reflects the gravimetric charge density of a polymer electrolyte. While in many cases increasing the IEC is synthetically facile, larger charge contents typically lead to high osmotic pressure and disproportionate water uptake, resulting in excessive swelling, charge dilution and diminished mechanical integrity. In light of this, there have been various attempts to limit water uptake while

11 Chapter 2 maintaining high charge concentrations via crosslinking, charge localization, composite reinforcement and hydrophobic grafts.

Figure 2.2 Illustration of nanophase-separation in Nafion. Sulfonic acid headgroups cluster to form water-rich domains for ion transport (white), while the hydrophobic backbone provides a stable matrix (grey). Figure adapted from Schmidt-Rohr and Chen.[18]

In addition to the lower mobility of the hydroxide, the ion-transport morphology of typical AEMs are not as well-defined as that of Nafion, the de facto standard PEM material. Nafion consists of a hydrophobic PTFE backbone with random fluoroether sidechains, the ends of which are capped by a sulfonic acid group.

While there are several proposed models for Nafion’s mesostructure, the general consensus is that the hydrophilic sulfonic acid groups phase-separate from the hydrophobic backbone, forming water-rich ionic clusters.[14,18] Aggregation of these clusters results in a percolating hydrophilic network throughout the polymer analogous to an efficient “highway” for proton transport (Figure 2.2). Comparatively, typical

AEMs often lack this well-defined phase separation, resulting in homogeneous

12 Chapter 2 morphologies characterized by tortuous ion transport. Consequently, there have been several recent reports focusing on effecting nanophase-separated morphologies in efforts to increase hydroxide transport efficiencies. Many of these involve functionalization of random copolymers using aliphatic spacers, large oligomer blocks, charge localization, or alkyl sidegroups to induce nanoscale ordering.

2.2.2 Poor Chemical Stability

In addition to low conductivity, many AEMs exhibit poor chemical stability. While radical-induced degradation leads to long-term degradation in PEM materials, their lifetimes are still several orders of magnitude longer than typical AEMs. Indeed, the lifespans of AEMs are often not limited by radical attack of the polymer, but instead by the innate instability of the materials in alkaline media -- the strong nucelophilic nature of hydroxide anion leads to multiple degradation pathways for both the pendant cation as well as the polymer backbone.

2.2.2.1 Cation Stability

Hydroxide attack of pendant cationic moieties leads to charge neutralization and rapid conductivity loss. The benzyltrimethylammonium (BTMA) cation, which is the most common headgroup in the literature, is susceptible to both direct SN2 nucelophilic attack at the α-carbon and E1 elimination at the α-hydrogen (Scheme 2.1). Abstraction of an α-H can also lead to ylide formation, which can further undergo Stevens or

Sommelet-Hauser rearrangements; however, DFT calculations[19] indicate that the ylide formation reaction is highly reversible, and experimental studies show that degradation via ylide rearrangement is not significant. Additionally, quaternary

13 Chapter 2 ammonium cations with β-hydrogens are also prone to Hoffman (E2) elimination.

DFT studies show that although increasing the length of alkyl functionalities can inhibit Hoffman elimination via steric interference, the energy barrier for E2 elimination of β-H is still significantly lower than SN2 attack.

Cation degradation is not solely linked to the nature of the cation itself and can be heavily influenced by external factors. Chempath et al. reported that well-solvated hydroxyl anions are less nucelophilic and lead to slower degradation.[21] Spectroscopic studies by Nunez and Hickner revealed that the stability of the BTMA headgroup varies depending on the polymer backbone[22], with the BTMA poly(arylene ether

14 Chapter 2 sulfone) showing headgroup degradation rates that were two orders of magnitudes higher than that of BTMA polystyrene. This large difference is attributed to the electron-withdrawing effect of the sulfone group, resulting in increased susceptibility of the benzylic methylene group to anionic attack. Tomoi et al. reported that using a linear aliphatic spacer to separate the cation from an aromatic backbone reduces the reactivity of the methylene carbon and thereby offers enhanced chemical stability of the cation.[23]

Enhancing the alkaline stability of pendant cations is therefore a complex issue, as the rate of degradation is contingent on a multitude of factors. Nevertheless, various approaches have been explored in the literature towards enhancing cation stability. Quaternary ammonium moieties have been investigated with different functional groups with the intent of steric and/or electronic stabilization. Others have employed different headgroups (e.g., guanidinium, phosphophonium, imidazolium) as means to mitigate hydroxide attack. A comprehensive review of these approaches can be found in Chapter 3.

2.2.2.2 Backbone Stability

It was originally believed that stability of the polymer backbone was not affected by functionalization with a cationic headgroup; i.e., that a polymer with high alkaline stability in its native state would retain its chemical resistance upon modification with tethered cations. Thus, many initial AEMs were designed by tethering cations to a base-stable polymer. However, qualitative observations suggested polymer stability becomes compromised on cationic functionalization. Despite synthesizing AEMs based on alkali-resistant backbones (e.g., polysulfone), various groups reported

15 Chapter 2 embrittlement and discoloration of the AEMs during accelerated degradation testing in alkaline media; these observations suggested cleavage of polymer backbones that were previously thought to be stable.

This phenomenon is underscored by spectroscopic analysis. Fujimoto et al. investigated the backbone stability of benzyltrimethylammonium functionalized poly(phenylene) (PP), fluorinated poly(arylene ether) (F-PAE) and non-fluorinated poly(arylene ether sulfone) (NF-PAES).[24] FTIR analysis showed phenolic alcohol formation after treatment of the NF-PAES and F-PAE materials under mild alkaline conditions (0.5 M, 80°C), indicating chain scission as a result of cleavage of the ether groups along the BTMA F-PAE and NF-PAES chains. Tensile testing of F-PAE and

NF-PAES supported this conclusion, showing rapid degradation of the mechanical properties of the F-PAE and NF-PAES membranes; both materials lost roughly 90% of their initial tensile stress and maximum strain after treatment in mild NaOH solution (0.1 and 0.5 M). In contrast, the fully aromatic PP materials showed no notable change in their mechanical properties.

16 Chapter 2

Scheme 2.2 Proposed degradation pathways for benzyltrimethylammonium bisphenol A polysulfone. The electron withdrawing effects of the pendant cation is believed to render the

(A) quaternary carbon and (B) ether groups susceptible to hydrolytic cleavage. Schematic adapted from Arges et al.[25]

Furthermore, 2D NMR studies by Arges et al. indicate hydroxide-induced polysulfone cleavage in benzyltrimethylammonium-functionalized polysulfone

(Scheme 2.2).[25] Their reports suggest that the electron withdrawing effects of pendant cations can render an otherwise stable polymer backbone susceptible to hydroxide attack. A similar study on a benylztrimethylammonium poly(phenylene oxide) material showed similar results with hydrolytic cleavage of the ether groups in the

PPO backbone.[26] Hence, it cannot be assumed that a polymer that is alkali resistant in its pure state will retain that property upon functionalization with cationic headgroups.

However, various reports suggest that it is possible to mitigate the effects of cation- induced destabilization through charge delocalization, aliphatic spacers, non-aromatic backbones, crosslinking, or other means.

17 Chapter 2

2.2.3 Carbon Dioxide Poisoning

One of the major advantages of an alkaline exchange membrane fuel cell over a liquid electrolyte alkaline fuel cell is that the polymer-bound ionic species do not form insoluble precipitates upon exposure to carbon dioxide. Nevertheless, CO2 poisoning continues to remain a concern as the hydroxide anions in alkaline exchange

2- membranes are energetically favored to convert to carbonate (CO3 ) and bicarbonate

- (HCO3 ) anions upon exposure to carbon dioxide. The dilute solution mobilities of these anions are 3-4x lower than that of the hydroxide anion owing to a combination of their larger mass and inability to undergo Grotthus-like diffusion. Accordingly, prolonged exposure to CO2 results in a loss of ionic conductivity The timescale for this conversion is function of various materials properties such as ion concentration and CO2 permeability, but near-total conversion of the free hydroxide ions have been reported to occur within an hour for several AEMs upon exposure to air.[27–29] Thus, care must be taken to operate AEMFCs under CO2 free conditions with purified reagents, or, alternatively, at elevated temperatures where CO2 solubility is limited.

2.3 Design Approaches

The need for enhanced stability and ion transport kinetics has led to the rapid research and development of a variety of different anion exchange membrane materials. Anion exchange membranes typically share a similar fundamental molecular design concept

– at a very basic level, modern AEM materials can be characterized by cationic functionalities covalently tethered to a polymer backbone. As a result, the development of a majority of these novel materials typically follows one of two predominant design paradigms.

18 Chapter 2

These paradigms can be described as either structure-property-driven or chemistry-driven. While both methods rely on chemical modification for enhanced performance and stability, their underlying approaches to solving the problem are intrinsically different. Materials that have been synthesized following the structure- property-driven approach typically follow a design rationale of tuning the polymer morphology towards enhanced macroscopic properties. Interestingly, while the discussion of ordered mesostructures in soft materials is often tied to di-block and triblock copolymers, the majority of the AEM materials developed following this paradigm tend to be random copolymers, whose fundamental phase-behaviors are less established in the literature. By contrast, the chemistry-driven approach tends to focus on pursuing different counterions and polymer backbone chemistries that may lead to better performance and stability by modifying parameters such as basicity of the pendant ionic group or taking advantage of charge delocalization effects.

19 Chapter 3

Chapter 3: Literature Review

The following literature review is divided into three major sections and subsections thereof. The first section discusses developments in homogeneous anion exchange membranes based on decoration of a stable polymer backbone with cationic headgroups. The second section details inorganic and organic composites in which the mechanical properties of a conductive ionomer are enhanced by a second, inert component. Finally, we conclude the literature review with an overview of phase- segregated membranes, where the molecular design principles are tailored to enhance macroscopic properties by tuning the nanostructure of the polymer membrane.

Concomitant with the review of nanostructured membranes is a brief overview on theoretical developments of phase-separation in copolymers and its influence on polymer electrolyte design.

3.1 Homogeneous Membranes

Homogeneous AEM materials can essentially be described as polymer chains stochastically decorated by pendant cationic functionalities. These materials are often not saturated with cationic functionalities, as the high charge content would lead to excessive water uptake and extremely poor dimensional stability. Thus, the majority of these materials are in essence random copolymers.

The design approach in these materials is based on modifying the chemistry of the pendant ion or the backbone towards better ionic conductivity, mechanical properties, and/or chemical stability. Given the ubiquity of the quaternary ammonium

20 Chapter 3 counterion in the literature, the first part of this section will focus exclusively on materials developed using this headgroup and its attachment onto both (1) aromatic and (2) aliphatic polymer backbones. The second part of this section will focus on different pendant cations such as phosphonium, imidazolium, and guanidinium.

3.1.1 Quaternary Ammonium Aromatic Random Copolymers

Aromatic polymers have been a popular choice as anion exchange membrane materials for two primary reasons. First, the aromatic groups are easily modified with halides or alkylhalides to facilitate chemical modification, allowing a variety of different functional groups to be tethered directly to the backbone. Second, the aromatic polymers that are typically used exhibit excellent alkaline resistance and mechanical robustness in their native states, though, as mentioned previously, this alkali resistance can become compromised on cation functionalization.

Quaternary ammonium cations were initially investigated due to their stability advantages over other cations at the time (e.g., alkylphosphonium). However, they have gained traction owing in part to their simple synthesis – quaternization of a tertiary amine occurs spontaneously in the presence of alkyl halide moieties via the

Menshutkin reaction.

3.1.1.1 Polysulfone

Polysulfones are one of the most popular classes of polyaromatic backbone materials for alkaline exchange membranes, having seen extensive use with a host of different sidegroups and functionalities. As with other polyaromatics, its popularity stems from its excellent mechanical, thermal and chemical robustness – neat polysulfones have

21 Chapter 3 elastic moduli on the order of 1-10 GPa, Tg around 150-200°C and high hydrolytic stability and alkaline resistance. Moreover, their aromatic groups are easily functionalized with various moieties. It comes as no surprise, then, that quaternary ammonium-functionalized polysulfones are one of the widest class of AEM materials in the literature.

Bisphenol A Polysulfone

First reported by Quellmalz and Zschocke for electrodialysis purposes[30], benzyltrimethylammonium bisphenol A polysulfone (BTMA PSf, Scheme 3.1) has since been prepared and characterized by several different research groups and, as will be apparent in this review, has often been used as a performance and stability benchmark. These materials are synthesized in a two-step protocol: the bisphenol A aromatic rings are first functionalized with labile chloromethyl groups that are then quaternized with trimethylamine via the Menshutkin reaction.

CH3 O O O S

CH3 O n H2C + N (CH3)3 OH-

Scheme 3.1 Benzyltrimethylammonium bisphenol A polysulfone (BTMA PSf).

The room temperature hydroxide conductivity of BTMA PSf have ranges from 1-50 mS cm-1 range and is largely dependent on the IEC, with an empirical percolation

22 Chapter 3

-1 [31–34] threshold observed around an IEC of 1.06 mEq g . In comparison to other

AEMs, BTMA PSf typically suffers from higher water uptake at similar IECs. This

- water uptake becomes disproportionate at higher IECs (~300% at 2.2 mEq g 1) , resulting in excessive swelling and extremely poor mechanical stability. Furthermore, as discussed in the section on alkaline stability, the BTMA headgroups are susceptible to nucleophilic attack by hydroxide anions. Crosslinking with diamines has been employed as a strategy for enhancing both the chemical stability of the cation and the dimensional stability of the AEM. Komkova et al. showed that the stability of tetramethylalkyldiammonium cations was dependent the length of the linear alkyl spacer.[35] Specifically, longer aliphatic spacers tended to provide better resistance against Hoffman elimination via increased electron density at the β-hydrogen.

However, longer spacers and bulkier alkyl sidegroups on the ammonium cation can also result in steric interference with hydroxide transport and lower ionic conductivity.[34] L. Zhuang et al. used a tertiary amine to introduce quaternary ammonium crosslinks between different polysulfone backbones. These membranes were exceptionally stable against dimensional swelling, exhibiting only a 3% swelling degree at 90°C.[36]

Other Polysulfones

In addition to bisphenol A polysulfone, the benzyltrimethylammonium (BTMA) headgroup has been used in other polysulfone systems in efforts to optimize material properties (Scheme 3.2). J. Wang et al. prepared a series of BTMA poly(phenyl

-1 sulfone)s (PPSf) with IECs ranging from 1.62 to 2.89 mEq g and with respective

23 Chapter 3

[37] room-temperature conductivities ranging from 16 mS cm-1 to 65 mS cm-1 .

-1 However, samples with IECs greater than 2.29 mEq g exhibited excessive water uptake at higher temperatures, losing mechanical integrity upon heating. All samples exhibited good alkaline stability with negligible changes in membrane conductivity after exposure to 4 M NaOH solution at 20°C for 48 h. Expanding on this study, the authors prepared BTMA-functionalized poly(phenyl sulfone)s where the diphenylsulfone groups are partially fluorinated.[38] Compared to BTMA PSf and

BTMA PPSf, the additional hydrophobic fluorine groups resulted in significantly reduced water uptake at similar IECs, allowing dimensionally stable membranes to be developed at higher than typical charge contents. Titrated IECs ranged from 1.47 to

-1 2.88 mEq g with corresponding gravimetric water uptake between 13 and 179% and hydroxide conductivity between 15 and 84 mS cm-1 at 20°C.

The effect of fluorination was also studied by Kohl et al. using a BTMA bisphenol A polysulfone analog in which the methyl groups in the BPA portion were replaced by trifluoromethyl groups.[39] The prepared membranes had degrees of substitution of 1.24 and 1.5 quaternary ammonium groups per repeat unit. The presence of the hydrophobic fluorine moieties reduced water uptake compared to unfluorinated BTMA PSf, with room-temperature water uptake measuring between 50 and 100%. Hydroxide conductivity was relatively low, however, measuring only ~10 mS cm-1 at room temperature despite the high charge functionalization.

24 Chapter 3

Scheme 3.2 Benzyltrimethylammonium-functionalized (A) bisphenol A polysulfone, (B) poly(phenyl sulfone), (C) poly(pthalazinon ether sulfone ketone), and (D) polyethersulfone

Cardo.

Other BTMA-functionalized polysulfone backbones include poly(phthalazinon ether sulfone ketone) (PPESK) and polyethersulfone Cardo (PES-C).[40,41] BTMA

PPESK was shown to exhibit a hydroxide conductivity starting at 5.2 mS cm-1 at 30°C and reaching up to 8.4 mS cm-1 at 80°C, but was subject to cation degradation at higher temperatures. A BTMA PES-C membrane (IEC=1.25 mEq g-1) showed a high conductivity of 40 mS cm-1 when immersed in a 1 M NaOH solution. However, immersion of the membranes in 2 M NaOH solutions resulted in the membranes turning white, prompting stability concerns.

Scheme 3.3 Bisphenol A polysulfone with benzyltrimethylammonium sidechains (PAES-Q).

Adapted from X. Li et al. [42]

25 Chapter 3

Whereas the aforementioned studies involved direct functionalization of the aromatic rings furthest removed from the sulfone functionality, X. Li et al.[42] prepared poly(phenyl sulfone)s with quaternary ammonium functionalities introduced near the aromatic rings directly adjacent to the sulfone group (PAES-Q, Scheme 3.3). The synthesis involved attachment, chloromethylation, and quaternization of phenyl sidegroups to the sulfone-adjacent rings. Hence, the benzyltrimethylamonium moieties are tethered to the end of an aromatic sidegroup of the polymer backbone as opposed to being a direct part of the main chain. (See figure). This formulation was compared to a conventional BTMA PPSf material, where the ammonium cations are directly tethered (via methylene group) to the aromatic rings of the biphenol portion of the polymer chain. The side-chain, sulfone-adjacent structure showed much higher hydroxide conductivity, lower swelling, and enhanced dimensional and chemical

-1 stability. For example, at 25°C, a side-chain material with an IEC of 1.49 mEq g exhibited a gravimetric water uptake around 37% and a hydroxide conductivity around

-1 22 mS cm-1 . In comparison, a BTMA PPSf benchmark with an IEC of 1.30 mEq g had around 91% water uptake and 10 mS cm-1 hydroxide conductivity. More importantly, when immersed in 1 M NaOH at 60°C, the main-chain approach resulted in mechanical failure of the film after 14 days. The side-chain material, on the other hand, remained mechanically robust and inly lost 25% of its initial conductivity after

30 days of immersion.

The authors also prepared self-crosslinked versions of the same material, taking advantage of the reactivity of the chloromethylated phenyl sidegroups with the aromatic rings of the biphenol portion of the polymer.[43] The remaining uncrosslinked

26 Chapter 3 chloromethyl groups were converted to quaternary ammonium moieities. The

-1 crosslinked materials had IECs between 1.33 and 1.64 mEq g and exhibited hydroxide conductivities between 44 and 58 mS cm-1 at 80°C; in contrast, the linear benchmark materials had hydroxide conductivities between 63 and 93 mS cm-1 at the same conditions. Despite the lower performance, the crosslinked materials exhibited better mechanical properties and much higher alkaline stability. Whereas the uncrosslinked material lost nearly half of its hydroxide conductivity after immersion for 28 days in a 60°C 1 M NaOH bath, the crosslinked materials showed negligible changes in conductivity, suggesting enhanced cation stability.

Jingling Yan and Michael Hickner[44] devised a synthesis protocol for selective ionization of the polysulfone backbone via copolymerization of tetramethyl bisphenol

A (tBPA) and biphenol (BPO) comonomers with phenyl sulfone. Bromination of this random copolymer resulted in reactive methylbromide groups only along the tetramethyl bisphenol A polysulfone portions of the backbone; the phenyl rings forming the biphenol comonomers remained unmodified. Thus, varying the concentration of tBPA to BPO allows control of ion content and distribution along the polymer chain – higher BPO content results in more hydrophobic blocks stochastically distributed along the polymer backbone, while higher tBPA content results in the opposite. As expected, the introduction of more hydrophobic biphenol content resulted in both lower water uptake and decreased ionic conductivity. Interestingly, SAXS measurements showed no local heterogeneity resulting from the stochastic distribution of non-ionic and ionic moieties along the polymer backbone. This might explain the results, which showed that that the performance of the membranes is primarily tied to

27 Chapter 3 water uptake and less so to any morphological significance from the homopolymer or copolymer nature of the polymer chain.

Other strategies with quaternary ammonium polysulfone AEMs have involved increasing the spatial charge density of the ammonium headgroups (Scheme 3.4). The high charge localization is believed to naturally lead to separation of the polymer chain into conductive and non-conductive regions, facilitating clustering of ionic domains. Guiver et al.[45] prepared bisphenol A polysulfones with pendant bis(phenyltrimethylammonium) (PTMA) functionalities. These materials (IEC=1.01 to

1.74 mEq g-1) showed higher hydroxide conductivity (15 to 38 mS cm-1, 20°C) and lower water uptake (8.7 to 18.4%) compared to mono-quaternized BTMA PSf

-1 (IEC=1.82 mEq g , 35 mS cm-1 and 40.5% WU at 20°C). Interestingly, the water uptake of PTMA PSf showed low temperature sensitivity, with negligible changes between 20°C and 60°C, resulting in lower dimensional swelling and a more mechanically robust material. This suppressed water uptake is attributed to the pendant PTMA groups, but the report offers no further discussion or conjecture on this rationale.

28 Chapter 3

A B

Scheme 3.4 Molecular designs with high concentration of ionic species. (A) PTMA PSf by

Guiver et al., (B) Tetrafunctional PPSf by Choi et al., and (C) PAESK with tetrafunctional fluorenyl groups by Watanabe et al. Figure adapted from their respective sources.

The use of tetra-functional groups in polysulfones further expands on the idea of spatially concentrating ionic headgroups by allowing up to four quaternary ammonium headgroups to be in placed close proximity, separated from the non-ionic portions of the polysulfone backbone. Consequently, at similar charge concentrations, poly(arylene ether sulfones) AEMs with tetra(quaternary ammonium) functionalities were found to exhibit higher higher conductivity and lower water uptake than their respective mono-quaternized BTMA PSf.[46] Specifically, AEMs with IECs of 0.57,

-1 0.84, and 1.20 mEq g exhibited high conductivities of 14.1, 21.7, and 36.7 mS cm-1 at

30°C. Incorporation into an MEA showed a peak power density of 77 mW/cm2 and

2 -1 maximum current density of 120 mA/cm for the 0.84 mEq g IEC material at 50°C.

In a separate study by Watanabe et al., poly(arylene ether sulfone ketone)s with tetra-functional fluorenyl groups were quaternized with trimethylamine to yield

-1 [47] AEMs with IECs between 0.68-2.54 mEq g . These membranes showed high

29 Chapter 3 conductivities between 10 and 57 mS cm-1 and low water uptakes between 5 and 140% at room temperature. Stability data were not reported. Although the enhanced performances were attributed to facilitated phase-separation due to localization of charges, neither study reported morphological characterization.

3.1.1.2 Poly(phenylene oxide)

Poly(2,6-dimethyl-1,4-phenylene oxide) (PPO) AEMs represent another class of popular aromatic backbones. As in the majority of polysulfone-based membranes, quaternary ammonium cations are typically introduced onto the PPO backbone by first forming a reactive intermediate, followed by quaternization with a tertiary amine.

Owing to the lack of an electron-withdrawing sulfone group, QA PPO membranes are suspected to have better alkali stability compared to QA PSf materials. A comparison between BTMA-functionalized poly(phenyl sulfone) and PPO showed rapid degradation of the PPSf ether group under alkaline conditions whereas the ether bonds in BTMA PPO showed no significant change. In addition, the BTMA cation in the

PPO lasted up to 10x longer than those tethered to PAS under alkali treatment.[48]

Xu et al. have studied several PPO-based AEMs. Their initial report investigated the quaternization of a 100% bromomethylated PPO (BPPO) homopolymer AEMs with a mixture of pyridine and trimethylamine.[49] Because the high charge content led to large water uptake and subsequent poor film formation, the

BPPO was first treated with ammonia to introduce covalent crosslinks prior to quaternization. The crosslinked pyridinium/quaternary ammonium PPO (IEC=1.6 mEq g-1) had a water uptake ~60% and an area resistance ~5 Ω cm-2. To exert more control over the functionalization the PPO, the authors developed a synthetic approach

30 Chapter 3 for controlled bromination of the PPO backbone.[50] Depending on the reaction temperature, bromination can occur at the benzyl and/or aryl position of the PPO backbone. These alkylbromide groups served as the reactive sites for quaternization with trimethylamine. The properties of the resulting AEM depended largely on the location of the quaternary ammonium group, with benzyl-subsitution resulting in increased water content and conductivity than aryl-substitution. In light of this, the authors developed an alternative method for selective quaternization of the benzyl position of the PPO via chloroacetylation of the phenyl ring.[51] A QA PPO membrane

-1 prepared in this manner to an IEC of 1.15 mEq g showed gravimetric water uptake between 40 and 60% and an area resistance of 0.2 Ohm/cm2.

In ensuing studies, Xu et al. prepared membranes based on CPPO and BPPO blends.[52,53] The lower reactivity of the chloroacetyl group with trimethylamine caused incomplete quaternization at higher CPPO contents, resulting in higher mechanical stability at the cost of lower conductivity. The unreacted chloroacetyl groups were found to undergo Friedels-Craft reaction with the phenyl hydrogen of

BPPO at elevated temperatures, resulting in chemical crosslinking. The crosslinked membranes showed enhanced mechanical properties with around a 3x increase in tensile strength from 34 MPa to 87 MPa but maintained high hydroxide conductivity

(32 mS cm-1 at room temperature).

Gopi et al.[54] presented a synthetic protocol for chloromethylation of PPO at the aryl position of the aromatic ring, followed with quaternization with trimethylamine. Because the ortho-methyl groups are unaffected, this method selectively introduces benzyltrimethylammonium groups at the meta position, further

31 Chapter 3 removing the cation from the electronegative ether linkages. The material showed

-1 relatively high water uptake around 30% despite a low IEC of 0.70 mEq g .

Nevertheless, the membrane exhibited suitable performance, with a chloride conductivity of 4.3 mS cm-1 and an exceptionally high peak power density of 111 mW/cm2 at 30°C.

Scheme 3.5 Click reaction tethering 1,2,3-triazole to PPO. Figure adapted from Binder et al.[55]

In a different approach, Binder et al. used brominated PPO as a basis for incorporating quaternary ammonium functionalities via click chemistry of 1,2,3- triazoles (Scheme 3.5).[55] These materials showed exceptionally high hydroxide

-1 conductivities up to 62 mS cm-1 at 20°C at an IEC of 1.80 mEq g . The diffusion of water, as measured by pulse gradient magnetic field NMR, was found to be two to three times slower when the triazole groups were present; at the same time, however, the hydroxide diffusion coefficient was roughly an order of magnitude higher. The high performance was thus attributed to the ability of the triazole groups to provide sites for hydrogen bonding with water, thereby creating a continuous hydrogen-

32 Chapter 3 bonded network that facilitates Grotthus-like diffusion of hydroxide anions. This is underscored by SAXS measurements that showed no indication of nanophase- separation, suggesting that a tuned polymer morphology was not responsible for the high performance. The ammonium functionality in these materials were nevertheless subject to nucelophilic attack as there was a precipitous loss of IEC after immersion in

1 M NaOH at 80°C.

3.1.1.3 Poly(ether ether ketone)

G. He et al. have performed several studies on quaternized poly(ether ether ketone)

(PEEK) and were specifically interested in the high density of electron-donating ether groups. Their initial investigation saw the quaternization of chloromethylated PEEK with trimethylamine, resulting in benzyltrimethylammonium functionalities. [56] The degree of quaternization was varied to produce materials with IECs between 0.43 and

1.35 mEq g-1, with hydroxide conductivities that scaled proportionally between 5.6 and 17 mS cm-1 (30°C). At similar IECs, these values are higher than in other polyaromatic materials quaternized with trimethylamine, whose conductivities typically range between 1 and 10 mS cm-1 . The enhanced performance is ascribed to the electron-donating ether bonds in the PEEK backbone, resulting in increased electron-density and basicity of the quaternary ammonium headgroup. However, these materials exhibited large water uptake in excess of 240% for IEC=1.35 mEq g-1.

A structure-property investigation of quaternized PEEK backbones was performed by comparing the performance and morphology of mono- and di- quaternized PEEK (Figure 3.1).[57] Specifically, chloromethylated PEEK was quaternized with 1,4,-diazabicyclo[2,2,2]octane (DABCO), a diamine. Because both

33 Chapter 3 tertiary amines are available for quaternization, crosslinking was avoided by reacting the chloromethylated polysulfone with a large excess of DABCO to create mono- quaternized PEEK. An additional alkylation step was then performed to create di- quaternized PEEK. The high cation concentration in the di-quaternized materials was suspected to lead to stronger hydrophobic-hydrophilic separation, resulting in better phase segregation. TEM imaging showed that both materials exhibited nanophase- segregation into hydrophobic and hydrophilic domains; however, the di-quaternized membranes had significantly larger characteristic domain sizes than their mono- quaternized equivalents. As a result, the di-quaternized membranes exhibited higher hydroxide conductivity (18 to 35 mS cm-1 vs. 3 to 12 mS cm-1) and effective hydroxide mobility (ca. 3x104 vs. 1x104 cm2 s-1 V-1) than the mono-quaternized materials.

However, both materials showed poor alkaline stability, with >40% loss in tensile strength and conductivity after 120 h aging in 2 M KOH at 60°C, indicating degradation of both the quaternized DABCO moieties as well as the PEEK backbone.

In a separate study, the authors investigated using DABCO as both a quaternization and crosslinking agent.[58] The results show that at 25°C, a crosslinked

-1 DABCO PEEK membrane with an IEC~1.4 mEq g had around 200% gravimetric water uptake value and a 33 mS cm-1 hydroxide conductivity, a marked improvement over the BTMA PEEK membrane. However, despite the crosslinking, long-term testing indicated poor hydroxide stability of both the PEEK backbone as well as the

DABCO cations, with >30% losses in both tensile strength and hydroxide conductivity after aging in 2 M KOH at 60°C for 120 h.

34 Chapter 3

Figure 3.1 Molecular structure and stained TEM images of similarly charged (IEC ~ 1.20 mEq g-1) mono- (left) and di-quaternized (right) PEEK membranes with a DABCO headgroup.

Dark regions indicate hydrophilic domains. Adapted from reference: [57]

3.1.2 Aliphatic Copolymers

There have been a number of aliphatic materials investigated as polymer backbones for AEM materials. Unlike polyaromatic AEMs, whose syntheses typically involve halogenation and Menshutkin quaternization, the synthetic approach for aliphatic

AEMs can differ dramatically from material to material.

Moreover, aliphatic polymers generally exhibit weaker mechanical properties and greater plastic deformation compared to aromatic polymers. As such, many aliphatic polymer electrolytes exhibit higher water uptake and swelling and must be dimensionally stabilized through the use of crosslinks. This section discusses several examples of aliphatic AEMs, and is organized into the major synthetic methods used to fabricate them.

35 Chapter 3

3.1.2.1 Covalent Crosslinking

Xiong et al.[59] prepared crosslinked, quaternary ammonium poly(vinyl alcohol) (PVA) membranes. Specifically, (2,3-epoxypropyl)trimethylammonium groups were grafted onto the PVA matrix in the presence of potassium hydroxide. The quaternized PVA was then crosslinked with glutaraldehyde during solution casting.

Although increasing the crosslink concentration resulted in a significant decrease in water uptake, it also decreased the hydroxide conductivity, presumably by hindering the formation of large ion transport channels. For example, increasing the crosslinking degree from 3.91% to 15.64% decreased both the conductivity from ~7 mS cm-1 to

~2.5 mS cm-1 and the water uptake from ~240% to ~90% at 30°C.

Herring et al.[60] prepared quaternary ammonium anion exchange membranes by reacting the methylchloride groups on chlorinated polypropylene chains with tertiary amine groups on branched poly(ethyleneimine), forming quaternary ammonium functionalities and a highly crosslinked structure (Scheme 3.6).

Interestingly, while the polypropylene-polyethyleneimine materials exhibit nearly

70% water uptake at 60°C, they showed no notable dimensional swelling, suggesting that the absorbed water is preferentially localized to the free volume of the high- crosslinked network. However, the anionic conductivities of these materials were extremely low, with chloride conductivities reaching only 0.293 mS cm-1 even at 90°C and 95% RH. Moreover, these materials were highly unstable in alkaline solution, as attempts to perform hydroxide exchange resulted in mechanical failure of the films.

The low performance and stability renders these materials unsuitable as AEMs.

36 Chapter 3

Scheme 3.6 Highly-crosslinked quaternary ammonium polypropylene-polyethyleneimine.

Figure adapted from Herring et al. [60]

3.1.2.1 Chitosan Modification

Chitosan has been proposed as an AEM material owing to its low toxicity and natural abundance (Scheme 3.7). Quaternized-chitosan membranes crosslinked by glutaraldehyde were synthesized by Y. Wan et al.[61] The hydroxide conductivity depended on the degree of quaternization, measuring between 4.8 and 7.5 mS cm-1 at room temperature. Degrees of quaternization above 35% led to poor mechanical properties as a result of excessive swelling, despite the incorporation of glutaraldehyde crosslinks.

Scheme 3.7 Quaternary ammoniun-functionalized chitosan (non-crosslinked). Adapted from

Y. Wan et al.[62]

In a followup study, Y. Wan et al.[62] investigated the use of an alternative crosslinker, ethylene glycol diglycidyl ether (EGDE), in quaternized chitosan

37 Chapter 3 membranes which limited water uptake to ~35% at a degree of quaternization of 71%.

The ionic conductivity was again found to be between 1 to 10 mS cm-1 at room temperature, with higher degrees of quaternization yielding higher conductivities. Q.

Liu et al. prepared AEMs from quaternized poly(vinyl alcohol) crosslinked with quaternized chitosan via glutaraldehyde, with hydroxide conductivities between 3 and

[63] 5 mS cm-1 at room temperature.

3.1.2.3 Radiation Grafting

Varcoe et al. were among the first to synthesize solid-state polymer electrolyte materials with pendant cations tailored specifically for alkaline fuel cells. Their initial report employed radiation-grafting to introduce quaternary ammonium groups onto poly(vinylidene fluoride) and poly(tetrafluoroethylene-co-hexafluoropropylene)

(FEP).[64] While the PVDF membranes physically degraded beyond practical use upon amination and hydroxide exchange, the FEP membranes were relatively stable, with hydroxide conductivities ranging from 10 mS cm-1 at 20°C to 35 mS cm-1 at 80°C under fully hydrated conditions. Nevertheless, due to radation-induced degradation of

FEP, the QA FEP membranes were not mechanically robust enough to incorporate into an MEA.

As a potential solution, Varcoe and Slade replaced the FEP backbone with radiation-resistant polymer -- poly(ethylene-co-tetrafluoroethylene) (ETFE).[65] The

QA ETFE materials showed similar conductivities as QA FEP, but were able to be incorporated into an MEA capable of producing 110 mW/cm2 peak power density at

60°C under H2/O2 flows.

38 Chapter 3

3.1.2.4 Ring Opening Metathesis Polymerization (ROMP)

Coates et al. have investigated the use of Ring Opening Metathesis Polymerization

(ROMP) to synthesize several different aliphatic AEMs. In their first report, the research group copolymerized tetraalkylammonium-norbornene with dicyclopentadiene (Scheme 3.8).[66] The highest performing material exhibited 18 mS cm-1 hydroxide conductivity at 20°C but relatively poor mechanical properties (2.3

MPa maximum tensile stress).

Scheme 3.8 Ring Opening Metathesis Polymerization of tetraalkylammmonium-norbornene with dicyclopentadiene. Scheme adapted from Coates et al. [66]

The low hydroxide conductivities of these norborene-based AEMs prompted the group to prepare a series of tetraalkylammonium-functionalized polyethylenes

(QA PE).[67] These materials exhibited relatively high gravimetric water uptake between 100 and 130%, but nonetheless showed good mechanical properties, able to reach ~130-170% elongation and ~6-9 MPa tensile stress prior to breaking. TEM

39 Chapter 3 characterization showed no evidence of microphase separation, which the authors attribute to the random distribution of the hydrophilic tetralkalylammonium ions.

Nevertheless, the materials demonstrated high hydroxide conductivites between 40 and 50 mS cm-1 at 20°C. The main advantage of these QA PE materials is that while they are insoluble in pure water and methanol, they exhibit excellent solubility in other alcohols and aqueous alcohol solutions. This solvent processability allows QA PE to be used as both an ionomer binder as well as a membrane separator.

The ROMP protocol for QA PE can be easily modified to introduce covalent crosslinking.[68] While this removes the solvent processability afforded by linear QA

PE, it enhances the dimensional stability of the membrane. As such, crosslinked QA

PE membranes could be synthesized with a much higher IEC (2.3 mEq g-1) compared to the linear alternative (1.3-1.5 mEq g-1). Consequently, these crosslinked QA PE

AEMs exhibited exceptionally high hydroxide conductivities between 68.7 mS cm-1 at

20°C and 111 mS cm-1 at 90°C, approaching ~80% of the proton conductivity of

Nafion-112 in the same temperature window.

3.1.3 Alternative Headgroups (Non-Quaternary Ammonium)

The poor alkali stability and low performance of quaternary ammonium headgroups has prompted investigations into several alternative cations. The advantages in the performance and stability of these alternative headgroups, however, are not always clear. For example, several studies contend that the basicity of the quaternary ammonium moieties is not high enough, resulting in incomplete dissociation of hydroxide anions and therefore lower conductivity.[69,70] In light of this rationale, the higher performance of several alternative headgroups has been attributed, often times

40 Chapter 3 in passing, to their presumably higher basicity. Yet a recent discussion by Varcoe et al.[20] highlights the fact that discussions about the low basicity of quaternary ammonium hydroxide stems from incorrect equilibria analysis; instead, they contend that quaternary ammonium hydroxides are usually fully dissociated under hydrated conditions, rendering the effects of headgroup basicity a moot point. Moreover, the effective stability of the headgroup is often contingent on several factors external to the identity of the cation itself, and attributing any enhanced stability solely to the chemical identity of the headgroup can often lead to false conclusions. Finally, while many headgroups are initially reported to offer significantly enhanced alkali stability, follow-up studies often reveal evidence to the contrary – this is especially apparent in the discussion on the stability of imidazolium headgroups.

3.1.3.1 Phosphonium

Y. Yan et al. were the first to report incorporation of a stable quaternary phosphonium as the pendant cation in AEMs by quaternizing chloromethylated bisphenol A polysulfone with tris(2,4,6-trimethoxyphenyl)phosphine (TTMPP).[71,72] The methoxyphenyl sidegroups were chosen to impart alkaline stability through (1) steric hindrance against hydroxide attack, and (2) electronic stabilization of the phosphonium cation via charge conjugation and electron donating effects of the methoxy functionalities. The importance of sidegroup stabilization of phosphonium cations cannot be underestimated, as tetraalkylphosphonium salts are prone to ylide formation and olfenination under basic conditions via the Wittig reaction.[73,74] Indeed,

Arges et al. reported that attempts to hydroxide-exchange benzyltriphenylphosphonium polysulfone failed due to rapid degradation of the

41 Chapter 3 phosphonium cation.[32] In contrast, an optimized TTMPP-quaternized material with

-1 an IEC of 1.18 mEq g exhibited a hydroxide conductivity of 45 mS cm-1 at 20°C, which is ~2.4 time greater than that of a quaternary ammonium polysulfone analog at similar charge content. The enhanced performance was attributed to the higher basicity of the phosphonium cation, a result of the electron donating potential of the trimethoxyphenyl sidegroups. Incorporation of the phosphonium material into an

MEA, as both the membrane and ionomer binder, resulted in a high peak power

2 2 density of 258 mW/cm and low internal resistance of 0.210 OHM cm under H2/O2 flows. These materials are also subject to self-crosslinking at elevated temperatures

(80°C) via covalent bonding between unconverted methylchloride groups of chloromethylated polysulfone with the trimethoxyphenyl sidegroups of the phosphonium cation.[75] This self-crosslinking effect was shown to reduce the water swelling by a factor of 5-10 while maintaining high hydroxide conductivity.

Building on their previous work that suggested that poly(ether ether ketone) may enhance the basicity of its pendant cations[56], G. He et al.[76] grafted tris(2,4,6- trimethoxyphenyl)phosphine along a PEEK backbone, producing AEMs with IECs in

-1 the 0.89 to 1.19 mEq g range. Compared to polysulfone membranes, the PEEK materials exhibited lower water uptake and swelling at the same IEC, an effect that the authors attributed to enhanced van der Waals interaction resulting from PEEK’s higher electron density. Compared with other headgroups, the phosphonium-functionalized

PEEK (IEC=1.19 mEq g-1) had a hydroxide conductivity (61 mS cm-1) that was nearly five times greater than that of BTMA PEEK (13 mS cm-1) at 20°C. Moreover, while imidazolium-functionalized PEEK shows similar conductivities, they also have nearly

42 Chapter 3 twice the IEC and therefore have significantly higher water uptake.[77] Alkali stability tests in 1 M KOH at 60°C showed virtually no change in hydroxide conductivity after

48 hours, underscoring the excellent stability of the quaternary phosphonium group initially reported by Y. Yan et al.

In another study involving tris(2,4,6-trimethoxyphenyl)phosphine, T. Xu et al.[78] tethered this group onto a poly(2,6-dimethyl-1,4-phenylene oxide) backbone,

-1 -1 creating materials with IECs between 0.49 mEq g and 1.40 mEq g . Tapping mode

AFM imaging revealed that increasing the degree of quaternization, and therefore the

IEC, resulted in larger and more interconnected hydrophilic ion-transport domains.

Consequently, the hydroxide conductivity at 80°C was directly proportional to the

IEC, with the highest IEC material exhibiting ca. 100 mS cm-1 while the lowest IEC material exhibiting ca. 40 mS cm-1 . These materials demonstrated moderate alkaline stability, losing roughly half of their initial conductivity following 200 h immersion in

1 M NaOH solution at 60°C. The loss in conductivity indicates degradation of the

TTMPP-based phosphonium headgroup, an effect that was not seen in their use with polysulfone[71,72,75] and poly(ether ether ketone)[76] backbones.

Scheme 3.9 Polyethylene functionalized with tetrakis(dialkylamino)phosphonium. Figure adapted from Coates et al.[79]

43 Chapter 3

Taking advantage of their experience in ROMP of quaternary ammonium

AEMs[67,80,81], Coates et al. used a similar approach to synthesize linear polyethylene with pendant tetrakis(dialkylamino)phosphonium cations (Scheme 3.9),[79] producing an optimized membrane with an IEC of 0.67 mEq g-1, 52% water uptake, and 22 mS cm-1 hydroxide conductivity at 22°C. This material showed remarkable alkaline stability, with negligible changes in hydroxide conductivity after 138 days in 15 M

KOH solution at 22°C. The particularly high stability was attributed to charge delocalization of the phosphonium cation.

3.1.3.2 Imidazolium

Resonance stabilization and charge delocalization afforded by the imidazolium ring has made the cation attractive as a potentially alkaline resistant headgroup. However, experimental reports have led to contradictory conclusions as to its actual stability under basic conditions. While initial studies suggested excellent alkali stability with virtually no degradation, more recent reports contend that the imidazolium cation is actually less stable than the quaternary ammonium it is trying to replace.[82] Still others report that the stability of the group can be tuned by choice of functional group.[83,84]

In any case, it is readily apparent that the stability advantage of imidazolium over quaternary ammonium is not as clear as once believed.

Fang et al.[85] synthesized a linear acrylic polymer with pendant imidazolium cations via copolymerization of 1-allyl-3-methylimidazolium chloride with methyl methacrylate and butyl methacrylate. The prepared membranes had relatively low

-1 imidazolium content and IECs between 0.15 and 0.2 mEq g ; however, they showed good conductivities ranging from 14 to 33 mS cm-1 at 30°C. Moreover, accelerated

44 Chapter 3 degradation testing of the membranes by immersing them in 6 M NaOH solution at

60°C showed no significant change in mass after 120 hours and ~10% decrease in conductivity, suggesting suitable alkaline resistance of both the imidazolium cations and the aliphatic backbone.

F. Yan et al.[86] copolymerized 1-vinyl-3-methylimidazolium with styrene, acrylonitrile, and divinylbenzene to create crosslinked imidazolium-functionalized poly(styrene-co-acrylonitrile) membranes. These materials exhibited ca. 30-50% water uptake and 25-40 mS cm-1 hydroxide conductivity at 30°C. Degradation testing by immersion in 1 M NaOH solutions at 60°C showed no significant change in IEC after

400 hours, indicating excellent alkaline resistance of the imidazolium cations.

However, a 4x decrease in the tensile modulus and 2x decrease in the tensile strength suggests degradation of the polymer backbone.

In response, F. Zhang et al.[87] hypothesized that using a bisphenol A polysulfone backbone, which exhibits excellent chemical stability, would yield an imidazolium AEM with better mechanical integrity under alkaline conditions; they therefore synthesized imidazolium polysulfone by quaternizing chloromethylated polysulfone with methylimidazole. The prepared materials had IECs between 1.39 and

2.46 mEq g-1, water uptake between 8.5% and 93%, and hydroxide conductivity

-1 between 16 and 21 mS cm at 20°C. H2/O2 MEA performance was relatively poor in

2 comparison to other AEMs, with a peak power density of 16 mW/cm at 60°C.

However, the imidazolium polysulfone materials showed poor alkaline stability: immersion in 3 M NaOH at 60°C resulted in membrane discoloration and a 23% loss in conductivity.

45 Chapter 3

Also using an aromatic backbone, Kim et al.[88] prepared ethyl imidazolium- functionalized poly(arylene ether sulfone) copolymers and compared it against a quaternary ammonium poly(arylene ether sulfone). The ethyl imidazolium- functionalized membranes showed higher hydroxide conductivity, ranging from 30 mS cm-1 to 100 mS cm-1 in the temperature range 20°C and 80°C, while the quaternary ammonium membranes only exhibited 21 mS cm-1 to 73 mS cm-1 . Moreover, whereas the ethyl imidazolium membrane showed neglible changes in IEC and retained mechanical integrity after immersion in 2 M NaOH at 60°C for 168 h, the quaternary ammonium membrane broke into pieces after only 48 h.

While these early experimental data indicate that imidazolium-based polymer electrolytes exhibit high alkaline resistance, more recent investigations show that the imidazolium cation may not be as stable as initially believed. A chemical stability study by Ye and Elabd[89] of a polymerized imidazolium ionic liquid, poly(1-[(2- methacryloyloxy)ethyl]-3-butylimidazolium hydroxide), showed that while the imidazolium headgroup was stable under mild alkaline conditions (<1 M KOH solution) it was susceptible to ring-opening degradation at higher KOH concentrations

(>1 M). Furthermore, a detailed investigation by Varcoe et al.[82] found that benzylmethylimidazolium groups offered no performance advantage over benzyltrimethylammonium. Stability studies of both small molecule and polymer- bound benzylmethylimidazolium using Raman spectroscopy indicate that the imidazolium groups actually have poorer alkaline stability than quaternary ammonium, leading the authors to contend that there is no real advantage of using pendant imidazolium moieties.

46 Chapter 3

D. Chen and M. Hickner[90] reported that poly(fluorenyl ether ketone sulfone) functionalized with imidazolium performed similarly to that functionalized with quaternary ammonium. In addition, immersing the membranes in 1 M NaOH at 60°C for 48 hours resulted in ~20% reduction in IEC for all samples, independent of cation identity. When the temperature was raised to 80°C, the imidazolium-functionalized membrane was actually less stable than the quaternary ammonium membrane and retained only 43% of its original IEC while the latter retained 78% of its original IEC.

The authors suggest that, despite offering charge delocalization, the planar structure of the imidazolium group provides no steric protection and renders it susceptible to hydroxide attack.

Various researchers have reported that the alkaline stability of imidazolium is highly dependent on its sidegroups. Page et al.[91] showed that 1-benzyl-2,3- dimethylimidazolium was more stable than 1-benzyl-3-methylimidazolium in 1 M

KOH at 60°C, although both headgroups were nevertheless less stable than benzyltrimethylammonium. F. Yan et al.[92] prepared a series of imidazolium salts and polymers where the C2 hydrogen was substituted with methyl, isopropyl, and phenyl groups and found the order of stability, at least for the small molecule model compounds, to be methyl > isopropyl > phenyl > unsubstituted. The stabilizing effects of these sidegroups were rationalized in terms of both their steric hindrance ability and their influence on the LUMO energy of the imidazolium cation.

Herring et al.[83] prepared an imidazolium PPO membrane in which the imidazolium headgroup had methyl functionalities at the C4 and C5 positions of the imidazolium ring and a 2,4,6-trimethoxyphenyl group at the C2 position. The electron

47 Chapter 3 donating and steric hindrance offered by the trimethoxyphenyl was suspected to increase alkaline stability, as the material showed no degradation after 24 h immersion in 1 M NaOH at 80°C.

On the other hand, a structure-stability study by Price et al.[93], examining the effects of different sidegroups the C2 position of the imidazolium ring, suggested that steric hindrance is the least effective method for enhancing cation stability. Instead, the authors contend that the major factor for increasing alkaline stability is to introduce functionalities at the C2 position that provide alternative, competing deprotonation reactions.

X. Yan et al. investigated imidazolium-functionalized PEEK AEMs

(IECs=1.56 to 2.03 mEq g-1) by quaternizing chloromethylated PEEK with 1- methylimidazole.[77] Their previous work with PEEK AEMs[56–58] suggested that the high electron density of the PEEK backbone led to enhanced van der Waal interactions and, consequently, better dimensional stability than other polyaromatics. Interestingly, the swelling ratio for these materials remained constant between 20 and 50°C and did not increase until 60°C. Compared to imidazolium-functionalized PSf, the PEEK materials at 60°C exhibited roughly half the swelling ratio at similar IECs. In-plane hydroxide conductivity ranged between 15 and 52 mS cm-1, depending on IEC, which is higher than other imidazolium-functionalized membranes as well as benzyltrimethylammoniun-functionalized PEEK. The chemical stability of these materials is unknown as no alkaline degradation tests were reported.

48 Chapter 3

Polybenzimidazole

Poly(benzimidazoles) (PBI) have imidazole groups as part of the backbone. PBIs represent an interesting class of materials that can be converted to either cation or anionic forms, depending on the counterion used. PBIs have been investigated as high- temperature PEM materials via acid-doping, but initial attempts to create poly(benzimidazolium) AEMs found that the materials were unstable in the hydroxide form.[94,95] As a solution, Holdcroft et al. introduced bulky mesitylene groups at the C2 position of the benzimidazolium to sterically stabilize the cation against hydroxide attack (Scheme 3.10).[96] Spectroscopic analysis via NMR showed no degradation of the cation after 10 days of immersion in 2 M NaOH at 60°C. However, the pure

-1 cationic PBI membrane had an extremely high IEC of 4.5 mEq g and was consequently soluble in water. To create dimensionally stable AEMs, the authors blended the sterically-crowded cationic PBI with an anionic PBI to create ionically

-1 crosslinked membranes with IECs of 2.0, 1.5 and 1.0 mEq g and respective water uptakes of 162, 119, and 82% and corresponding hydroxide conductivities of 9.6,

10.1, and 13.2 mS cm-1 at 21°C. Submersion of the membranes in 2 M KOH at 60°C showed no notable change in IEC after 13 days, showing that the crosslinked materials retained the excellent alkaline stability of the cations.

Scheme 3.10 Alkaline stable anionic PBI with mesitylene group. Adapted from Holdcroft et al. [96]

49 Chapter 3

L. Jheng et al. investigated quaternized PBI AEMs with methylimidazolium sidegroups (Scheme 3.11).[97] Their synthesis allowed for the flexibility of introducing imidazolium moieties on the main PBI chain via quaternization of the benzimidazole

(mQPBI), as a sidechain via covalent grafting of methylimidazolium (sQPBI), or both

-1 (msQPBI), preparing AEMs with respective IECs of 1.42, 0.96, and 1.49 mEq g .

The sQPBI and msQPBI membranes exhibited similar water uptakes (~41%) compared to the mQPBI membrane (18.4%). Despite similar hydration levels, the msQPBI membrane had much higher hydroxide conductivity (27.2 mS cm-1) than the sQPBI membrane (10.7 mS cm-1) at 80°C. Alkaline treatment in 1 M KOH at 30°C showed rapid decrease in conductivity across all samples, and NMR characterization revealed degradation of both benzimidazolium and methylimidazolium cations.

Scheme 3.11 Permutations of ionized/non-ionized PBI with and without imidazolium functionalities reported by L. Jheng et al. [97]

50 Chapter 3

3.1.3.3 Guanidinium

The guanidinium headgroup has been proposed as an alternative to quaternary ammonium in light of its high basicity and presumed stability resulting from charge delocalization. S. Zhang et al.[98] tethered guanidinium to chloromethylated

-1 poly(phenyl sulfone), preparing membranes with IECs between 0.79 and 2.15 mEq g

. Water uptake ranged from 12% to 88% and hydroxide conductivity was between 5 mS cm-1 and 67 mS cm-1, and were directly proportional to the IEC. Immersion of an

-1 IEC=1.68 mEq g membrane in 1 M NaOH solution at 60°C showed no change in conductivity after 30 days, indicating exceptional alkaline stability.

S. Zhang et al. also reported poly(aryl ether sulfone)s with pendant hexaalkylguanidinium groups.[99] Stained TEM imaging showed large hydrophilic

-1 clusters ca. 40-50 nm in diameter. A membrane with an IEC of 1.39 mEq g exhibited hydroxide conductivity between 20 mS cm-1 and 42 mS cm-1 in the 15°C to 60°C temperature range, with only moderate water uptake (36%) at 60°C. This guanidinium-functionalized membrane also showed excellent alkaline stability -- immersion in 10 M and 2 M KOH solution for 24 h showed no notable change in the conductivity.

Bai et al.[100] prepared guanidinium-functionalized, self-crosslinked PPO

-1 (IEC=0.80 mEq g ) with conductivity between 14 mS cm-1 and 51 mS cm-1 in the 20 to 70°C temperature range. Immersing the membrane in 1 M NaOH at 60°C moderate

5-10% decrease in conductivity after 2 days. However, the tensile stress at break of the membranes decreased by nearly 50% from 3.8 MPa to 2.1 MPa. Moreover, after 5 days the membranes were reported to have lost mechanical integrity. Thus, while the

51 Chapter 3 guanidinium headgroup was found to be relatively stable in alkali media, the functionalized PPO backbone suffered from chain scission.

3.1.3.4 Pyrrolidinium

Recent reports on pyrrolidinium cations show promising stability. J. Qiao et al. synthesized AEMs with pyrrolidinium cations by crosslinking poly(vinyl alcohol) with poly(acrylamide-co-diallyldimethylammonium chloride) using glutaraldehyde as the crosslinking agent (Scheme 3.12).[101] While these materials had moderate IECs

- (0.91 to 1.63 mEq g 1) , they exhibited poor hydroxide conductivities (0.77 to 3.03 mS cm-1) and high water uptake (72.2 to 121.1%) at 23°C. However, both the cation and polymer backbone showed excellent alkaline stability – there were no significant changes in either conductivity or mechanical properties even after aging the membranes in 6.0 M KOH at 80°C for 250 h.

Scheme 3.12 Pyrrolidinium-based random copolymer membrane reported by J. Qiao et al. [101]

In light of the reported superb chemical resistance, F. Yan et al.[102] prepared a series of pyrrolidinium-based AEMs, systematically studying the effects of different substituents at the pyrrolidinium nitrogen group on alkaline stability and found that the choice of subsitutent heavily influences the long-term alkaline stability of the pyrrolodinium cation. NMR characterization of pyrrolidinium cations showed that the

52 Chapter 3 ethyl-substituted pyrrolodinium([EMPy]+) yielded the highest stability, with a degradation degree of 8.16% after aging in 1 M NaOH at 80°C for 96 h. In comparison, imidazolium and benzyltrimethylammonium showed 37.62% and 31.37% degradation, respectively. Characterization of AEMs, prepared by photoinitiated copolymerization of pyrrolidinium with styrene, acrylonitrile and divinylbenzene, supported the NMR results. Stability testing showed no change in IEC for the

[EMPy]+ AEM after 168 h of immersion in 1 M NaOH at 80°C; in contrast, the

BTMA membrane ~40% of its initial IEC under similar conditions. This stability is reflected in the hydroxide conductivity, with the [EMPy]+ membranes showing no notable change in conductivity after 168 h in 1 M NaOH at 80°C. Despite the relatively low conductivity of reported pyrrolidinium-based membranes (ca. 10 mS cm-1), their exceptional stability makes them promising for future development.

3.1.3.5 Sulfonium

Y. Yan et al.[103] investigated triarylsulfonium (TAS) as a cationic group tethered to a bisphenol A polysulfone backbone (Scheme 3.13), looking specifically at triarylsulfonium headgroups with and without a methoxyl subsituent (PSf-

MeOTASOH and PSf-TASOH, respectively). Both materials were produced with an

IEC around 0.68-0.69 mEq g-1. At 20°C, the PSf-MeOTAS membrane exhibited twice the conductivity (15.4 mS cm-1) as the PSf-TAS membrane (7.7 mS cm-1) , attributed to increased cation basicity as a result of the electron donating ability of the methoxyl group. The methoxyl subtituent was also found to be critical to alkaline stability.

Whereas the MeOTAS cation was stable in 1 M KOD solution at 60°C for 10 days, the

53 Chapter 3

TAS cation completely degraded. Moreover, immersion of PSf-MeOTAS in 1 M KOH for 30 days showed no conductivity loss.

Scheme 3.13 Tertiary sulfonium with methoxyphenyl sidegroup tethered to PSf backbone.

Image adapted from Y. Yan et al. [103]

3.1.3.6 Ruthenium

Hickner et al.[104] reported the first AEMs with non-organic pendant cations, instead using covalently bound divalent ruthenium-complexes (Scheme 3.14). The AEMs are based on a polynorbornene backbone with bis(terpyridine)ruthenium(II) and dicyclopentadiene functionalities. The divalent nature of the Ru complex allows the pendant cation to associate with two anions, providing potential for higher IEC

-1 materials. The membranes were prepared with IECs of 1.0, 1.4, and 2.0 mEq g resulting in respective water uptakes of 30%, 126%, and 432% and hydroxide conductivities of 14.1, 28.6, and 19.6 mS cm-1 at 30°C.

54 Chapter 3

Scheme 3.14 Ruthenium-functionalized polynorbornene. Figure adapted from Hickner et al.[104]

Further investigation[105] showed that diffusion coefficients for ion transport in the Ru(II) complex membranes were roughly an order of magnitude lower than the dilute solution diffusivity and that hydration numbers optimal for ion transport were much higher than in AEMs with other cations, indicating the Ru(II) complex introduces barriers to ion transport. This inhibition is ascribed to the large size of the

Ru(II) complex, causing steric hindrance, and its uniform electrostatic surface potential, which provides no local preference for water association.

3.2 Composite Membranes

Composite AEMs are characterized by ionomer materials in which the conductive portion is chemically distinct from an inert matrix. This is often manifested by embedding an ionomer in an inert polymer film, typical of organic composites, or as inert materials dispersed within an ionomer matrix, typical of inorganic-organic composites. The decoupling of conductivity from mechanical properties provides a powerful platform for tuning these properties separately.

55 Chapter 3

3.2.1 Pore-Filled Composites

Filling the void spaces of a porous, inert polymer scaffold is a popular approach in creating AEMs with high dimensional stability. These materials have a naturally segregated morphology in which the conductive ionomer is spatially separated from the inert matrix, often resulting in high conductivities.

To separate the charged species from the inert backbone, X. Zhang et al. developed a novel processing method to create a QA PSf material with a grid-plug microstructure. They first prepared a porous polysulfone film using a phase-inversion technique and then filled the pores with (3-acrylamidopropyl)trimethylammonium chloride monomers. The impregnated film was subject to UV to photopolymerize the

QA monomers within the pore spaces, resulting in a percolating ion-conducting phase within the polysulfone film. Their best performing membrane exhibited an in-plane

2 hydroxide conductivity around 23 mS cm-1 and peak power density of 55 mW/cm at

60°C. While it is likely that this approach provides enhanced mechanical properties compared to traditional solution-casting methods, the authors did not report any mechanical characterization.

Y. Cao et al. polymerized quaternary ammonium-functionalized poly(vinylbenzyl chloride) (QA PVBz) within the pore spaces of a preformed PTFE film.[106] Compared to a neat QA PVBz membrane, the pore-filled membrane showed significantly reduced water uptake (13% vs 43%) and dimensional swelling (7% vs

21%), while maintaining similar hydroxide conductivity (ca. 18 mS cm-1, 20°C). An

H2/O2 MEA fabricated from the composite material showed good performance,

56 Chapter 3

2 reaching a peak power density around 160 mW/cm and a maximum current density

2 around 630 mA/cm at 25°C.

In a similar study, X. Wang et al. swelled a porous PTFE sheet with DABCO- functionalized bisphenol A polysulfone.[107] The composite material had better tensile strength (32 MPa vs 22 MPa) and lower water uptake (61% vs 81%) compared to neat

DABCO PSf while exhibiting similar hydroxide conductivities (ca. 35-40 mS cm-1,

30°C). Fuel cell performance reached 146 mW/cm2 peak power density at 50°C under

H2/O2 flows.

X. Zhang et al. recently reported a crosslinked, imidazolium-functionalized poly(epichlorolydrin) ionomer embedded into porous PTFE support.[108] The composite materials exhibited similar hydroxide conductivities as the neat material

(14-18 mS cm-1, 30°C) but had significantly improved mechanical properties. Whereas the neat membranes exhibited tensile strengths and Young’s moduli < 2.5 MPa, the

PTFE-reinforced membranes had tensile strengths between 53-67 MPa and Young’s moduli between 82-110 MPa. Aging of the composite membrane in 1 M KOH at 60°C for 15 days showed no significant change in conductivity. Fuel cell performance at

50°C under H2/O2 flows was relatively low compared to other reports, with a 23 mW/cm2 peak power density.

H. Yu et al.[109] dispersed quaternary ammonium poly(vinylbenzyl- divinylbenzene) (QA PVD) and poly(vinylbenzyl-divinylbenzene-hexa fluorobutyl methacrylate) (QA PVDh) within porous PTFE films, preparing composite

-1 membranes with IECs between 1.35 and 1.76 mEq g . The composites showed high hydroxide conductivity (30 to 57 mS cm-1 vs 22 mS cm-1) and good tensile strength

57 Chapter 3

(100 to 115 MPa). The imidazolium headgroup was also found to be highly stable in 1

M KOH at 60°C, with no changes in conductivity in IEC after immersion for nearly

1000 h. Incorporation of the composite AEM in a H2/O2 MEA yielded excellent performance, with an optimized cell delivering a peak power density of 370 mW/cm2.

L. Zhuang et al.[110] impregnated porous PTFE with benzyltrimethylammonium polysulfone, significantly enhancing the tensile strength of the hydrated AEM from 5.3 MPa in the neat BTMA PSf membrane to 25.4 MPa in the composite membrane. Moreover, the use of a PTFE matrix reduced the swelling ratio and water uptake by more than half to 14.3% and 54.7%, respectively. At the same time, the composite material preserved the high hydroxide conductivities of pristine

BTMA PSf, measuring ~16 mS cm-1 at 25°C. Because of the high mehcnaical strength provided by the PTFE film, the composite AEM could be synthesized to a thickness of

20 um without compromising dimensional integrity. The low thickness results in lower resistive loss, culminating in excellent H2/O2 single-cell performance with 315 mW/cm2 peak power density at 50°C. However, the BTMA PSf was found to slowly leach out of the PTFE matrix after prolonged operation at elevated temperatures. To inhibit this loss, the BTMA PSf ionomer was replaced with a tertiary amino polysulfone, which undergoes quaternization via a self-crosslinking process.[36,111]

Immersion of the crosslinked composite for 100 h in an 80°C water bath found no evidence of leaching. In contrast, the composite filled with linear BTMA PSf lost roughly 35% mass over the same period. Incorporation into a H2/O2 fuel cell yielded

550 mW/cm2 maximum power density at 60°C and 0.1 kPa backpressure.

58 Chapter 3

T. Guo et al.[112] swelled porous PTFE films with polyepichlorohydrin

(PECH). The PECH was then quaternized with tetramethylhexanediamine to introduce ionic headgroups. The authors examined composites with different amine contents,

-1 -1 resulting in IECs between 1.10 mEq g and 1.70 mEq g . Characterization at 30°C showed that water uptake (70-100%) and hydroxide conductivity (5-35 mS cm-1) scaled proportionally with IEC. Moreover, 24 h immersion of the composite membranes in 8 M KOH at 30°C showed negligible changes in the conductivity, suggesting some adequate stability of the quaternary ammonium groups. The authors suggest this stability is a result of increased electron density at the β-carbon in the quaternary ammonium groups from THMDA, inhibiting deprotonation and eventual elimination of the charged species.

3.2.2 Electrospun Composites

Pintauro et al. have reported dimensionally robust electrospun fiber mats as AEM materials.[113–116] These composites are created by electrospinning chloromethylated polysulfone with neat poly(phenyl sulfone). The highly porous fiber mat is then compressed at high pressure (2,000 psi) and treated with chloroform vapor to increase the fiber volume from 25% to 50%. The membrane was then soaked in trimethylamine to quaternize the chloromethyl groups. The charged composite was compressed at

15,000 psi to further reduce the void space volume and then solvent-annealed with

THF vapor to flow the poly(phenyl sulfone) into the remaining pore spaces. This series of processing steps is critical as high void volumes can greatly compromise the mechanical properties of the mat.[117]

59 Chapter 3

The inert poly(phenyl sulfone) provides structural support, reducing water swelling by a factor of ~2-2.5x from 48-170% in the neat membranes to 25-75% in the electrospun composite membranes. Hydroxide conductivities of the composite and neat membranes were similar, although the highest IEC (2.49 mEq g-1) material showed higher conductivity in the composite (~40 mS cm-1) than in the neat material

(32.6 mS cm-1) as a result of lower charge dilution. The high IEC composite membrane exhibited a tensile strength of 25 MPa and good flexibility; in contrast, its corresponding neat membrane could not be tested as it swelled excessively in water and was extremely brittle when dry. The properties of the fiber mats were further optimized by using aliphatic diols or diamines to crosslink the conductive quaternized polysulfone chains.[115,116] A diamine-crosslinked composite with 35 wt.% inert

-1 poly(phenyl sulfone) and an effective IEC of 2.01 mEq g exhibited a high hydroxide conductivity (65 mS cm-1) at 23°C with 144% water uptake and 14 MPa tensile strength.

3.2.3 Inorganic Composites

Inorganic fillers such as SiO2, Al2O3, TiO2 and Bentonite have been extensively studied as plasticizers in AEMs based on polymer-salt complexes. These polymers, typically poly(vinyl alcohol) or poly(ethyelene oxide), lack a pendant cation headgroup and instead rely on solvation of salts via coordination of electron-donor groups (e.g., ether or alcohol groups) on the polymer. Anion transport in these materials occurs through segmental motion of the polymer chains, and nanoparticle fillers have been shown to enhance ion transport kinetics by disrupting rigid

60 Chapter 3 crystalline domains within the polymer. This mode of ion transport is significantly slower than the Grotthus-like transport present in water-swollen, pendant cation polymer electrolytes, and the ionic conductivity of these polymer-salt complexes are too low to be implemented in alkaline fuel cell stacks. Nevertheless, impregnation of nanoparticles in ion-functionalized materials have been shown to lead to advantageous polymer-particle interactions, resulting in improved conductivity and mechanical properties.

Zhao et al. embedded quaternary ammonium-functionalized silica nanoparticles (QA SiO2, 5 to 20 wt.%) within a crosslinked poly(vinyl alcohol) (PVA) matrix.[118] Because the cationic moieties were tethered to the silica, higher silica loading resulted in higher performance. However, compared to homogeneous materials, these composites exhibited lower hydroxide conductivity (4-8 mS cm-1,

20°C) and higher water uptake (ca. 120%). C. Yang et al.[118] investigated a similar material by blending crosslinked QA PVA with QA SiO2 (5 to 20 wt.%). The hydroxide conductivity (30°C) improved upon QA SiO2 incorporation from 1.24 mS

-1 -1 cm in pure QA PVA to 2.37 mS cm in QA PVA/QA SiO2(20 wt.%).

C. Yang et al.[119] also reported crosslinked QA PVA membranes impregnated with Al2O3 nanoparticle fillers. Inclusion of 10 wt.% Al2O3 in a crosslinked QA PVA membrane increased the ionic conductivity (4 M KOH, 30°C) from 13.7 mS cm-1 to

35.0 mS cm-1 and offered enhanced dimensional stability.

X Li et al. dispersed zirconica (ZrO2) nanoparticles to 2.5 to 20 wt% in a quaternary ammonium-functionalized poly(arylene ether sulfone) (QAPAES) ionomer

[120] matrix. The weakly-electronegative ZrO2 nanoparticles were reported to coordinate

61 Chapter 3 with quaternary ammonium cations and hydroxide anions, facilitating the aggregation of ionic domains. Moreover, because of this coordinating potential, titration measurements revealed that the ZrO2 particles were able to increase the effective IEC of the material. The combination of these two factors resulted in higher conductivities upon incorporation of the ZrO2 fillers compared to bare QAPAES; 10% incorporation

-1 - of ZrO2 increased the hydroxide conductivity (20°C) from 9.5 mS cm to 23.1 mS cm

1. In addition, XRD measurements suggest that, because of their interaction with the quaternary ammonium groups, the ZrO2 nanoparticles are able to induce crystallization within the QAPAES matrix. Consequently, the swelling of the composite membranes were reduced and their tensile strengths increased compared to bare QAPAES, despite having higher water uptake. The authors also investigated the

[121] same ZrO2 filler approach in multiblock copoly(arylene ether sulfone) AEMs.

However, (the effects of ZrO2 incorporation) performance enhancements in these multiblock materials as a result of ZrO2 incorporation were not as significant due to inhibited cation-ZrO2 interactions.

3.3 Phase-Segregated Membranes

In an effort to enhance anion transport efficiency, recent experimental approaches have shown novel design strategies towards creating AEMs with phase-segregated nanostructures akin to Nafion. Interestingly, many of these structured materials are based on functionalization and modification of random copolymers, as opposed to traditional diblock copolymers typically associated with phase-separated morphologies. That random copolymers can have well-defined local ordering is unsuprising, given that Nafion itself can be described as a random graft copolymer;

62 Chapter 3 however, little attention has been afforded to a fundamental understanding of the phase-separation behavior of these types of copolymers.

3.3.1 Random Copolymer

While there have been many recent reports on inducing phase-separated structures in random copolymer-based alkaline exchange membranes, their general design principles can be categorized as linear random copolymers, graft copolymers or multiblock copolymers.

3.3.1.1 Linear Random Copolymers

Tsai et al.[122] prepared polyisoprene-co-poly(vinylbenzyltrimethylammonium chloride) materials by random nitroxide-mediated copolymerization of isoprene and 4- vinylbenzyl chloride, followed by quaternization of the benzyl chloride groups by trimethylamine (Scheme 3.15). The ionomer was solution cast and thermally crosslinked to create dimensionally stable AEMs. The IECs of the membranes ranged from 0.77 to 2.34 mEq g-1, with respective water uptakes between 27.8 and 515.6%.

Chloride conductivities at 60°C ranged from 0.07 mS cm-1 to 16.8 mS cm-1, where the

-1 highest value was obtained from the sample with an intermediate IEC of 1.54 mEq g

; more highly charged membranes exhibited charge dilution from excessive water uptake. All hydrated materials, even the low conductivity sample with an IEC of 0.77 mEq g-1, showed distinct scattering peaks under SAXS analysis, with characteristic domain sizes between 5.0 to 6.3 nm. The lowest IEC and conductivity material had the largest domain size of 6.3 nm, while all other materials had domain sizes in the 5.0 to

5.6 nm range. Moreover, there is a significant increase in conductivity from 0.07 mS

63 Chapter 3

-1 cm-1 to 9.12 mS cm-1 as the IEC is increased from 0.77 to 1.26 mEq g . Based on this information, the authors contend that there is an optimal domain size in these materials for the hydrophilic ionic clusters to form a percolating ion-transport network. The alkaline stability of these membranes were not reported.

Scheme 3.15 Polyisoprene-co-poly(vinylbenzyltrimethylammonium chloride) random copolymer. Figure adapted from Tsai et al.[122]

3.3.1.2 Graft Copolymers

Graft copolymers comprise the largest category of phase-separated alkaline exchange membranes. At a very fundamental level, these materials are described by oligomers randomly grafted along a polymer backbone. There is large diversity in the choice of oligomers (e.g., whether they are hydrophobic or hydrophilic, ionic or non-ionic) and how they are tethered to the backbone in relation to the cationic headgroup (e.g., as aliphatic spacers/tails or co-grafted along the main chain).

Aliphatic Tail

N. Li et al. have performed several studies investigating the choice of functional groups on the quaternary ammonium cation. Their initial report showed that substituting one of the methyl groups on the ammonium cation with a longer

64 Chapter 3 hexadecyl chain induces phase-separation in QA PPO membranes.[123] The synthesis followed the typical halogenation and Menshutkin functionalization approach.

However, the authors developed developed a new protocol in which bromination occurred selectively at the benzyl position of PPO. The bromomethyl groups were treated with dimethylhexadecylamine (DHMDA) to introduce quaternary ammonium groups with a long aliphatic tail, creating a brush-comb architecture. SAXS and AFM characterization found phase-separated domains on the order of 3 nm. In contrast, the

PPO membranes treated with trimethylamine were homogeneous (i.e., benzyltrimethylammonium groups upon quaternization), exhibiting no scattering under SAXS. As a result of domain formation, the DHMDA PPO samples showed high hydroxide conductivity (up to 35 mS cm-1 at room temperature) while retaining low gravimetric water uptake (<20%) and dimensional swelling (<6%). Indeed, at similar IECs, the comb-shaped DHMDA PPO material showed nearly three times the hydroxide conductivity and half the water uptake compared to the homogeneous

TMA-functionalized PPO. Moreover, whereas the TMA PPO benchmark catastrophically failed after only 80 hours in accelerated degradation conditions (1 M

NaOH, 80°C), the DHMDA PPO materials showed excellent alkaline resistance and maintained 80% of its initial conductivity even after 500 hours. The authors attribute this stability to the steric hindrance of hydroxide attack by the large hexadecyl- sidechain.

65 Chapter 3

Scheme 3.16 Comb-shaped QA PPO membrane with aliphatic tail. Figure adapted from N. Li et al.[124]

The effects of the length of the aliphatic tails were also explored by using dimethylalkylamines with alkyl sidechains of 6, 10, and 16 carbons.[124] The domain size, estimated from SAXS, scaled directly with the length of the alkyl sidechain. The longer hexadecyl sidechain produced the most distinct phase-separation and highest conductivity efficiency; however, increasing the degree of substitution to produce

-1 materials with IECs greater than 2.5 mEq g resulted in poor film-forming ability, presumably due to incompatibility between the long hexadecyl sidechain and the PPO backbone. This is again evident in the mechanical stabilities of these materials as ionomer binders in an MEA. While the hexadecyl sidechain offered the highest peak power density (77 mW/cm2 at 50°C), it suffered a short lifetime compared to the MEA created with amine with hexyl sidechains.

N. Li et al.[125] also reported using a combination of hexyl and undecene- functionalized dimethylalkylamines to quaternize PPO. The undecene group provided a synthetic pathway for covalent crosslinking via Grubbs II-catalyzed olefin metathesis, producing dimensionally stable membranes with high IECs ranging

66 Chapter 3

-1 between 2.42 and 3.20 mEq g . Increased crosslinking resulted in decreased water uptake and swelling while preserving high hydroxide conductivity, with an optimal material reaching 40 mS cm-1 at only 8% swelling ratio at 25°C. SAXS characterization showed that crosslinking not only retained the comb-like structure reported in the linear systems, but actually facilitated structure formation by co- localizing the alkyl chains. In addition to their high dimensional stability and performance, these materials exhibited excellent alkaline resistance with no notable changes in IEC or conductivity after immersion for 30 days in 1 M NaOH solution at

80°C.

Scheme 3.17 Poly(arylene ether nitrile) random copolymer. Dark regions in TEM image indicate stained hydrophilic domains. Figure adapted from Q. Liu et al.[126]

In a similar study, Q. Liu et al.[126] recently reported poly(arylene ether nitriles) prepared with diamines separated by an aliphatic chain (Scheme 3.17). Mono- quaternization results in an aliphatic tail on the quaternary ammonium moitie, while di-quaternization yields self-crosslinking. Membranes were produced with IECs between 1.56 and 2.51 mEq g-1 and, as a result of the crosslinking, had water uptake below 40% even at high IEC values. Moreover, the materials exhibited good hydroxide conductivity, with the highest IEC membrane measuring 50.6 mS cm-1 at

67 Chapter 3

30°C. Stained TEM imaging showed distinct hydrophilic ion clusters, especially at higher charge contents. Stability testing from immersion of the materials in 2 M

NaOH at 60°C showed that the materials lost roughly 30% to 40% of their initial conductivity after 1000 hours of submersion. At the same time, the authors detected no indication of hydrolysis of the backbone, suggesting that the loss in conductivity is a result of nucleophilic attack of the quaternary ammonium species.

Spacer Chains

Drawing inspiration from the use of aliphatic spacers in QA PSf materials[23,127], Jannasch et al.[128] devised a synthesis protocol that introduced an aliphatic spacer between the PPO backbone and the quaternary ammonium ions

(Scheme 3.18). The hypothesis was that the flexible aliphatic spacer would (1) facilitate phase separation of the quaternary ammonium groups from the rigid PPO backbone, resulting in more efficient anion transport; and (2) decouple the electron- withdrawing effect of the cation from the backbone, leading to enhanced alkaline stability. The introduction of a seven-carbon long aliphatic spacer was shown to give rise to ionic clusters with a character size of 3.3 nm (measured via SAXS). In contrast

(and in agreement with other literature reports), the benchmark benzyltrimethylammonium PPO material, where the quaternary ammonium group is separated form the PPO backbone by a single methylene group, showed no indication of domain formation. Compared to the BTMA-PPO benchmark, a comparable nanostructured material exhibited a 4x increase in hydroxide conductivity (20 mS cm-1 vs 5 mS cm-1 at 20°C). Interestingly, although the micro-structured materials exhibited

68 Chapter 3 higher conductivity at equal water uptake compared to the BTMA-PPO benchmark, their equilibrium water uptakes were nevertheless higher despite more hydrophobic character as a result of the aliphatic spacers. Most importantly, the aliphatic spacers imparted exceptional alkaline stability, with no notable degradation in hydroxide conductivity after soaking in a 1 M NaOH solution at 80°C for 196 hours.

A B

Scheme 3.18 Quaternary ammonium PPO with spacer chains reported by (A) Jannasch et al.[128] and (B) Bai et al.[129] Figures have been adapted from their respective sources.

Bai et al.[129] incorporated 2,4,6-tri(dimethylaminomethyl)-phenol onto PPO to fabricate QA PPO materials with trifunctional ammonium moieties (PPO-TQA)

(Scheme 3.18). The trifunctional materials had IECs ranging from 0.84 to 1.50 mEq g-

1 and their properties were compared to BTMA PPO with IECs of 0.87 and 1.97 mEq

-1 g ; the use of trifunctional ammonium moieties resulted in higher conductivity, lower water uptake, and enhanced alkaline stability. The PPO-TQA with an IEC of 1.50

-1 mEq g had a hydroxide conductivity of 71.7 mS cm-1 at 60°C, which was nearly 4

-1 times greater than that of BTMA PPO with an IEC of 1.97 mEq g . In a more direct

-1 comparison, PPO-TQA with an IEC of 0.84 mEq g had roughly three times the hydroxide conductivity (29.8 mS cm-1 vs 11.6 mS cm-1) and half the water uptake

-1 (8.5% vs 15.9%) compared to a BTMA PPO sample with an IEC of 0.87 mEq g .

69 Chapter 3

SAXS of the dry PPO-TQA samples showed no obvious phase separation; however, the authors did not investigate the scattering of hydrated materials. Thus, the authors attribute the enhanced performance to an increase in local ion concentration as opposed to morphological enhancements. While the authors reported increased alkaline stability, they did not conjecture as to a possible cause.

Ionic Grafts

T. Xu et al.[130] investigated a different approach for synthesizing comb-shaped PPO copolymers. Instead of using an aliphatic spacer or tail, they used ATRP to graft quaternary ammonium-functionalized 4-vinylbenzyl chloride to the PPO backbone, resulting in a ion-dense sidechain between 4 and 6 monomers in length at a graft density between 10 and 12%. TEM and SAXS characterization suggest that this graft approach results in the formation of hydrophilic domains on the order of 60 nm. The graft copolymers had much higher water uptake (75% vs 25%, 20°C) compared to

BTMA PPO. As a result of the higher water content and morphology, the comb- shaped copolymers exhibited nearly 4x higher hydroxide conductivity (39 mS cm-1 vs

9 mS cm-1, 20°C) compared to BTMA PPO. However, the comb-shaped structure did not influence alkaline stability, as testing revealed similar losses in conductivity and

IEC compared to BTMA PPO. Notably, while the conductivity decreased by roughly

50% after immersion for 288h h in 2 M NaOH at 60°C, the material retained roughly

85% of its original mechanical properties. The dominant mechanism for degradation in these materials can likely be traced to nucleophilic attack of the quaternary ammonium cation as opposed to deterioration of the PPO backbone.

70 Chapter 3

Scheme 3.19 PPO with ionic crafts synthesized via ATRP, adopting a rod-coil architecture.

Figure adapted from T. Xu et al.[131]

In a follow-up study, T. Xu et al.[131] investigated the same graft copolymer concept, but with lower graft density (4%) and longer graft length (6-12 repeat units), thereby creating a rod-coil architecture with flexible ionically-active side-chains tethered to a rigid PPO backbone (Scheme 3.19). As a result, the rod-coil polymers had roughly 67% of the gravimetric water uptake as BTMA PPO at similar IEC. Real- space imaging with AFM and TEM revealed a phase-separated structure characterized by interconnected hydrophilic domains on the order of 3-5 nm, resulting in high hydroxide conductivities (ca. 50 mS cm-1 at 30°C). The rod-coil material showed exceptional alkaline stability, retaining ~80% of its original conductivity after immersion in 2 M NaOH at 60°C for over 1400 h.

Co-grafting

Using molecular dynamics simulations, L. Zhuang et al.[127] investigated various methods of modifying QA polysulfone towards producing an enhanced nanostructure conducive to ion transport. Their approach focused on introducing additional

71 Chapter 3 hydrophobic moieties in order to drive aggregation of ions into efficient ion transport clusters. Specifically, they compared the simulated hydrated nanostructures of QA PSf materials in which hydrophobic aliphatic chains are (1) used as a spacer to separate the

QA group from the backbone; (2) attached directly to the QA group as a ‘tail’ (i.e., benzyldimethylalkylammonium); and (3) co-grafted alongside BTMA groups (Scheme

20).

Scheme 20 Different molecular designs for quaternary ammonium polysulfone explored by L.

Zhuang et al. (A) QA groups are directly tethered to the polysulfone backbone (e.g., BTMA

PSf). (B) A hydrophobic aliphatic spacer is used to separate the QA group from the PSf backbone. (C) A hydrophobic aliphatic chain is used as ‘tail’ on the QA group. (D)

Hydrophobic aliphatic chains are grafted alongside BTMA groups.

The simulation results indicated a co-grafting approach would yield the most efficient ion transport morphology. The authors tested this experimentally by synthesizing a series of BTMA polysulfones onto which were co-grafted linear

-1 aliphatic chains of different lengths. The IECs for all samples were around 1.0 mEq g to provide direct comparison to Nafion. Compared to the benchmark BTMA PSf material, the co-grafted membranes exhibited much higher conductivity. Surprisingly, the hydroxide conductivity of the highest performing sample (110 mS cm-1) exceeded that of the proton conductivity of Nafion 112 (100 mS cm-1) at 80°C. The high

72 Chapter 3 conductivity is attributed to the nanoscale hydrophobic-hydrophilic phase separation evidenced by SAXS data and TEM micrographs; these empirical data were consistent with the MD simulation predictions. Aggregation of the hydrophobic grafts results into clusters results in a hydrophilic “ionic highway” in the interstitial regions.

3.3.1.3 Multiblock

Multiblock copolymer AEMs have been explored by several groups as means to induce ion transport nanostructures. These copolymers are typically synthesized from polycondensation of hydrophobic and hydrophilic oligomers. Depending on the end- group functionality of the oligomers, the blocks are distributed in either a perfectly alternative or stochastic fashion along the polymer chain. Hence, the thermodynamics of multiblock copolymers are essentially analogous to those of random copolymers, but with proportionally larger length-scales of separation. Yet, the experimental literature often do not capture this idea and generally treat these random multiblock copolymers in the same qualitative sense as di-block and tri-block copolymers, despite exhibiting fundamentally different thermodynamics.

This idea is manifested in a study by W. Kim et al., where the authors compare random and multiblock poly(arylene ether sulfone) copolymers tetra-functionalized with quaternary ammonium groups.[132] The block copolymers were synthesized from a polycondensation reaction of hydrophilic quaternary ammonium oligo(arylene ether sulfone) with hydrophobic oligosulfone, both with Mn~7000 to 8000 Da; the random copolymer was synthesized from simple monomer constituents. The endgroups of the oligomers were not reaction selective, and as such the oligomers are stochastically distributed along the multiblock copolymer chain in the same manner as the

73 Chapter 3 monomers are in the random copolymer. AFM imaging showed that the block copolymer had larger phase-separated domain; this is not unexpected as the lengthscale of phase-separation is dictated by the characteristic size of the hydrophobic/hydrophilic regions of the polymer chain. Performance-wise, the block copolymer had slightly higher hydroxide conductivity (21 mS cm-1 vs 18 mS cm-1) and lower water uptake (48% vs 59%) compared to the random copolymer at similar conditions (IEC=1.43 mEq g-1, 80°C). Immersion in 1 M KOH at 60°C showed similar rates of degradation, with both samples losing ca. 40 to 50% of their initial conductivity.

Watanabe et al. have prepared several multiblock copolymer materials through coupling of hydrophobic and hydrophilic oligomers, with the intent of localizing ionic groups onto specific segments of the polymer chain. Their prior work with quaternization of tetra-functional fluorenyl groups in poly(arylene ether sulfone)s showed that local concentration of ionic moieties led to enhanced conductivity and dimensional stability.[47] However, that work was based on a pure random copolymer in which hydrophilic and hydrophobic regions were stochastically distributed along the polymer chain; the oligomer approach provides a higher degree of compositional control on the characteristic lengths of hydrophobic and hydrophilic portions of the polymer chain. The chemistry they employed in all their studies theoretically yields a perfectly alternating multiblock copolymer, where a hydrophobic block is always attached to a hydrophilic block, and vice-versa.

In one manifestation, they coupled hydrophobic oligo(arylene ether sulfone ketone)s with hydrophilic oligo(arylene ether sulfone)s with quaternized fluorenyl

74 Chapter 3 groups.[133] The lengths of the hydrophobic and hydrophilic oligomers were varied to study compositional effects on morphology and macroscopic properties. Whereas

TEM imaging of the purely random quaternary ammonium poly(arylene ether sulfone) showed no indication of phase separation, the random multiblock copolymers prepared in this manner showed ionic domains on the order of ~5 nm. Moreover, increased block length, and thereby increased regularity, yielded lower water uptake and higher conductivity. The best performing material had hydrophobic and hydrophilic block

-1 sizes of 16 and 11 repeat units, respectively, with an IEC of 1.93 mEq g . The hydroxide conductivity of this membrane (96 mS cm-1) was found to be 3.2x higher than that of the pure random copolymer (30 mS cm-1) at 40°C. However, prolonged exposure (500 h) to 80°C water revealed a ~60% loss in IEC of this material.

To improve chemical stability, the authors prepared a multiblock copolymer without electron-withdrawing sulfone and ketone groups by coupling a partially fluorinated oligo(arylene ether) hydrophobic block with a quaternary ammonium- functionalized oligo(phenylene) hydrophilic block, producing membranes with overall

-1 [134] IECs of 1.3, 1.8, and 2.0 mEq g . TEM imaging of these AEMs showed a phase-

-1 separated structure, with hydrophilic domains ~ 2 nm. The 2.0 mEq g IEC membrane exhibited a high hydroxide conductivity near 80 mS cm-1 at 30°C with a water uptake of only ~70%, suggesting highly efficient ion transport. Alkaline stability was improved over the sulfone/ketone-containing material, as immersion in 1 M KOH at

40°C showed little change in conductivity after 300 h.

W. Kim et al.[132] investigated a similar set of multiblock copolymers containing tetra-functional fluorenyl groups by polycondensation of two different

75 Chapter 3 oligo(arylene ether sulfone)s with Mn~7000-8000 Da: one with tetra-functional fluorenyl groups for quaternization with trimethylamine (hydrophilic block) and one without (hydrophobic block). The hydrophilic blocks constituted 15, 20, and 25% by

-1 mol corresponding to membranes with respective IECs of 1.14, 1.66, and 1.89 mEq g

; overall Mn were in the 64,700 to 79,300 Da range. A random copolymer (IEC=1.59 mEq g-1) synthesized from small molecules (vs. oligomers) was used a benchmark.

AFM imaging of the multiblock copolymers showed distinct phase separation between hydrophilic/hydrophobic domains. Consequently, compared to the random copolymer, the multiblock copolymer of similar IEC exhibited slightly lower water uptake (32 vs.

42%) and higher hydroxide conductivity (21.37 vs. 17.91 mS cm-1) at 30°C. However, both materials suffered from alkaline degradation, losing 30~40% of their initial conductivity after aging in 1 M KOH at 60°C.

X. Li et al.[135] prepared a polysulfone-based multiblock copolymer AEM consisting of hydrophobic oligosulfone and hydrophilic quaternary ammonium- functionalized oligo(tetraphenyl methyl sulfone). Both oligomers were synthesized to degrees of polymerization (DP) of 8, 15, and 30 with polydispersity indices (PDI) ranging from 1.4 to 2.0. Due to the polycondensation synthesis, the oligomers were distributed in an alternating fashion along the polymer backbone to yield polymers with overall Mn~59,500 to 87,000 and PDI~1.0 to 1.8. Each aromatic ring in the tetraphenyl methyl moieties is capable of functionalization with a quaternary ammonium headgroup, thus providing a high local concentration of ionic charges.

Tapping-mode AFM imaging showed distinct domain formation, with characteristic sizes that scaled proportionally with the length of the oligomer blocks. Samples with

76 Chapter 3 equal DP of the hydrophobic and hydrophilic oligomers were found to offer the highest conductivities (~40-50 mS cm-1 at 60°C). However, longer hydrophobic blocks led to reduced water uptake.

3.3.2 Diblock and Triblock Copolymers

There is a common belief that the well-defined chemical compositions of diblock and triblock copolymers allow precise tuning of polymer morphology and better control of structure-property relationships. Indeed, self-consistent mean field theory of diblock copolymers have led to theoretical predictions, and subsequent experimental verification, of well-ordered morphologies. However, it is often overlooked that, in addition to molecular weight, phase-segregation and nanostructure of these materials are strongly affected by properties such as chain stiffness and polydispersity. In addition, processing conditions can significantly affecy the ultimate structure. In light of this, it is difficult to synthetically realize these materials in a reproducible and scalable fashion. Nevertheless, diblock and triblock copolymers offer a simple conceptual platform for exploring structure-property relationships. Surprisingly, there have been few reports of di/triblock copolymer anion exchange membranes, and fewer still that examine their properties in the hydroxide form.

Balsara et al. prepared a series of di-block polystyrene-block- polychloromethylstyrene copolymers (PS-b-PCMS), where the PCMS block was quaternized to give either benzyltrimethylammonium (BTMA) or n-butylimidazolium

(IM) groups with chloride counterions. The volume fraction of the hydrophilic BTMA and IM blocks ranged from 0.26 to 0.50 and 0.35 to 0.60, respectively. The resulting

AEMs showed lamellar structures regardless of cation, block length, or volume

77 Chapter 3 fraction, although the domain size (ca. 1-5 nm) did scale with the molecular weight of the blocks. The SAXS peaks for the BTMA copolymers were broader than those of the of IM copolymer, suggesting more disperse domain sizes. Despite different headgroups, the polymers exhibited similar water uptakes and ionic conductivities.

Elabd et al. investigated structure-property relationships of poly(styrene-block-

4-vinylbenzylhexylimidazolium) (PS-b-PVBHexIm) with a bis(trifluoroethanesulfonylimide) (TFSI) anion. A cylindrical phase was observed at

PVBHexIm volume fractions below 0.29, a combination of cylindrical and lamellar phases was observed at a volume fraction of 0.34, and a lamellar phase was observed at a volume fraction of 0.50. The nanostructure of the material was found to have profound influence on TFSI conductivity, with the pure lamellar structure exhibiting a conductivity that was an order-of-magnitude greater than that of the lamellar/cylindrical mixture due to the presence of grain boundaries and defects in the latter.

Q. Liu et al.[136] prepared triblock AEMs of polystyrene-block-poly(ethyelene- ran-butylene)-block-polystyrene (SEBS, Mn=118,000) by chloromethylation and quaternization of the aromatic polystyrene groups with trimethylamine, resulting in a hydrophilic-hydrophobic-hydrophilic arrangement. The triblock AEM had a lower- than-average IEC of 0.3 mEq g-1, resulting in both low water uptake (12%) as well as low conductivity (5-10 mS cm-1) . Although the stated purpose of using the tri-block design was to enhance conductivity through a more efficient ion-transport mesostructure, no morphological characterization data were reported.

78 Chapter 3

3.3.3 Theory and Simulation of Phase Separation in Random Copolymers

From the literature review of phase-segregated AEMs, it is readily apparent that the majority of these microstructured copolymers – whether they are graft, multiblock, etc. – can be characterized as a random block copolymer. In fact, there have been few reports of di-block and tri-block copolymers, and none that attempt to tie structure- property understanding with a commercially viable AEM. Yet, at the same time, the fundamental thermodynamics of phase-separation of random copolymers and its concomitant effect on morphology and macroscopic properties have not been adequately explored in the literature. Instead, the vast majority of theory and simulation efforts have focused exclusively on mono-disperse diblock and triblock copolymers, highlighting a disconnect between empirical AEM design and fundamental polymer theory.

However, the few theoretical studies that have been reported offer a glimpse into the fundamental differences between phase-separation in random and diblock copolymers. In contrast to diblock copolymers, where the lengthscale of phase- separation is dictated by the overall length of the polymer chain, the lengthscale of phase-separation in random copolymers is governed by the characeteristic, stochastically-determined size of an oligomeric block. Indeed, the statistical distribution of monomers along the chain, a function of their reactivity ratio, plays a critical role in determining not only the order-disorder transition of random copolymer systems, but also the characteristic lengthscale of their phase-separated structures.

This additional consideration is less relevant in diblock copolymers, where the primary parameter for determining the phase-separated morphology is the fractional

79 Chapter 3 composition of the two different monomer components. Moreover, consistent with experimental results, simulations of random copolymer melts have shown that these systems are dominated by local interactions and lack the long-range order associated with diblock copolymers.

These fundamental studies have been supported by molecular dynamics simulations of specific materials such as Nafion,[137,138] 3M PEMs,[138] and even quaternary ammonium polysulfone.[139] Nafion presents an especially interesting case, as it is a statistical random copolymer with short sulfonic acid side-chains, yet has been experimentally found to exhibit a well-separated morphology. Molecular dynamics simulations of Nafion by Goddard et al. show that the nanostructure of

Nafion is highly dependent on the monomer reactivity ratio between its polar and nonpolar monomers, which governs the statistical size of the average block of a certain monomer type. In agreement with fundamental predictions, low reactivity between the comonomers, and therefore a more blocky system, results in larger heterogeneities (water-rich domains) and better phase-separation.

While these simulations may provide additional insight or alternative perspectives on ion transport in those specific materials, there lacks a general platform of understanding that cohesively ties together the molecular structure of random copolymers, their thermodynamics behavior, and their macroscopic (e.g. ion transport) properties. This is evident in the design of many phase-separated AEMs, where the general design principles are not based on some fundamental understanding of the thermodynamics, but rather on qualitative arguments that lead to structure formation, or, in other cases, a molecular design that happens to yield a phase-separate

80 Chapter 3 morphology. Indeed, there appears to be a general disconnect between experimental

AEM development and random copolymer theory – aside from a study reported by J.

Yan and Hickner,[44] we are unaware of any experimental AEM design that leverages the fact that the stochastic distribution of monomers in random copolymers can affect phase-behavior. In addition, only L. Zhuang et al.[127] rationalized, a priori, a phase- separated AEM deisgn by taking advantage of MD simulations.

One obstacle, perhaps, is that current simulation and theory do not adequately reflect reality. The theory and simulations are typically adapted from those of diblock copolymers and are based on a mean-field treatment of polymer melts with Gaussian- chain statistics. The fundamental issues stem from the difference in the lengthscale of interaction between diblock copolymers and random copolymers. For example, the mean-field approach does not adequately capture the large local density fluctuations near the phase-transition. While this effect may be insignificant in the context of large diblock copolymers, the local interactions associated with random copolymers are highly sensitive to these fluctuating fields. As a result, the mean-field paradigm may not accurately capture the phase behavior of random copolymers. Furthermore, while the Gaussian-chain approximation is adequate for probing polymer behavior at lengthscales much greater than the Kuhn length, it fails to capture proper chain statistics at smaller lengthscales associated with the local interactions in random copolymers. Indeed, our recent work show that chain stiffness can significantly impact the ODT and phase-separated morphology of random copolymer melts. This is a particularly important to understanding and furthering experimental design as random

81 Chapter 3 copolymers can vary in their rigidity depending on their functional groups or backbones.

82 Chapter 4

Chapter 4: Materials and Methods

This chapter provides an overview of common synthesis protocols and characterization methods that were employed in all studies. These protocols have been adapted from the published works that form the basis of Chapter 5 and Chapter 6 and the references within.[140,141] Methods that are specific to a certain study are discussed in detail in the respective chapter.

4.1 Materials

Styrene and divinylbenzene monomers were purchased from Sigma Aldrich. The 4- tert-butylcatechol inhibitors were removed by passing the monomers through an alumina column prior to use. All other chemicals were used as purchased without further purification. Trimethylamine (4.2 M in ethanol), chlorotrimethylsilane, paraformaldehyde, stannic chloride, and 2-hydroxy-2-methylpropiophenone, poly(ethylene glycol) monomethyl ether, and sodium hydride (60% w/w in mineral oil) were purchased from Sigma Aldrich. Udel P-3500 MB8 polysulfone was provided by Solvay Chemicals. Potassium hydroxide and all solvents (chloroform, ethanol, petroleum ether, dimethylformamide, etc.) were purchased from Fisher Scientific.

4.2 Synthesis Protocols

4.2.1 Chloromethylation of Polysulfone

The protocol for polysulfone chloromethylation was adapted from Avram et al.[142]

Chlorotrimethylsilane (42.5 mL) and paraformaldehyde (10.0 g) were added to a round-bottom flask containing a polysulfone solution (14.88 g polysulfone in 750 mL

83 Chapter 4 of chloroform). After 30 minutes of mixing, stannic chloride (0.392 mL) was added dropwise to the reaction solution. The reaction flask was then fitted with a reflux condenser and heated to 50°C in a silicone oil bath; the mixture was stirred with heating for up to 72 hours, depending on the desired degree of substitution. Upon completion, excess reagents were filtered out and the filtrate was precipitated into ethanol with a 3:1 ratio of ethanol to filtrate. The precipitate was then washed with excess ethanol and collected as a white powder. The powder was redissolved in choloroform, repurified following the precipitation process just described and dried in a vacuum desiccator at room temperature for 48 hours to yield chloromethylated polysulfone (CMPSf). The degree of chloromethylation was determined by 1H NMR.

4.2.2 Quaternization of Chloromethylated Polysulfone

Chloromethylated polysulfone prepared via the method presented in Section 3.2.1 was quaternized by treatment with trimethylamine. Specifically, chloromethylated polysulfone was dissolved to 5 to 10 wt.% solution in dimethylformamide, to which was then added 3x molar excess of trimethylamine (4.2 M solution in ethanol). The mixture was stirred for 12-24h at room temperature, after which it was poured onto a glass slide and dried in a vacuum desiccator for 48 h, forming a membrane. The membrane was removed by soaking the slide in de-ionized water and peeling off the edges with the aid of a razor blade.

84 Chapter 4

4.3 Characterization Protocols

4.3.1 1H NMR Characterization

All 1H NMR measurements were performed on a Varian Mercury 400 MHz FT-NMR using deuterated chloroform solvent. Integration values were normalized against either the bisphenol A methyl hydrogens or the β-hydrogens of the sulfone group. A representative NMR spectra of chloromethylated polysulfone is presented in Figure 1.

The degree of chloromethylation (DSCl), which represents an average number of chloromethyl groups per repeat unit, is then determined by: DSCl = A(b)/2 where A(b) is the integration value of peak b in Figure 4.1 upon normalization (i.e., the average number of chloromethyl hydrogens per repeat unit).

Figure 4.1 Representative 1H NMR spectra of a chloromethylated polysulfone. The degree of substitution here is approximately 1.09.

85 Chapter 4

4.3.2 Ion Exchange Capacity (IEC)

The theoretical IEC was calculated by the following equation:

1000 ∗ �� = !" ��!"#"$%& where DSCl is the average number of CH2Cl groups per polysulfone monomer (as

1 determined by H NMR), and MWMonomer is the average molecular weight of a repeat unit.

Experimental IEC was determined using a back-titration method. Here, membranes in the hydroxide form were immersed in 30 mL HCl (0.01 N) for 18 hours. The resulting solution was then titrated with KOH (0.01 N) using phenolphthalein indicator. Following titration, the membrane was rinsed with 18.2

MΩ·cm water, dried in a vacuum desiccator for 24 hours, and weighed. The experimental IEC was then be calculated by:

� � − � � �� = !"# !"# !"#$ !"#$ �!"# where CHCl/CNaOH and VHCl/VNaOH denote the molar concentration and volume, respectively, of HCl or NaOH, and mdry is the mass of the dried membrane following titration.

4.3.3 Water Uptake

The gravimetric water uptake was determined by the following equation:

� −� �� % = 100% ∗ !"# !"# �!"#

For temperature-dependent water uptake, membrane samples were immersed in water baths (18.2 MΩ·cm at room temperature) at set temperatures ranging from 20°C to

86 Chapter 4

80°C. After 45 minutes of immersion, the hydrated samples were removed from the water bath and quickly dabbed with KimWipe to remove surface water. The hydrated mass was then measured gravimetrically using a mass balance. The dimensions of the hydrated samples were then measured and recorded. The dry mass and dimensions of the membranes were determined after drying the samples in a vacuum desiccator for

24 hours.

For water uptake kinetics, dry membrane samples in the hydroxide form were swollen in water for set time intervals, after which the hydrated samples were removed, quickly dabbed with a KimWipe to remove surface water, and weighed. The gravimetric water uptake was then plotted against immersion time.

The λ value is defined as the number molecules per cationic headgroup and is calculated via:

��(%) � = 1000 ∗ ��!!! ∗ �� where the IEC used was based on the experimental titrated value if available.

4.3.4 Small Angle X-ray Scattering (SAXS)

All SAXS measurements were performed on beam line 1-4 at the Stanford

Synchrotron Radiation Lightsource, a Directorate of SLAC National Accelerator

Laboratory. Fully hydrated samples in the chloride form were folded onto themselves to an approximate thickness between 0.5 mm to 1 mm, loaded into windowless Teflon sample holders, and subjected to 5-minute exposure times. The beam center and sample-to-detector distance were calibrated using a silver behenate standard. Radial averaging and background subtraction of the raw scattering data were facilitated by

87 Chapter 4 the Nika package provided by Argonne National Laboratory.[143] All data fitting were done in Igor Pro using an iterative Levenberg-Marquardt algorithm.

4.3.5 Alkaline Stability

Accelerated degradation studies were performed by soaking sections of hydroxide- exchanged membranes in KOH at varying concentrations and temperatures, the specifics of which depended on the membrane material and are consequently detailed in the relevant chapter. Samples were removed at various exposure times and rinsed thoroughly with 18.2 MΩ·cm water prior to measurement. The in-plane conductivity was measured at 22°C. The mass was measured gravimetrically after drying the samples in a vacuum desiccator for 12 to 24 hours.

4.3.6 Thermogravimetric Analysis

Thermogravimetric Analysis (TGA) measurements were performed with a Mettler

Toledo TGA/sDTA 851e. Samples were dried in a vacuum desiccator for 48 hours prior to TGA testing. In a typical test, samples of roughly 3 mg (exact mass measured using the TGA instrument) were loaded into aluminum crucibles and heated from 25

°C to 600 °C, under nitrogen, at a rate of 10°C per minute. Differential TGA (dTGA) plots were generated numerically from the raw data.

4.3.7 Mechanical Testing

Mechanical properties were measured using an Intron 5844 system with tensile load clamps at a crosshead speed of 1 mm min-1. All membranes were exchanged to the hydroxide form prior to mechanical testing. For “dry” testing, hydrated rectangular membrane samples measuring 3 mm x 5 mm were allowed to equilibrate with the

88 Chapter 4 ambient environment (20°C, 35% RH) for 48 hours. Tensile testing was then performed under ambient conditions. For characterizing the fully-hydrated membranes, membranes were swollen in nitrogen-purged water for 10 minutes and then cut into 3 mm x 5 mm strips. Tensile testing was performed with the samples submersed in liquid water (20°C) using a BioPuls bath attachment. The elastic modulus was determined from the initial slope of the stress-strain curve according to the ASTM D882 protocol. The average modulus, strength, and elongation-at-break and standard deviations thereof were determined from testing five different samples at each composition.

4.3.8 Conductivity Measurements

In-plane conductivity was measured using a BekkTech BT-552 Conductivity Test

System. In a typical test, a 5 mm wide strip of the polyelectrolyte film to be tested was placed in a BekkTech BT-112 conductivity cell with a 4-point probe setup. The cell was then loaded into a Fuel Cell Technologies 5 cm2 fuel cell test fixture under a 300

SCCM stream of fully hydrated nitrogen gas. A cyclic DC sweep between -0.15 to

+0.15 V was applied with a Keithley 2400 sourcemeter. The resulting voltage-current data were linearly fit to determine the overall resistance of the film. The in-plane conductivity σ was then calculated from the measured resistance, R, given the film thickness, T, film width, W, and inter-electrode distance, L, via the following equation:

� � = � ∙ � ∙ �

The temperature dependence of the conductivity was determined by ramping up the temperature in 5°C to 10°C intervals, holding the film at each temperature setpoint for

89 Chapter 4

60 minutes to allow equilibration; the temperature of the water saturators were set equal to the temperature of the test cell to maintain 100% RH.

4.3.9 Fuel Cell Tests

Single cell performance was evaluated using membrane electrode assemblies (MEA), the preparation and construction of which are described as follows. Anode catalyst ink was prepared by combining 11 mg of Pt/C (TKK TEC10E50E) and 26 µL ionomer solution (Fumion FAA-3) in 413 µL of a 60% v/v aqueous isopropanol solution. The cathode ink was prepared by combining 25 mg of PtIr/C (ETEK) and 59 µL ionomer solution in 938 µL of a 60% v/v aqueous isopropanol solution. The inks were sonicated for 10 minutes at room temperature and then painted onto 6.25 cm2 squares of woven carbon paper (Sigracet GDL35 BC) to a Pt loading of 0.5 mg cm-2 for the anode and 2 mg cm-2 for the cathode. The MEA was prepared by sandwiching the hydroxide-exchanged polyelectrolyte film between the anode and cathode in a Fuel

Cell Technologies 5 cm2 fuel cell test fixture, with Teflon gaskets (0.25 mm thick) at both electrodes to prevent puncturing by the serpentine flow channels.

MEA fuel cell performance was evaluated using a BekkTech BT-552 test system with a water-saturated gas flow rate of 200 SCCM O2 at the cathode and 150

SCCM H2 at the anode. An Agilent 6060B load box was used to apply a set load from

1.0 V to 0.200 V in increments of 0.05 V; the current was recorded after 60 seconds at each load setpoint to generate a polarization curve.

90 Chapter 5

Chapter 5: A Semi-Interpenetrating Network Approach for

Stabilizing Highly Charged Anion Exchange Membranes for

Alkaline Fuel Cells

Note: This chapter has been adapted from He, S. S., Strickler, A. L. and Frank, C. W.,

ChemSusChem 2015 (8), pp. 1472-1483

5.1 Introduction

As discussed in Chapter 2, the unfavorable performance of AEM fuel cells can be partially attributed to the lower ionic conductivities of typical AEM materials, which are often several times lower than those of Nafion and other PEMs.[144] Attempts to resolve the performance issue typically focus on either modifying the chemistry of the pendant cation or changing the identity of the polymer backbone. For example, by replacing the typical quaternary ammonium cation with a more basic quaternary phosphonium cation, Gu et al.[71,72] were able to increase room-temperature conductivities from around 10 mS cm-1 to 38 mS cm-1 . More recent reports have begun to focus on effecting a more efficient ion transport architecture for hydroxide transport. Li et al.[123,124] showed that using a quaternary ammonium cation with a short (8-16 carbon) aliphatic chain prompts microphase separation and higher performance. Pan et al.[127] reported that grafting linear aliphatic chains along the main polymer backbone promotes hydrophobic clustering, resulting in the creation of an ion transport “highway” in the interstitial, water-rich regions. Moreover, we recently

91 Chapter 5 reported that grafting poly(ethylene glycol) chains onto quaternary ammonium polysulfone results in nanophase separation between the hydrophilic grafts and the hydrophobic polymer backbone, producing water-rich ion transport channels and, consequently, higher ion transport efficiency.[140]

However, the larger size of the hydroxide ion relative to that of protons results in slower diffusion kinetics, fundamentally limiting the performance of AEMs compared to PEMs. Indeed, the dilute solution mobility of the hydroxide anion is only

57% of that of the proton.[70,145] Hence, to first approximation, for a given polymer system the hydroxide concentration must be roughly 1.8 times greater than the proton concentration to yield similar conductivities. This effect is evident from the literature, where many high-performing AEMs must have ion exchange capacities much higher than Nafion’s to achieve similar performance.[20,144,146,147] Of course, one might imagine that increasing the IEC even further could lead to even higher performance; however, the higher ion content typically leads to increased water uptake and swelling, yielding diminishing returns on effective ion concentration as a result of dilution.

Moreover, at excessively high IECs, the membranes simply lose mechanical integrity and rupture from the uptake of too much water.[20,144]

Here, we report the ability to significantly reduce water swelling and enhance the mechanical strength of highly charged anion exchange membranes by reinforcing the linear polyelectrolyte chains with a crosslinked matrix of a robust hydrophobic material, poly(styrene-co-divinylbenzene), essentially creating a semi-interpenetrating network (semi-IPN). A semi-IPN is defined as a system in which a linear polymer is homogeneously dispersed (at least on the length-scale of the polymer chains) within a

92 Chapter 5 covalently crosslinked polymer network; semi-IPNs have been reported previously in the fuel cell literature for enhancing the methanol resistance of Nafion and other

PEMs.[148–152] Additionally, Wang et al.[153] synthesized a semi-IPN quaternized chitosan-polystyrene anion exchange membrane, although in that study they crosslinked the conductive chitosan network and used the polystyrene as a linear hydrophobic component. The approach we present here is applicable to a variety of anion exchange membranes already reported in the literature, provided that the polyelectrolyte backbone has suitable compatibility with styrene and divinylbenzene.

Specifically, we chose benzyltrimethylammonium polysulfone (QA PSf) as a benchmark material due to its ubiquity in the literature[31–33,38,39,70,154] and incorporated a polystyrene (PS) or poly(styrene-co-divinylbenzene) (PS-co-DVB) network as a structural “scaffold” to limit swelling and enhance mechanical integrity. We found

-1 that films formed from a QA PSf material with an IEC of 2.99 mEq g ruptured and solubilized in water, even under ambient temperatures; however, the addition of a secondary hydrophobic polymer provided mechanical stability even up to 80°C, depending on composition. Furthermore, the presence of this secondary hydrophobic network did not greatly diminish the in-plane hydroxide conductivity. All membranes exhibited room-temperature hydroxide conductivities between 38 mS cm-1 and 50 mS cm-1 and showed similar Arrhenius activation energies, suggesting that the semi-IPN structure did not introduce excessive tortuosity in the ion transport pathways of the conductive quaternary ammonium polysulfone.

93 Chapter 5

5.2 Results and Discussion

5.2.1 Synthesis and Characterization

Detailed synthesis and characterization methods can be found in the Experimental section. Benchmark benzyltrimethylammonium polysulfone membranes QA PSf-299

(IEC=2.99 mEq g-1) and QA PSf-226 (IEC=2.26 mEq g-1) were prepared by quaternization of chloromethylated polysulfone (CMPSf) with trimethylamine and solvent casting onto a glass slide.

The same CMPSf used for QA PSf-299 formed the basis for the semi-IPN membranes, which were prepared by swelling CMPSf films in a solution of styrene, divinylbenzene (DVB), and 2-hydroxy-2-methylpropiophenone (HMPP) photoinitiator. The composition of the monomer soaking solutions is presented in

Table 5.1. The monomer-swollen membrane was then subjected to UV irradiation

(365 nm) to promote photopolymerization. The films were finally soaked in a trimethylamine/ethanol solution to convert the labile chloromethyl groups into quaternary ammonium moieties.

An illustration of a semi-IPN AEM prepared by this method is presented in

Scheme 5.1. The semi-IPN membranes are denoted as QA sIPN [X/Y/Z], where ‘X’,

‘Y’ and ‘Z’ denote the wt.% of chloromethylated polysulfone, styrene and divinylbenzene, respectively; the total monomer content (styrene and divinylbenzene) was determined gravimetrically after UV polymerization, while the specific wt.% of styrene and divinylbenzene was estimated from their volumetric ratio in the soaking solutions. Note that QA sIPN [76/24/0] is, in a strict sense, a polymer blend and not a

94 Chapter 5 semi-interpenetrating network as there are no covalent crosslinks; we chose this nomenclature for the sake of consistency.

Table 5.1 Composition of monomer soaking solutions and sample nomenclature. Sample Styrene [mL] Divinylbenzene [mL] HMPP [µL] QA sIPN [76/24/0] 5.00 0.00 200 QA sIPN [76/23/1] 4.75 0.25 200 QA sIPN [73/22/5] 4.00 1.00 200 QA sIPN [72/20/8] 3.50 1.50 200

CH3 O O O S CH3 O m CH2 + H3C N CH3 - CH3 OH

Scheme 5.1 Illustration of semi-IPN AEM. Blue lines represent quaternary ammonium polysulfone, red lines represent poly(styrene-co-divinylbenzene), and hexagons represent covalent divinylbenzene crosslinks.

Polymerized films of the monomer soaking solutions were prepared by heating the solutions at 80°C for 2 hours and subjected to TGA to verify that inhibitors had been successfully removed and to characterize the degradation temperatures for polystyrene (PS) and poly(styrene-co-divinylbenzene) (PS-co-DVB).

95 Chapter 5

Figure 5.1 Differential TGA plot of polymerized monomer soaking solutions.

A differential TGA (dTGA) plot of the polymerized soaking solutions (Figure

5.1) shows a mass loss peak in the 390°C to 430°C range, corresponding to the degradation of polystyrene and/or poly(styrene-co-divinylbenzene). Moreover, because styrene and divinylbenzene have boiling points of 145°C and 195°C, respectively, the lack of mass loss below 200°C suggests complete polymerization of the monomers. The sample resulting from polymerization of the pure styrene solution, which contained no divinylbenzene crosslinking functionalities, showed a characteristic mass loss temperature of 391°C. The solutions that contained DVB all show a notable upshift in the degradation temperature to the 410°C to 430°C range as a result of stabilization by the covalent crosslinking. As expected, increasing the degree of crosslinking by increasing the proportion of DVB-to-styrene further raises the degradation temperature. The 95%/5% styrene/divinylbenzene solution showed a characteristic degradation temperature around 415°C. This shifts to 421°C for the

80%/20% solution and to 425°C for the 70%/30% solution as the higher DVB concentration leads to a greater extent of crosslinking. These data are consistent with

96 Chapter 5 previous literature on the thermal stability of polystyrene-divinylbenzene copolymers.[155]

Figure 5.2 Differential TGA plot of QA PSf-226, QA sIPN [76/24/0] and QA sIPN [72/20/8] alkaline exchange membranes.

A similar characterization was carried out for the membranes, where dTGA data for the semi-IPN films were compared with the unmodified QA PSf-299 membrane to confirm successful polymerization of the styrene and DVB monomers within QA PSf-299 (Figure 5.2). All membranes share a mass loss below 100°C as a result of losing residual water and around 180°C due to decomposition of the benzyltrimethylammonium sidegroups.[39,156] The unmodified QA PSf-299 sample shows a broad mass loss above 350°C, which is ascribed to thermal decomposition of polysulfone. In comparison, QA sIPN [76/24/0] exhibits a sharp mass loss at a characteristic degradation temperature of 410°C, attributed to the presence of the polystyrene; the higher apparent characteristic degradation temperature compared to bare polystyrene (391°C as reported in the previous paragraph) is due to overlap with the thermal degradation of the polysulfone. As with the dTGA of the bare solutions

97 Chapter 5

(Figure 5.1), the introduction of divinylbenzene in QA sIPN [72/20/8] results in covalent crosslinks and an upshift in the degradation temperature to 428°C. The presence and composition-dependent behavior of these mass loss peaks suggests successful polymerization of the styrene/DVB monomers within the QA PSf-299 membrane.

5.2.2 Monomer Uptake and Ion Exchange Capacity

Poly(styrene-co-divinylbenzene) was chosen as the reinforcing matrix due to its excellent thermal, mechanical, and chemical robustness. Moreover, the structures of the styrene and divinylbenzene monomers are sufficiently similar to each other and to the polysulfone backbone to facilitate homogeneous mixing between the various components. This property is manifest in their similar solubility parameters: polysulfone has a solubility parameter[157] of 19.9 (J/cm3)1/2 while styrene and divinylbenzene have solubility parameters[158] of 17.8 (J/cm3)1/2 and 17.4 (J/cm3)1/2, respectively, as estimated by the Hansen group contribution method. At the same time, the difference in solubility parameter between polysulfone and styrene/divinylbenzene is sufficiently high enough to prevent complete dissolution of the membrane when placed in a monomer solution. Consequently, after soaking for 24 hours in a styrene and/or divinylbenzene solution, all CMPSf films showed monomer uptake ranging from 30% to 40% by mass (Figure 5.3). Moreover, the swollen CMPSf films remained clear and colorless, suggesting there is no macroscale demixing and heterogeneity between the CMPSf and the styrene and/or divinylbenzene monomers.

98 Chapter 5

Figure 5.3 Left Axis (Red Bars) - Gravimetric monomer uptake of the CMPSf films after soaking in styrene/divinylbenzene monomer solution for 24 hours. Right Axis (Blue Squares)

– Theoretical IEC as calculated by the monomer mass uptake based on the maximum IEC of

-1 2.99 mEq g for unmodified QA PSf-299.

Given the small difference in the solubility parameter between styrene and

DVB, the increased gravimetric monomer uptake on increasing the concentration of

DVB cannot be attributed to more preferable interaction between the polysulfone and

DVB. Instead, the increased mass uptake is due to the higher molecular weight of

DVB (130 g/mol) compared to styrene (104 g/mol). Assuming that the ratio of styrene to divinylbenzene in the film is equivalent to that of the swelling solution, all samples showed similar molar uptake (3.0 to 3.4 mmol per gram of CMPSf). Consequently, given the same initial theoretical IEC, the increased mass from monomer uptake leads

-1 to a decrease in the overall IEC of the membranes from 2.28 mEq g in the QA sIPN

-1 [76/24/0] to 2.15 mEq g in QA sIPN [72/20/8]. In comparison, the unmodified QA

-1 PSf-299 and QA PSf-226 benchmark materials have theoretical IECs of 2.99 mEq g and 2.26 mEq g-1, respectively.

99 Chapter 5

5.2.3 Water Uptake and Temperature Stability

Increasing charge concentration (i.e., IEC) can potentially lower resistivity by introducing more free ions in the system. At the same time, however, the favorable solvation energy of ion pairs results in a significant increase in water uptake upon increasing the IEC, leading to excessive swelling, mechanical weakening, and eventual catastrophic rupturing of the membranes. Moreover, as water uptake typically scales disproportionately with increasing IEC, dilution effects begin to take place at higher charge contents. Our blend and semi-IPN membranes offer a solution to this issue, as the secondary hydrophobic component limits water uptake (Figure 5.4) and constrains dimensional swelling (Figure 5.5). The absence of data for the QA PSf-299 starting material is a result of complete mechanical failure and dissolution of the membrane in water, even at room temperature. In contrast, our modified membranes that incorporate the linear QA PSf-299 within either a polystyrene or poly(styrene-co- divinylbenzene) matrix remained as tough flexible films upon hydration, underscoring the effectiveness of this approach.

Figure 5.4 Gravimetric water uptake as a function of temperature.

100 Chapter 5

Figure 5.5 Thickness swelling of membranes at equilibrium water uptake at different temperatures.

As a result of the instability of QA PSf-299, the lower IEC QA PSf-226

-1 membrane was used as the primary benchmark; its 2.26 mEq g IEC is in the vicinity

-1 of the blend and sIPN materials, which range from 2.28 mEq g for QA sIPN

-1 [76/24/0] to 2.15 mEq g for QA sIPN [72/20/8]. We note that the introduction of a linear hydrophobic component (polystyrene) to create a blend results in a significant decrease in the water uptake, especially at higher temperatures. For example, the room temperature water uptake is reduced from 301% (QA PSf-226) to 227% (QA sIPN

-1 [76/24/0]) despite both membranes having similar theoretical IECs (2.26 mEq g and

2.28 mEq g-1, respectively). More importantly, the QA PSf-226 benchmark completely ruptures above 50°C; the membranes became extremely fragile around this temperature and often fractured when measuring their thickness. In contrast, QA sIPN

[76/24/0] maintains its mechanical integrity even at 80°C. Introducing DVB crosslinks further reduces water uptake, where even a small amount of DVB content in QA sIPN

[76/23/1] is able to substantially reduce the room temperature WU to 123%, roughly

101 Chapter 5 half of that of QA sIPN [76/24/0]. This effect is enhanced at higher DVB contents where QA sIPN [72/20/8] exhibits 76% water uptake at room temperature, nearly a 4x decrease over the unmodified QA PSf-226 material.

To better elucidate the effects of styrene and DVB content on the temperature dependence of water uptake, we define a normalized water uptake value

WU(T) = 100 ! ! -!(°) (1) !"#$ !(°) where m(T) represents the hydrated mass at temperature T. This normalized value,

WU(T)norm, reflects the percent gain in mass of the fully hydrated sample at temperature T over its fully hydrated mass at room-temperature and better illustrates the temperature response in water uptake (Figure 5.6).

The effect of temperature on water uptake, as shown in Figure 5.6, can be categorized into three distinct temperature regimes: Region 1 between room temperature and 35°C where there is a notable increase in the water uptake for all samples; Region 2, between 35°C and 65°C, where the water uptake shows a limited dependence on the temperature, resulting in a plateaued temperature response; and finally Region 3, above 65°C, where the water uptake begins to again increase notably with increasing temperature.

102 Chapter 5

Figure 5.6 Temperature response of water uptake as normalized against the room temperature hydrated mass.

The extent to which each membrane’s water uptake thermally responds is highly dependent on composition. Our rationalization for the composition-dependent response in each of the regions, as well as for the shared features in the three regions, is based on the competition between the free energy associated with the osmotic pressure exerted by free ions within the membrane and the energy required to dimensionally swell the polymer matrix. The increase in water uptake between room temperature and 35°C (Region 1) is attributed to a higher degree of ion solvation on increasing the temperature; the polymer matrix in this region is sufficiently compact such that the energy required to expand the matrix is lower than the energy gained from additional water incorporation. This hypothesis is consistent with the composition dependence of the trend in water uptake, where membranes with a higher concentration of chemical crosslinks (inferred from the DVB content) show the lowest overall water uptake increase because the chemical crosslinks inhibit chain mobility.

103 Chapter 5

Following this rationale, the plateau in Region 2 is ascribed to finite extensibility of the polymer chains. The high extension of the polymers leads to a significant decrease in the number of accessible states, leading to large entropic losses; the energy required to strain the polymer network beyond a certain extension, therefore, becomes exponential with network strain. Consequently, the gravimetric water uptake exhibits minimal changes in this region as the energy released from additional water uptake is lower than that required to expand the polymer network.

In Region 3, all the samples, save for the highly crosslinked QA sIPN

[72/20/8], exhibit a large deviation from this plateau behavior at temperatures above

65°C, resulting in escalated water uptake at higher temperatures. We suspect that this increase can be attributed to increased mobility of the polymer chains. Specifically, this thermal transition in the water uptake behavior coincides with the β transition

[159,160] temperature for atactic polystyrene (ca. 55°C to 65°C). This sub-Tg transition is a result of local reorientation and rotation of the phenyl rings, leading to overall conformational changes in the polystyrene backbone and, consequently, localized cooperative motion of the polystyrene chain. Hence, the β relaxation of the secondary polystyrene network, coupled with its thermodynamic mismatch with quaternary ammonium polysulfone, leads to chain migration and relaxation of the previously strained network, resulting in increased water uptake. The extent of this effect is, as expected, diminished with increasing degree of chemical crosslinking, as highlighted by the fact that the highly crosslinked QA sIPN [72/20/8] sample did not exhibit a significant increase in water content at 80°C.

104 Chapter 5

Thus, by introducing polystyrene and poly(styrene-co-divinylbenzene) networks, we were able to reinforce overall membrane structure, as exemplified by the decrease in gravimetric water uptake. Moreover, we note that the introduction of the

DVB crosslinks significantly limits membrane swelling at room temperature as well as water uptake at elevated temperatures.

5.2.4 Swelling Kinetics

Time-dependent water uptake data revealed second-order uptake kinetics, where the initial high rate of swelling becomes increasingly retarded by chain-stretching, asymptotically approaching equilibrium. Consequently, we employed Schott’s model for second-order swelling kinetics[161] to probe the effects of composition on the rate of water absorption within the films (Figure 5.7). The Schott model explains the empirical second-order behavior by assuming that the observed swelling rate is directly proportional to the remaining swelling capacity. This is mathematically expressed as:

2 dW(t) #W(∞)−W(t)& (2) = K % ( dt $ W(∞) ' or, solving and rearranging, as:

t 1 t (3) = 2 + W(t) K ⋅W(∞) W(∞)

where W(t) is the water uptake at time t, W(∞) is the water uptake at equilibrium, and

K is the intrinsic rate constant for water swelling. Replotting the water uptake kinetic

105 Chapter 5 data in the form of Equation 3 shows excellent agreement with the Schott model

(Figure 5.8).

Table 5.2 Intrinsic rate constant for water uptake at 20°C. -1 -1 Sample K [gpolymer gwater s ] QA PSf-226 0.020 QA sIPN [76/24/0] 0.132 QA sIPN [76/23/1] 0.670 QA sIPN [73/22/5] 0.703 QA sIPN [72/20/8] 0.721

The intrinsic rate constants for water uptake in the different membranes were calculated from the intercept of the linear correlation and equilibrium water uptake

(Table 5.2). We note that the unmodified QA PSf-226 membrane exhibited the lowest

-1 -1 rate constant (0.02 gpolymer gwater s ). The incorporation of a hydrophobic polystyrene

-1 -1 component (QA sIPN [76/24/0]) increases the rate constant to 0.132 gpolymer gwater s , while the addition of divinylbenzene crosslinks further increases the rate constant to

-1 -1 between 0.67 and 0.72 gpolymer gwater s . This trend is rationalized through composition-dependent chain relaxation processes, where QA PSf-226 shows the slowest water uptake rate as it is limited by the stress-relaxation kinetics of the linear polymer chains, an effect that is exacerbated by the large dimensional swelling of the film. Introducing a hydrophobic component reduces water uptake and consequently reduces strain, limiting the influence of stress-relaxation. Finally, covalent crosslinking constrains the overall system and inhibits segmental motion of the polymer chains, resulting in reduced swelling and a more elastic mechanical behavior; consequently, the water diffusion process becomes decoupled from the large-scale reorientation of the polymer chains, leading to a high intrinsic rate of water uptake.

106 Chapter 5

Figure 5.7 Water uptake kinetics for the AEMs.

Figure 5.8 Water uptake kinetics for the AEMs plotted according to the Schott second-order kinetics model. The dashed lines represent linear fits, showing excellent agreement to the

-1 Schott model. The water uptake here is defined as a ratio (gwater gpolymer ) instead of a percentage to facilitate calculation of the intrinsic rate constant.

107 Chapter 5

5.2.5 Mechanical Properties

Pure polystyrene and poly(styrene-co-divinylbenzene) are brittle polymers with high elastic moduli and exhibit little plastic deformation. We performed tensile tests to explore how their incorporation in the semi-IPN architecture affects the mechanical properties of the ionomer membranes in both the dry and fully-hydrated states. Because the QA PSf-299 starting material was extremely delicate under these conditions, QA PSf-226 was again used for comparison.

Table 5.3 Mechanical properties of dry membranes (20°C and 35% RH). Sample Modulus [MPa] Tensile Strength [MPa] Max Strain [%] QA PSf-226 277±15 9.6±1.2 9.5±1.3 QA sIPN [76/24/0] 326±23 13.2±1.6 7.1±2.2 QA sIPN [76/23/1] 401±37 16.4±2.8 6.7±1.3 QA sIPN [73/22/5] 783±27 18.8±1.3 6.3±1.1 QA sIPN [72/20/8] 1243±42 21.2±1.9 5.4±0.8

The mechanical properties of the dry membranes were determined after equilibration with the ambient environment (20°C, 35% RH); characterization was performed under the same conditions and the results are presented in Table 5.3. The introduction of a PS or PS-co-DVB network results in increased elastic modulus and tensile strength compared to a linear QA PSf-226 membrane of similar IEC. As hinted at by the water-uptake kinetics discussed previously, the stress-strain curves of the semi-IPN membranes show a highly elastic response compared to the plastic behavior of the linear QA PSf-226 membrane. At higher DVB content there is a significant increase in both the modulus and the tensile strength as a result of more crosslinking, with QA sIPN [72/20/8] exhibiting an approximately 550% higher modulus and 100%

108 Chapter 5 higher tensile strength than QA PSf-226. At the same time, as a result of the brittle character of both PS and PS-co-DVB, the elongation at break of the semi-IPN membranes was roughly 50% to 60% of the linear QA PSf-226 membrane. The combination of these effects led to qualitatively stiff semi-IPN membranes in the dry state, breaking upon bending beyond ~15°; however, they were nevertheless flexible enough to be mechanically stable under careful handling and were robust enough to be loaded into the tensile testing clamps without fracturing.

Table 5.4 Mechanical properties of hydrated membranes (20°C in water bath). Sample Modulus [MPa] Tensile Strength [MPa] Max Strain [%] QA PSf-226 3.50±0.32 0.20±0.13 12.1±6.3 QA sIPN [76/24/0] 10.1±2.6 1.15±0.29 14.5±3.1 QA sIPN [76/23/1] 21.0±2.8 3.81±0.75 14.8±2.6 QA sIPN [73/22/5] 73.8±4.7 8.32±0.93 13.6±2.7 QA sIPN [72/20/8] 96.8±8.5 10.2±1.3 12.5±1.9

Tensile tests of the hydrated membranes were performed inside a water bath at

20°C (Table 5.4). Water-induced plasticization is a well-known phenomenon in polyelectrolyte membranes[24,115] and is manifested here as an increase in the maximum strain and a decrease in the modulus and tensile strength across all samples upon hydration. Most notably, as a result of high water uptake (301%), the linear QA

PSf-226 material was fragile with a tensile strength below 0.5 MPa and an elastic modulus of only 3.5 MPa. In comparison, the QA sIPN [72/20/8] material’s modulus and tensile strength were nearly two orders of magnitude higher. Despite its relatively low strain-at-break, the ~100 MPa modulus and ~10 MPa strength of QA sIPN

109 Chapter 5

[72/20/8] is comparable to those of fully-hydrated Nafion 117 (114 MPa modulus and

14 MPa tensile strength)[162].

We present a simple scaling analysis to illustrate the structure-property relationship in the crosslinked membranes. Intuitively the elastic modulus should increase with a higher crosslink density and a lower degree of swelling. Accordingly, the elastic modulus, E, of the hydrated materials scales with ρx, the crosslink density,

[163,164] and Φp, the volume fraction of dry polymer at equilibrium swelling, as follows :

!/! !!/! �~�!�! ~�!� (4)

Assuming that the dry polymer has a similar mass density as water and that the ideal crosslink density scales directly with the volume fraction of divinylbenzene, we arrive at the following approximate scaling for the elastic modulus:

!"#$% !!/! � ~�!"#� (5) where ΦDVB is the volume fraction of DVB and Q is the swelling ratio of the total hydrated mass to the dry mass of the material at equilibrium water uptake.

Table 5.5 Comparison between empirical and predicted elastic modulus of hydrated semi-IPN materials. -1/3 ideal Sample ΦDVB Q E [MPa] E [MPa] QA sIPN [76/23/1] 0.01 0.765 21.0±2.8 - QA sIPN [73/22/5] 0.05 0.807 73.8±4.7 88.6 QA sIPN [72/20/8] 0.08 0.830 96.8±3.5 136.6

Using this scaling argument (Equation 5) and assuming that the volume fraction can be approximated by the mass fraction due to similar densities, we

110 Chapter 5 predicted the modulus for QA sIPN [73/22/5] and QA sIPN [72/20/8] based on the experimentally measured modulus for QA sIPN [76/23/1]. Table 5 presents the comparison of the predicted modulus, which is based solely on ideal scaling with respect to DVB content and swelling ratio, to the empirical modulus as determined by tensile testing. The estimated elastic modulus based on this crude scaling analysis is close to the experimentally measured values, underscoring the idea that an increase in

DVB content of the soaking solution directly contributes to a higher crosslinking density and enhanced mechanical properties.

5.2.6 Anion Conductivity

The in-plane hydroxide conductivities (σ) of the membranes as a function of temperature are presented in Figure 5.9. Again, QA PSf-299 data could not be included due to dissolution at room temperature. In order to inhibit conversion of the hydroxide ions into carbonate/bicarbonate, all measurements were performed in an enclosed chamber under hydrated nitrogen gas flow.

The QA PSf-226 membrane measured a hydroxide conductivity of 38 mS cm-1 at 30°C, consistent with literature reports of QA PSf materials with similar IECs.[114]

However, the material exhibited extremely poor temperature response, showing a slight drop in the hydroxide conductivity as it was heated up to 55°C. This is attributed to the excessive water uptake and swelling of the membrane, which reaches over

400% at 40°C and 1600% at 55°C. Indeed, raising the temperature above 55°C resulted in rupturing of the membrane due to excessive water uptake. In contrast, introducing a hydrophobic polystyrene component in QA sIPN [76/24/0] provides conductivity stability up to 65°C by reducing swelling. Although the conductivity

111 Chapter 5 begins to drop after 65°C, the QA sIPN [76/24/0] membrane nonetheless remains mechanically stable and does not exhibit the catastrophic failure of QA PSf-226.

Figure 5.9 In-plane hydroxide conductivity as a function of temperature (100% RH). Dashed lines are for guiding the eye.

The introduction of DVB to form a crosslinked poly(styrene-co- divinylbenzene) matrix around the conductive quaternary ammonium polysulfone

(creating a semi-IPN architecture) further enhances temperature stability, with QA PSf sIPN-80/20 and QA PSf sIPN-70/30 exhibiting stability even at 80°C despite their high IECs. In particular, QA PSf sIPN-70/30 showed the greatest absolute conductivity, measuring 89 mS cm-1 at 80°C. Ultimately, the conductivity data verifies that the increased stability and decreased water uptake obtained through adoption of the semi-IPN architecture translates to improved performance and better thermal stability.

112 Chapter 5 σ

Figure 5.10 Arrhenius plot of hydroxide conductivity at 100% RH. Dashed lines represent

Arrhenius fit in the temperature range where Arrhenius scaling is observed.

An Arrhenius plot of the conductivity-temperature relationship is provided in

Figure 5.10. Interestingly, deviation from Arrhenius behavior is correlated with the temperatures at which the water uptake exits the plateau region described previously.

For example, QA sIPN [76/23/1] begins showing a decrease in conductivity after around 60°C to 70°C, which corresponds to the temperature range in which the water uptake begins to sharply increase again (Figure 6). We suspect that these two behaviors are intrinsically tied, wherein the proposed β relaxation of poly(styrene-co- divinylbenzene) leads to reorientation of the secondary network, consequently altering the ion transport morphology. We suspect that this chain migration, coupled with dilution of charges from increased water uptake, leads to the observed non-Arrhenius behavior. This effect is underscored by comparing the water uptake and conductivity behavior for QA sIPN [72/20/8], a sample in which the large degree of crosslinking inhibits migration of the poly(styrene-co-divinylbenzene) network. These crosslinks

113 Chapter 5 lead to a continued plateau behavior in the water uptake at temperatures greater than

65°C and a concomitant adherence to Arrhenius scaling of the conductivity.

We were initially concerned that the presence of a crosslinked secondary network may introduce increased tortuosity in the ion transport pathways, leading to lower ionic conductivities and a trade-off in mechanical stability versus performance.

Experimentally, however, we found that was not the case, with all samples exhibiting similar room-temperature conductivities. The independence of room-temperature conductivity from both the presence and composition of the secondary network suggests that the hydrophobic reinforcing scaffold is sufficiently phase-separated from the hydrophilic ion transport domains as to not interfere with the ion transport mechanism. This is further evidenced in the fact that all semi-IPN membranes exhibit activation energies around 11 kJ/mol as calculated in the temperature regime where

Arrhenius behavior is present, suggesting similar ion transport mechanisms.

Moreover, these values for the activation energy are comparable to those for hydroxide transport in aqueous solution,[7,145] again suggesting that the hydrophobic poly(styrene-co-divinylbenzene) network has little influence on ion transport morphology at the high IECs and water uptakes investigated here.

Bicarbonate and chloride conductivities measured at 20°C and 100% RH

(water-saturated nitrogen gas) are presented in Table 6. Note that the QA PSf-226 has greater conductivity than the semi-IPN membranes when in the chloride form, in contrast to the trend in the hydroxide and bicarbonate forms. This is likely due to the much lower water uptake exhibited by membranes with the chloride counterion, mitigating issues resulting from charge dilution and dimensional stability.

114 Chapter 5

Table 5.6 AEM conductivity at 20°C and 100% RH for different counterions. - -1 - -1 Sample HCO3 [mS cm ] Cl [mS cm ] QA PSf-226 10.6 13.3 QA sIPN [76/24/0] 10.3 11.1 QA sIPN [76/23/1] 11.2 10.5 QA sIPN [73/22/5] 11.4 10.2 QA sIPN [72/20/8] 12.1 10.3

5.2.7 Leaching and Alkaline Stability

The long-term stability of AEMs is critical to their device viability. Given that the linear ionomer component (i.e., QA PSf) within the semi-IPN membranes presented here is chemically decoupled from the crosslinked poly(styrene-co-divinylbenzene) matrix, gradual demixing and leaching out of the ionically conductive QA PSf component from the crosslinked PS-co-DVB network presents a valid concern. This concern is exacerbated by recent literature reports[24,25] which suggest that, upon cationic functionalization, the polysulfone backbone itself becomes vulnerable to nucleophilic attack; backbone cleavage would yield smaller fragments that would exhibit faster phase separation and migration from the crosslinked matrix.

To characterize the leaching stability, we monitored changes in both the mass and conductivity of the QA sIPN [72/20/8] membrane under prolonged conductivity testing at 40°C and 100% RH (Table 5.7). Because the hydroxide anion is vulnerable to conversion to bicarbonate and carbonate, the membrane was kept in the chloride form for the long-term leaching test. Over the course of 20 days, we found negligible changes in either mass or conductivity, suggesting that leaching of the active QA PSf

115 Chapter 5 polyelectrolyte from the crosslinked PS-co-DVB matrix is insignificant during extended operation under aqueous conditions.

Table 5.7 Leaching Stability of QA sIPN [72/20/8] in Cl- form. 1 Day 5 Days 10 Days 15 Days 20 Days % Initial Mass 100% 99.7% 100% 100% 99.3% % Initial σ 99.0% 97.8% 97.1% 98.5% 97.6%

Quaternary ammonium polysulfone is known to exhibit chemical and mechanical degradation when exposed to highly alkaline environments. For example, the pendant benzyltrimethylammonium cation is vulnerable to nucleophilic attack by hydroxide anions.[165,166] Furthermore, recent reports have shown that the electron withdrawing effect of pendant cations makes the polysulfone backbone susceptible to hydrolytic cleavage, resulting in the loss of mechanical integrity.[22,25] σ

Figure 5.11 Alkaline stability of QA PSf-226 and QA sIPN [72/20/8] in 6 M KOH solution at

40°C.

To investigate whether the semi-IPN structure had any influence on alkaline stability, we subjected both QA PSf-226 and QA sIPN [72/20/8] to accelerated

116 Chapter 5 degradation testing, monitoring changes in conductivity and mass after soaking in a highly alkaline (6 M KOH) solution at 40°C (Figure 5.11). The QA PSf-226 membrane showed rapid loss in both conductivity and mass, becoming extremely fragile and exhibiting catastrophic mechanical failure within 12 hours of exposure, indicating significant degradation of the polymer backbone. On the other hand, the QA sIPN [72/20/8] sample showed enhanced mechanical integrity and was able to withstand exposure to the 6 M KOH solution for 30 hours prior to brittle failure. A

PS-co-DVB sample subjected to the same conditions showed no significant change in mass, suggesting suitable alkaline resistance. The ultimate stability enhancement is therefore ascribed to the presence of the PS-co-DVB matrix.

Despite the improvement, the overall system stability is fundamentally limited by the inherent issues of benzyltrimethylammonium polysulfone described previously.

A potential solution is to form a full IPN by crosslinking the QA PSf chains, partially mitigating the stability issues brought about by cleavage of the polysulfone. Moreover, while we used QA PSf as a benchmark to test our design, the synthesis is adaptable to other aromatic backbones (specifically those with similar solubility parameters) and/or other cation groups (provided the reagents are able to diffuse into and react within the semi-IPN film).

5.2.8 Membrane Electrode Assembly Performance

QA sIPN [72/20/8] was incorporated into a membrane electrode assembly (MEA) to assess device viability. Figure 5.12 shows the polarization curve for both QA sIPN

[72/20/8] at 35°C and 80°C and QA PSf-226 at 35°C under H2/O2 flows. Note that the

~40 mV difference in OCV between 35°C and 80°C for the QA sIPN [72/20/8] MEA

117 Chapter 5 is larger than would be expected from increased reaction kinetics alone. As we used the MEAs as-fabricated and without any additional activation protocol, we suspect that the load-cycling from the 35°C measurement helped “break in” the MEA, resulting in better catalyst activity[167] and a concomitant increase in the OCV when conducting the 80°C measurement.

Figure 5.12 Polarization (left axis, open markers) and power density (right axis, filled markers) for QA PSf-226 and QA sIPN [72/20/8] based MEAs at 35°C and 80°C at 100% RH.

Back pressure set to 200 kPa absolute.

Qualitatively, we found that the high water uptake of the unreinforced QA PSf-

226 baseline material rendered it mechanically delicate and prone to tearing during the fabrication of the MEA. This is reflected in the marked performance contrast between the QA sIPN [72/20/8] and the QA PSf-226 MEAs at 35°C, despite the two membranes exhibiting similar in-plane conductivities. Most notably, the open circuit voltage (OCV) for the QA PSf-226 membrane measured only ~600 mV compared to

~950 mV for the QA sIPN [72/20/8] membrane despite similar electrode materials, indicating significant fuel crossover effects that likely result from poor mechanical

118 Chapter 5 stability of the membrane (e.g., cracks and/or pinhole artifacts). This ultimately resulted in the baseline QA PSf-226 MEA exhibiting performance metrics (peak power density, maximum current density, etc.) that are roughly 50% of those of the more mechanically robust QA sIPN [72/20/8] MEA.

The poor mechanical stability of the QA PSf-226 membrane also led to an inability to test the MEA at higher temperatures; elevating the temperature past 40°C resulted in a sharp drop in the OCV and failure of the device. In contrast, the QA sIPN [72/20/8] MEA was stable up to 80°C, yielding a maximum current density around 670 mA/cm2 and a peak power density (PPD) of 236 mW/cm2. These results confirm that our mechanical reinforcement of a high IEC, linear alkaline polyelectrolyte directly translates to better overall device stability and performance.

5.3 Conclusions

We have demonstrated the ability to enhance the chemical and mechanical stability of a highly charged (IEC=2.99 mEq g-1) benzyltrimethylammonium polysulfone (QA PSf-299) alkaline exchange membrane material by reinforcing the linear polyelectrolyte chains with a poly(styrene-co-divinylbenzene) matrix, producing a semi-interpenetrating network architecture. Unlike the base QA PSf-299 material, which ruptured in water even at room temperature due to excessive water uptake, the semi-IPN membranes exhibited mechanical stability up to 80°C even with a high IEC

-1 in the 2.20 to 2.30 mEq g range.

The enhanced stability is attributed to a dramatically lower gravimetric water uptake and better mechanical properties. The higher dimensional stability of the semi-

IPN membranes translated to better conductivity stability at higher temperatures.

119 Chapter 5

Moreover, the room-temperature conductivities for semi-IPN samples did not vary drastically with the composition of the poly(styrene-co-divinylbenzene) network, suggesting that this secondary network did not interfere with the ion transport mechanics of QA PSf. This conclusion is underscored by the fact that all the semi-IPN membranes had similar Arrhenius activation energies. Finally, the highly charged semi-IPN membranes were stable up to 80°C while operating in a membrane electrode assembly, with the QA sIPN [72/20/8] MEA exhibiting a peak power density (PPD) of

236 mW/cm2 and a maximum current density around 670 mA/cm2.

While we chose benzyltrimethylammonium polysulfone as a model material due to its popularity in the literature and our previous experience[140] with it, its poor alkaline stability limits the long-term stability of the semi-IPNs based on it. Although we found that the extra support provided by the poly(styrene-co-divinylbenzene) matrix is able to nearly triple the lifespan of the membrane under accelerated degradation conditions (6 M KOH, 40°C), commercial viability would nevertheless demand a more robust polyelectrolyte. However, the semi-IPN approach can easily be adapted to more stable AEM backbones such as poly(phenylene oxide) and/or more stable cations such as quaternary sulfonium and phosphonium. Ultimately, we offer our results as a framework that can be applied to more chemically robust polyelectrolytes.

120 Chapter 6

Chapter 6: Facilitating Hydroxide Transport in Alkaline

Exchange Membranes via Hydrophilic Poly(ethylene glycol)

Grafts

Note: This chapter has been adapted from Steve S. He and Curtis W. Frank, J. Mater.

Chem. A, 2014 (2), pp. 16489 – 16497.

6.1 Introduction

Given the high performance of acidic proton exchange membranes (PEM), it is no surprise that their structure and chemistry have heavily influenced AEM design. As in PEMs, the general motif for synthesizing AEMs has been to attach pendant ionic salts along a robust hydrophobic polymer backbone. This approach typically manifests as aryl- or benzyl- substituted cations along an aromatic polymer chain. Whereas sulfonate is the pendant counter-anion of choice for PEMs, the pendant counter-cation in AEMs has been more varied, with the aim of improving hydroxide conductivity and alkaline stability. Recent approaches have involved membranes based on quaternary ammonium[31,38,70,123,156,168,169], imidazolium[87] , guanidinium[99], triazole,[55] phosphonium[71] and sulfonium cations[103], amongst others. Quaternary ammonium based on trimethylamine is by far the most widely studied of these pendant cation groups and has been introduced on an assortment of different polyaromatic backbones, including polysulfone[31,38,70,156], poly(phenylene oxide)[123,168] and polyetheretherketone.[169]

121 Chapter 6

However, in these systems, the close proximity of the pendant cation to the rigid polymer backbone inhibits the formation of strongly segregated hydrophilic- hydrophobic domains.[130,133] Consequently, these membranes are often characterized by poorly defined water-rich phases, leading to ion transport occurring in highly constricted and tortuous pathways. A natural approach for improving AEM performance, then, is to design ionomers with better-defined ion transport domains[170,171] by promoting strong microphase separation between the pendant counterion and the polymer backbone. Indeed, several groups have recently implemented this concept by introducing linear spacers either between the backbone and the counterion (a la Nafion)[23] or as a side-chain of the counterion.[123,124]

Another avenue for modification is to graft side-chains along the same backbone as the pendant counterion. A recent study by Pan et al[127] showed that grafting alkyl chains along a model AEM polymer material facilitated clustering of the hydrophobic species, resulting in a interstitial water-rich ion transport “highway”. In this work, we show that structure formation in these random graft copolymer systems can be achieved through a different mechanism. Specifically, we show that introducing flexible, hydrophilic poly(ethylene glycol) (PEG) grafts along a typical pendant counterion-hydrophobic polyaromatic backbone ionomer can induce local phase separation and leads to enhanced hydroxide conductivity.

The synthetic approach we adopt is applicable to several AEMs previously reported in the literature. As a model system, we grafted PEG moieties along the highly-studied benzyltrimethylammonium-functionalized polysulfone (QA PSf) AEM to yield quaternary ammonium polysulfone-graft-poly(ethylene glycol) (QA PSf-g-

122 Chapter 6

PEGx) (Scheme 6.1). The design rationale for promoting phase separation is two-fold.

First, the hydrophilic PEG graft has a repulsive χ interaction parameter with the hydrophobic polysulfone backbone and favorable interaction with water (highlighted by its reported use as a PEM humidifying agent[172]). Moreover, PEG’s electron-rich ether groups have been shown to complex with cationic quaternary ammonium salts[173,174] and are expected to interact favorably with the pendant benzyltrimethylammonium (BTMA) species along the polysulfone backbone. The culmination of these effects leads to co-localization and concentration of the quaternary ammonium groups into efficient, water-rich hydroxide transport domains, resulting in increased ionic conductivity.

Scheme 6.1 Synthesis of QA PSf-g-PEGx where ‘x’ corresponds to the molecular weight of the PEG graft. Red and blue denote hydrophobic and hydrophilic portions, respectively, of QA

PSf-g-PEGx.

123 Chapter 6

6.2 Methods

6.2.1 PEGylation of Chloromethylated Polysulfone

Chloromethylated polysulfone was prepared in the protocol detailed in Chapter 4. The

PEGylation reaction follows the modified Williamson-Ether Synthesis described by

Park et al.[175] A typical PEGylation procedure is described as follows. In a nitrogen glove bag, 1.6 mmol of sodium hydride (60% w/w suspension in mineral oil) was first dissolved in THF at a concentration of 0.01 g/mL, and then added to a 25 wt% poly(ethylene glycol) monomethyl ether (1.2 mmol) solution in THF. The reaction was allowed to proceed for 2 hours with stirring, after which it was added dropwise to a solution of chloromethylated polysulfone (1.0 g of DS 1.05 in 20 mL THF). The solution was reacted for 20 hours at room temperature and then precipitated into a 3:1 mixture of petroleum ether to ethanol. The precipitate was collected by vacuum filtration to yield polysulfone-graft-poly(ethylene glycol) monomethyl ether (PSf-g-

PEG). The degree of PEGylation was determined by 1H NMR.

6.2.2 Quaternization of PSf-g-PEG

QA PSf-g-PEGx was synthesized by in situ quaternization of trimethylamine via the

Menshutkin reaction. Specifically, trimethylamine (4.2M in ethanol) was added dropwise to a solution of either chloromethylated polysulfone or polysulfone-graft- poly(ethylene glycol) (1 g in 10 mL DMF) at 3x molar excess. The reaction mixture was stirred under ambient conditions for 48 hours.

124 Chapter 6

6.3 Results and Discussion

6.3.1 Synthesis

The 1H NMR spectrum of CMPSf in shown in Figure 6.1a. Successful chloromethylation of polysulfone was verified by the emergence of a singlet at 4.52 ppm associated with the Ar-CH2-Cl hydrogens. The integration of this peak estimates an average of 1.14 chloromethyl groups per polysulfone monomer.

A representative NMR spectrum of PSf-g-PEG350 is provided in Figure 6.1b. The broad feature between 3.5 ppm and 3.8 ppm corresponds to the hydrogens in the -

CH2CH2O- repeat unit. Integration of this peak gives 7.9 ethylene glycol repeat units per PEG350 side-chain, corresponding to a molecular weight ~350 Da, as expected.

The emergence of an additional peak at 4.46 ppm is a result of the partial conversion of Ar-CH2-Cl to Ar-CH2-PEG; integration of this peak gives an average degree of

PEGylation of 0.22 PEG350 side-chains per polysulfone repeat unit. The singlet at

3.36 ppm represents the methoxy hydrogens terminating the end of each PEG monomethyl ether chain; integration of this peak quantifies the number of PEG chains in the membrane and was found to be 0.21. That the total number of PEG chains per repeat unit is equivalent (within error) to the PEG graft density verified that there were no excess, unreacted PEG reagents in the purified product. The graft density of ~0.22

PEG350 chains per repeat unit translates to a mass composition of 12% w/w PEG350.

A similar 1H NMR analysis performed on PSf-g-PEG750 reveals a graft density of

~0.10 PEG750 chains per repeat unit, or a mass composition of 12% w/w PEG750.

125 Chapter 6

Figure 6.1 1H NMR spectra of (a) CMPSf and (b) PSf-g-PEG350.

6.3.2 Morphology

We performed small angle x-ray scattering (SAXS) to characterize the morphology of the PEGylated membranes. The SAXS profiles of QA PSf, QA PSf-g-PEG350 and

QA PSf-g-PEG750 are shown in Figure 6.2a. The absence of notable scattering features for QA PSf suggests a homogeneous morphology. This result is consistent with the expectation that the short methyl linkage between the polysulfone backbone and the quaternary ammonium species inhibits strongly separated hydrophobic- hydrophilic domains, and is corroborated by morphological studies of pendant BTMA in similar systems.[44,123] On the other hand, the introduction of PEG grafts gives rise to a broad scattering peak in the mid-q region (0.1 nm-1 to 1 nm-1), indicating the formation of distinct microphase separated regions. This PEGylation-induced phase-

126 Chapter 6 separation is attributed to the strong χ repulsion between the flexible PEG graft and the PSf backbone.

The Teubner-Strey (TS) bi-continuous model was used to help elucidate the structure associated with the SAXS data. Originally developed to describe oil-water- surfactant microemulsions, this model has since been employed to study the structure of the ionomer phase in sulfonated PEMs.[176–178] The microemulsion analogy is rationalized by the surfactant-like behavior of covalently bound hydrophobic- hydrophilic groups within the polymer and is supported by recent theoretical work on phase-separated random copolymer morphologies.[179]

Figure 6.2 (a) Radially averaged SAXS data for QA PSf and QA PSf-g-PEGx. (b) Illustration of proposed morphological influence of the PEG grafts. Red and blue correspond to hydrophobic and hydrophilic regions, respectively. Restricted phase-separation in QA PSf

(left) leads to poorly defined mesoscale morphology, resulting in tortuous ion transport. The inclusion of PEG grafts (right) leads to more well-defined phase-separated domains on the order of 5 to 10 nm in diameter, resulting in more efficient anion transport.

127 Chapter 6

Table 6.1 Structural and Performance Data. (a) IEC [mmol OH- g-1] determined by NMR and back-titration (in parentheses). (b) OH- Conductivity [mS cm-1] at 60°C. (c) OH-

Conductivity at 60°C normalized against titrated IEC values [mS g cm-1 mmol-1]. (d,e) Domain spacing d [nm] and correlation length ξ [nm] from Teubner-Strey fitting of the SAXS scattering profiles.

(a) (b) (c) (d) (e) IEC σ σnorm d ξ QA PSf 2.02 (1.98) 40.0 20.2 - - QA PSf-g-PEG350 1.50 (1.36) 48.2 35.4 6.20 1.93 QA PSf-g-PEG750 1.64 (1.57) 63.2 40.3 7.86 1.63

The TS model proposes a structure factor of following form:

1 � � = ! ! + �� ! + �!� + �!�

A regression fit of the structure factor proposed by this model to the QA PSf-g-PEGx scattering curves yielded excellent agreement, suggesting that the QA PSf-g-PEGx membranes adopt a mesostructure characterized by a percolating ionomer network co- continuous with a hydrophobic matrix.

The two distinct length scales associated with this fit are given by:

!! ! ! 1 � ! � � = 2� ! − ! 2 �! 4�! and

!! ! ! 1 � ! � � = ! + ! 2 �! 4�! where d is a length scale representing a quasiperiodic spacing in the pair correlation function and can be physically interpreted as the average size of an ionomer

128 Chapter 6 domain[180]; ξ is a characteristic length for correlation falloff. The values of these length scales derived from the TS fit are listed in Table 6.1. The scattering data can be rationalized by the chain architecture for both QA PSf-g-PEG350 and QA PSf-g-

PEG750. The larger domain size for QA PSf-g-PEG750 is ascribed to the longer contour length of the higher molecular weight PEG750 (5.8 nm) compared to PEG350

(2.8 nm). Water uptake in these hydrophilic channels results in swelling and domain sizes exceeding the contour length of the PEG grafts. An illustration of our morphological interpretation of the SAXS data is provided in Figure 6.2b.

The weaker domain correlation (as manifest in the lower ξ) of QA PSf-g-

PEG750 can be explained by its lower PEG grafting density, which is roughly half that of QA PSf-g-PEG350 given the same PEG weight composition in the two systems.

The larger average spacing between the PEG chains coupled with their random placement leads to decreased grafting regularity and diminished long-range interactions. This effect is again evidenced in the polydispersity of the domain sizes, as reflected by the ξ/d ratio, where a lower value corresponds to higher polydispersity.

The ratio for QA PSf-g-PEG750 (0.20) is two-thirds of that for QA PSf-g-PEG350

(0.31), suggesting a broader distribution of hydrophilic domain sizes. In summary,

PEGylation of QA PSf gives rise to a broad scattering feature analogous to that of a bi-continuous microemulsion; this scattering is attributed to microphase separation of hydrophilic, PEG-rich ionomer channels from the hydrophobic polysulfone matrix.

Given the same weight composition, QA PSf-g-PEG350 has narrower, but less disperse domain sizes compared to QA PSf-g-PEG750.

129 Chapter 6

Figure 6.3 Photographs of PSf-g-PEG350 (top) and QA PSf-g-PEG350 (bottom) films.

Visual comparison between polysulfone-graft-poly(ethylene glycol) films with and without quaternary ammonium functionalization showed striking macroscopic differences (Figure 6.3). The highly turbid PSf-g-PEG350 film suggests micron-scale phase-separation arising from the incompatibility between the PEG and PSf. In contrast, the introduction of benzyltrimethylammonium groups in QA PSf-g-PEG350 results in an optically clear film, indicating that the presence of the charged quaternary ammonium species inhibits macrophase-separation and underscores the interaction between the PEG and the pendant quaternary ammonium ions.

6.3.3 Scaling Analysis

The Alexander-de Gennes scaling theory describes the physical conformations of surface-tethered polymer chains. This theory contends that the thickness (or extensional length) H of the polymer brush scales as

� �~� � where N is the single-chain molecular weight, g is the size of a de Gennes blob and ξ is the correlation length defined by the average distance between grafting points.

130 Chapter 6

We adopted the principles of this scaling analysis to elucidate the local structure of the anion transport domains by modelling the graft-copolymer architecture of QA PSf-g-

PEGx as a simple one-dimensional polysulfone main-chain with tethered PEG sidechains. In this case, the correlation length simply scales inversely with the linear graft density Λ:

1 �~ Λ and the blob size scales as

� !/! �~ � where b is the Kuhn length of the graft and v is the Flory scaling exponent. Hence, we have the following scaling relationship between the brush thickness, H, and the linear graft density, Λ :

! ! ! !! �~ Λ! ��!~Λ!��!

Here, α is the scaling exponent relating brush thickness and graft density and is equal to 0.7 assuming real chains in a good solvent (i.e., v=0.588).

From our SAXS data, we posit that the domain size d ~ H and assume that both

PEG350 and PEG750 have the same Kuhn length b. Moreover, we note that at the same weight composition, the linear graft density ratios should simply be the inverse of the molecular weights. With these assumptions, we arrive at the following relationship:

� Λ ! � � !!! !"#$%& ~ !"#$%& !"#$%& ~ !"#$%& �!"#$%& Λ!"#$%& �!"#$%& �!"#$%&

131 Chapter 6

Inputting the empirical domain sizes and molecular weights of the PEG grafts, we find that α=0.689, which is very close to the α=0.7 value assuming real chains in a good solvent.

The close conformity between the empirical scattering data and the Alexander- de Gennes scaling theory suggests that the PEG side-chains adopt a brush-like conformation in the ionomer channels. This local structure can be rationalized by the thermodynamic propensity of the hydrophilic PEG side-chains to extend away from the hydrophobic PSf backbone and into a water-rich region upon hydration.

6.3.4 Hydroxide Conductivity

Hydroxide conductivity is a critical performance metric for AEMs. The hydroxide conductivity of un-PEGylated QA PSf was 40.0 mS cm-1 at 60°C, consistent with previously reported data in the literature.[32] The introduction of 12% w/w PEG350 and PEG750 grafts increased the conductivity to 48.2 mS cm-1 and 63.2 mS cm-1, respectively. The temperature dependence of the hydroxide conductivities followed an

Arrhenius trend (Figure 6.4). Concomitant with the higher hydroxide conductivity was a decrease in the apparent Arrhenius activation energy, from 17.5 kJ/mol (QA PSf) to

15 kJ/mol (QA PSf-g-PEG350) and 11.2 kJ/mol (QA PSf-g-PEG750). The enhanced performance with PEGylation supports the notion that a microphase-separated structure is beneficial to ion transport.

The structural influence on the enhanced performance is underscored by comparing the ion exchange capacities (IEC) of the materials (Table 6.1). Of particular interest is that the PEGylated membranes have lower IECs than QA PSf.

While a decrease in IEC typically manifests in decreased ionic conductivity as a result

132 Chapter 6 of lower charge content, both QA PSf-g-PEG350 and QA PSf-g-PEG750 show increased performance. Hydroxide conductivity normalized by the IEC has been used as a qualitative metric for assessing ion transport efficacy[72,75,103] and is presented in

Table 6.1. PEGylation increases the IEC normalized conductivity by 81% and 100% for QA PSf-g-PEG350 and QA PSf-g-PEG750, respectively, highlighting the influence of structure formation on ion transport properties.

Figure 6.4 Temperature dependence of hydroxide conductivity for QA PSf and QA PSf-g-

PEGx at 100% RH. The dashed lines represent Arrhenius fits.

6.3.5 Water Uptake

Water content strongly influences conductivity in ion exchange membranes.

Generally, under moderate water uptake conditions (e.g., barring notable dilution of ion concentration, mechanical degradation of the membrane, etc.), ionic conductivity scales directly with water content.[144] Table 2 presents the water uptake data for the prepared membranes.

133 Chapter 6

-1 Table 6.2 Water Uptake Data at 22°C. (a) Gravimetric water uptake. (b) λ [mol H2O mol

BTMA] as calculated from water uptake and titrated IEC. (c) λ-normalized OH- conductivity

-1 -1 [mS cm mol H2O mol BTMA].

(a) (b) (c) WU% λ σλ

QA PSf 146% 41 0.48

QA PSf-g-PEG350 81% 33 0.85

QA PSf-g-PEG750 96% 34 1.12

Despite the presence of hydrophilic PEG grafts, the PEGylated membranes nonetheless exhibit lower water uptake compared to bare QA PSf due to their significantly lower IECs. That the influence of ion concentration on water uptake outweighs that of PEG incorporation is unsurprising given that the ion-dipole interaction associated with quaternary ammonium cation solvation is stronger than the hydrogen-bonding interaction involved in PEG hydration. This effect is evident in the hydration numbers of their small molecule analogs. The tetramethylammonium cation has been calculated to have a hydration number of approximately 23;[181] in comparison, PEG is estimated to contain a maximum of one bound water molecule per ethylene oxide repeat unit via hydrogen bonding at the ether position.[182]

The λ parameter denotes the average number of water molecules per pendant BTMA group. Whereas bare QA PSf has λ=41, QA PSf-g-PEG350 and QA PSf-g-PEG750 have λ=33 and 34, respectively. Juxtaposing the λ values for the different samples against their conductivities, it becomes clear that PEGylation yields higher conductivities at lower water content. This effect is highlighted by the λ-normalized

134 Chapter 6 conductivities, where QA PSf-g-PEG350 and QA PSf-g-PEG750 show 77% and

133% improvements over QA PSf. This increase in efficiency again corroborates the idea that PEGylation-induced structure formation is facilitating hydroxide transport.

6.3.6 Alkaline Stability

Quaternary ammonium polysulfone AEMs are known to have questionable alkaline stability. It is well established that the benzyltrimethyammonium cations in these

AEMs are subject to nucleophilic attack by hydroxide anions, resulting in decreases in ion concentration and conductivity.[165] Moreover, Arges et al have shown that the polysulfone backbone, otherwise stable at high pH, becomes vulnerable to hydrolytic cleavage upon quaternary ammonium functionalization, leading to embrittlement and mechanical degradation.[25] In practice, the extent and rate of degradation is subject to environmental conditions (e.g., pH, temperature, etc.) and chemical makeup (e.g., grafting degree, IEC, etc.). Here we aim to assess the effects, if any, that PEGylation may have on alkaline stability, again using QA PSf as a baseline for comparison.

Specifically, we exposed all three membranes to accelerated degradation conditions (1

M and 6 M KOH solutions at 60°C) and monitored changes in both the mass (Figure

6.5a) and ionic conductivity (Figure 6.5b) over time.

Exposure to extremely caustic conditions (6 M KOH) resulted in the rapid degradation of all membranes. The significant mass loss (>20%) for all samples after

24 hours of exposure is attributed primarily to the degradation of the quaternary ammonium-functionalized polysulfone backbone. This effect was manifest physically in the embrittlement and discoloration (yellowing) of the samples. Ultimately, all samples shattered into several pieces after removal from the conductivity cell at the

135 Chapter 6

24-hour mark. It has been suggested that the primary mechanism for quaternary ammonium polysulfone backbone degradation under these conditions is hydrolytic cleavage of the ether bond weakened by the electron withdrawing effects of the quaternary ammonium and sulfone functionalities.[25] Hence, a priori we had expected that the PEG graft would stabilize this ether bond via the electron donating potential of the PEG ether oxygen. However, the rapid degradation for all samples indicates that

PEGylation had minimal effect on the mechanical robustness of the samples under these accelerated conditions. Similarly, the conductivity of all samples dropped precipitously (>40%) within 24 hours attributed to a combination of severe mechanical degradation (via attack on the polysulfone backbone) and decrease in ion concentraton (via attack on the quaternary ammonium cation).

136 Chapter 6

Figure 6.5 Stability of samples exposed to 1 M and 6 M solutions of KOH at 60°C as characterized by changes in (a) gravimetric mass and (b) in-plane conductivity at 22°C.

Testing under mild alkaline conditions (1 M KOH) revealed slight stability differences between QA PSf and QA PSf-g-PEGx. Whereas QA PSf lost nearly a quarter of its mass and became brittle over a 5-day period, the PEGylated membranes retained 85-90% of their initial mass and remained flexible, suggesting improved backbone stability. Although differences in cation stability, as inferred by the conductivity drop, were notable after a 5-day exposure period (e.g., 12% retained conductivity for QA PSf vs. 29% for QA PSf-g-PEG350), they were nonetheless

137 Chapter 6 pragmatically insufficient given that all membranes lost more than 70% of their initial conductivity. We note that the degradation rate of the membranes were directly related to their IECs, with QA PSf-g-PEG350 showing the highest stability with an initial IEC of 1.36 mmol OH- g-1 and QA PSf showing the lowest stability with an initial IEC of

1.98 mmol OH- g-1. As such, the moderate stability improvement of the PEGylated membranes under these conditions is likely due to their lower charge concentrations

(i.e., IEC).

Ultimately, our data suggests that the PEG grafts had marginal influence on overall alkaline stability. While the the PEG side-chains do not accelerate the degradation of the membranes, they nevertheless do not resolve the inherent robustness issues associated with both the polysulfone backbone and the benzyltrimethylammonium cation. As noted previously, while we used QA PSf as a platform for testing our design strategy, the synthesis is inherently adaptable to other aromatic backbones and cationic functional groups. Thus, using a more stable backbone (e.g., poly(phenylene oxide)[48]) and/or cation (e.g., phosphonium[72] or sulfonium[103]) may yield more practical materials.

6.3.7 Fuel Cell Performance

Membrane electrode assemblies (MEAs) using the PEGylated membranes were fabricated to assess prototypical H2/O2 fuel cell performance. MEA performance of the PEGylated membranes, as characterized by the polarization and power density curves (Figure 6.6), mirrors their enhanced hydroxide conductivity. The introduction of PEG grafts yielded up to a 50% increase in the peak power density from 120 mW

138 Chapter 6 cm-2 for QA PSf to 183 mW cm-2 for QA PSf-g-PEG750 as a result of a more efficient ion transport architecture.

Figure 6.6 Polarization and power density curves of MEAs fabricated using QA PSf and QA

PSf-g-PEGx membranes (60°C, 100% RH).

The theoretical open circuit voltage (OCV) is the water-splitting potential

(~1.2 V). Deviations from the ideal OCV are commonly attributed to catalytic overpotentials and fuel crossover. The OCV for all the MEAs were approximately 1.0

V and did not change noticeably with the introduction of the PEG side-chains, indicating that any effect on H2/O2 permeability resulting from the morphological and chemical changes induced by the PEG grafts were negligible.

6.4 Conclusions

We have demonstrated that grafting hydrophilic poly(ethylene glycol) chains along a benzyltrimethylammonium polysulfone AEM can promote nanoscale hydrophobic- hydrophillic domain formation. Co-localization and concentration of the ion conducting quaternary ammonium moieties as a result of this phase separation results in the formation of a percolating, water-rich anion transport phase, leading to more

139 Chapter 6 efficient hydroxide transport as demonstrated by increased in-plane conductivity and device performance.

A scaling analysis suggests that the local, nanoscale structure is characterized by a brush-like conformation of the poly(ethylene glycol) grafts; this architecture is rationalized by the thermodynamic propensity of the hydrophilic sidechains to repel themselves from the hydrophobic backbone and extend into water-rich regions upon hydration. Energetic frustrations arising from the random grafting of the sidechains result in a lack of structures at length-scales greater than that of the nanometer-sized domains. Consequently, as demonstrated by a close agreement of the SAXS data with the Teubner-Strey model, the mesoscale morphology is characterized by a percolating network of these nanoscale anion transport domains within a continuous hydrophobic matrix.

We chose benzyltrimethylammonium polysulfone as a model system for our study due to the availability of several literature sources against which we could benchmark and verify our data. However, we believe the same thermodynamic considerations and design rationale can be applied to similar polyaromatic backbones

(e.g., polyphenylsulfone) as well as to other ion exchange groups (e.g., imidazolium).

Moreover, because the PEGylation chemistry shares the same alkyl halide reactive site that is commonly used to quaternize tertiary amines, the synthesis is inherently adaptable to several polyaromatic systems reported in the literature. Ultimately, we hope that this grafting approach offers new design considerations for future polyelectrolyte membranes.

140 Chapter 7

Chapter 7: Simulation and Theory of Random Copolymers

Towards Understanding the Morphologies of Anion Exchange

Membrane Materials

Note: The work in this section is a result of a collaborative effort with Shifan Mao,

Michael Eissien, Elyse Coletta, and Andrew Spakowitz. Parts of this chapter have been adapted from Mao et al. (in preparation) and figures, unless otherwise stated, were adapted from those provided by Shifan Mao. Shifan Mao developed the underlying mathematical theory as well as the code for both Monte Carlo simulations and numerical solutions to the theory. I, Steve He, aided with code development,

Monte Carlo simulations, data processing and interpretation, and offered advice on experimental implications.

7.1 Introduction

Transport in polymer electrolyte membranes is highly dependent on the material’s morphology and microstructure. As discussed in Chapter 2, ion transport in proton exchange membranes at high water content is analogous to that of rapid proton transport in dilute solution, where proton migration occurs through the Grotthus mechanism. Anomalous hydroxide transport in alkaline anion exchange membranes has been proposed to share a similar mechanism.[7,145] However, the structural diffusion process is highly influenced by the microstructure of the polymer electrolyte

141 Chapter 7 materials, as the effective diffusivity is contingent on morphological considerations such as percolation and tortuosity of the ion transport pathway.[15]

To date, the majority of alkaline exchange membrane materials are based on a random copolymer design. Typical AEMs are synthesized by decorating a polymer backbone with cationic head groups, or by polymerizing a mixture of ionic, polar monomers and non-polar monomers. In either approach, the polymer chains are characterized by a stochastic distribution of ionic and non-ionic groups. The same concept underscores the composition of block copolymer AEMs. As evident in

Chapter 3, the majority of reported block copolymer AEMs are multiblock copolymers, where the oligomeric components are either stochastically distributed or are perfectly alternating with each other along the main chain. These multiblock materials synthesized from oligomeric blocks share the same underyling thermodynamic behavior with random copolymers synthesized from individual monomers, but evaluated at a larger length scale.

In contrast, much of the theory and simulation development on copolymer phase-separation have focused on diblock copolymers. These studies are traditionally based on a mean-field treatment of Gaussian chains in a polymer melt. Similar efforts in elucidating the phase-behavior of random copolymers have followed an analogous treatment. Frederickson et al. reported that the stochastic chemical composition of random copolymers gives rise to quenched disorder and a lack of the large-scale ordering associated with diblock copolymers, resulting in a “disordered microphase” dominated by local interactions upon demixing.[183,184] Other studies and simulations have supported this idea, reporting that the characteristic length scale of phase-

142 Chapter 7 separation in random copolymers is dictated by the statistical size of a monomer block[179,185,186]; this is in contrast to diblock copolymers, where the relevant length scale is the size of the polymer chain.

However, the compositional differences between diblock and random copolymers bring some of the simplifying assumptions in these theoretical treatments into question. For example, the mean-field approach does not adequately capture the divergent density fluctuations near the order-disorder transition. While this may be acceptable for diblock copolymers, where the length scale of phase-separation is often much larger than that of local density fluctuations, these density fluctuations can have profound impact on the predicted morphology of random copolymers, where the relevant length scales are on the order of monomers. Furthermore, the Gaussian chain approximation becomes invalid as the characteristic length scale of interest approaches that of the Kuhn length, and a semi-flexible model needs to be employed to truly elucidate chain statistics.

Here I aim to present an introduction to collaborative efforts towards addressing some of these issues via polymer theory and Monte Carlo simulations, with the ultimate goal of elucidating the morphology of random copolymers as it relates to the development, design and understanding of alkaline anion exchange membranes.

From the theory, we find that statistical distribution of monomers and the overall stiffness of the polymer chain significantly influence the order-disorder transition of the system. In addition, our simulation results show that these two parameters profoundly affect the phase-separated morphology of the polymer. The primary focus

143 Chapter 7 is on discussing the qualitative aspects of these approaches and assessing their results in the context of AEM materials design.

7.2 Method

For simplicity and illustrative purposes, we focus specifically on a two-component multiblock system composed of A and B-type blocks, each with G monomers (i.e., G is the single block size). Unless otherwise stated, all data and analysis were prepared assuming equal number fractions of A and B (fA=fB=0.5). The two primary parameters of interest are the chemical correlation factor, λ, and the chain rigidity, Nb. The correlation factor, adapted from Frederickson et al.,[183] is related to the reactivity ratio between A and B and reflects the statistical preference for A or B to be covalently bound to similar (e.g., A with A) or dissimilar (e.g., A with B) blocks. A λ > 0 represents positive correlation, where an A block is more likely to be bound with another A block (and similarly for B); λ = 1 represents the upper limit of a homopolymer. A λ < 0 represents negative correlation, where an A block is more likely to be tethered to a B block, and vice-versa; the λ = -1 limit defines a perfectly alternating copolymer (i.e., A-B-A-B-…, etc.). The λ = 0 value reflects an ideal random copolymer, where the statistical distribution of A and B depends only on their respective number fractions fA and fb. An illustration of the effect of λ on the composition of a polymer chain is provided in Figure 7.1.

The chain rigidity is parameterized by Nb, which is defined as the ratio of the end-to-end length L of a single block to the Kuhn length (i.e., twice the persistence length lp):

144 Chapter 7

� �! ≡ 2�!

. Hence, Nb >> 1 reflects flexible Gaussian chain statistics, while Nb << 1 represents the rigid rod limit; interim values characterize a semi-flexible polymer that can be described by worm-like chain statistics.

Figure 7.1 Effects of chemical correlation λ on the composition of polymer chains. Red and blue represent A and B blocks.

Both theory and simulation efforts were employed to evaluate the effects of λ and Nb on the thermodynamic behavior of a polymer melt. The free energy of the system is at the core of both efforts and consists of: (1) the conformational free energy

145 Chapter 7 of the polymer chains and (2) the interaction free energy between A and B-type blocks determined by the χ interaction parameter. The chain rigidity is accounted for in a

WLC model of the conformational free energy adapted from Spakowitz et al.,[187–191] while the stochastic distribution of chemical identities along the chain is inherently manifested in the interaction free energy.

The complexity of the system renders a direct extraction of the spinodal from the free energy expression analytically intractable. In light of this, the theory developed by Mao et al. adopts the random phase approximation, which condenses individual particle-particle interactions into perturbations by a fluctuating mean-field; details of the mathematical development of this theory can be found in Mao et al.[192]

Moreover, the stochastically determined composition of the polymer chains inherently introduces quenched disorder in the system; this is accounted for by averaging over all possible chemical identities with the aid of the “replica trick”. The theory ultimately allows prediction of the spinodal and order-disorder transition of random copolymers subject to their rigidity and composition. To explore the morphological implications of these parameters, we performed a series of simulations based on a theoretically- informed coarse-grained formalism introduced by De Pablo et al.[193] The system is defined as a box filled with polymer chains, the compositional distributions of which are statistically determined by λ. The chains are then subject to Monte Carlo moves towards minimizing the overall free energy of the system as a function of the interaction parameter. Our initial assessment of the conformational free energy was evaluated using a simple freely-jointed chain model; however, we have since adapted a coarse-grained WLC model to capture the effects of semi-flexibility. This particle-

146 Chapter 7 based simulation is particularly advantageous over traditional field-theoretical simulations in that it inherently accounts for local density fluctuations and consequently offers a more accurate account of polymer behavior near and above the order-disorder transition.

7.3 Results and Discussion

Instability and phase-transition are dictated by a divergence in the density correlation at a wavevector k* – this wavevector is inversely related to a characteristic length scale of instability. For example, a large k* indicates local instability and the propensity to form heterogeneous microphases, while k*=0 indicates an energetic preference for macroscopic demixing (e.g., a liquid-liquid phase). This latter concept is easily illustrated in the limit of λ=1 homopolymers, where the phase-separated polymer blend exhibits a bimodal distribution of A and B chains. The same partitioning phenomenon is envisioned with other λ values at the k*=0 instability, where the fractionation is no longer bimodal, but becomes weighted by the statistical composition of each chain.[183]

147 Chapter 7

Figure 7.2 Effect of chemical correlation λ and chain rigidity Nb on the non-dimensional wavelength of instability. Nb ranges from 1000 (red) to 0.01 (blue). The solid black line represents the flexible Gaussian chain limit, while the dashed black line represents the rigid rod limit.

The critical wavector k* was non-dimensionalized with Rb, the average end-to- end distance of a block, to facilitate comparison. Plotting k*Rb against the chemical correlation λ reveals interesting differences between the rigid and flexible random copolymers; namely, the characteristic domain size for a flexible polymer is much more responsive to λ. In contrast, increasing chain rigidity results in a more muted response to chemical correlation, with the more rigid chains exhibiting a plateau behavior in k*Rb at low λ. This can be physically rationalized by the idea that the conformational limitations of the more rigid chains inhibit local rearrangement, and the rigidity, rather than composition, defines the characteristic phase-separated length scale; this effect is especially pronounced at low λ, where the average block sizes are

148 Chapter 7 much lower than that of the Kuhn length. In contrast, the flexible polymers are easily able to maneuver to maximize local interactions and are much more sensitive to the distribution of blocks along the polymer chain.

Figure 7.3 Non-dimensional characteristic domain size as a function of chemical correlation and chain rigidity.

The rigidity of the polymer chain also affects the size of local domains. Figure

7.3 plots non-dimensional domain size versus rigidity, where D=2π/k* is a Bragg-like

1/2 domain spacing and 2lpNb is the average end-to-end distance of a flexible Gaussian chain. The conformational limitations resulting from increased rigidity “locks in” a characteristic domain size. As a result, chemical correlation as little effect on the characteristic domain sizes for rigid polymers. Conversely, increasing the flexibility of the chains facilitates clustering, resulting in a monotonic increase in the non- dimensionalized domain size towards the Gaussian chain limit. Accordingly, chemical correlation for more flexible polymers has a much significant influence on domain

149 Chapter 7 spacing, with more correlated (i.e., more positive λ) chains exhibiting larger domain sizes as a result of increased statistical block sizes.

Note that all chains, spanning the entire rigidity spectrum, feature a λ value above which is defined by the k*=0 instability. The λ at which this occurs captures both the k*=0 as well as the k*≠0 instabilities and, as such, defines a Liftshitz point λL featuring coexistence of the two phases. This Lifshitz point is a notable function of rigidity, ranging between λ=-0.26 for the Gaussian chain to λ=-0.10 for the rigid rod.

For λ< λL, the heterogeneous phase is defined by a locally heterogeneous microphase, while for λ> λL the theory predicts a spinodal for both a liquid-liquid phase and a disordered microphase. Again, the conformational limitations in rigid polymers inhibit large-scale partitioning, resulting in in a more positive λL.

Figure 7.4 Order-disorder transition for a random copolymer as a function of chemical correlation and chain rigidity. The rigidity Nb ranges between Nb=1000 (red) and Nb=0.01

(blue). The solid black line represents the flexible Gaussian chain limit, while the dashed black line represents the rigid rod limit.

150 Chapter 7

Figure 7.4 shows the effect of chemical correlation and rigidity on the order- disorder transition. We note at low levels of chemical correlation that a lower Nb, and therefore a more rigid polymer, results in a suppression of the ODT. This behavior is analogous to that found in diblock copolymers and can be rationalized by the fact a stiffer chain has a lower number of accessible conformations; this reduction in the entropy emphasizes the effect of the enthalpic interactions of the system, which drive de-mixing between the two components. However, this effect disappears at higher chemical correlations and the ODT falls onto a single universal curve for λ> λL. In fact, positive λ values show no dependence of the ODT on chain rigidity. One might intuitively suspect that chain rigidity plays a much larger role in determining the ODT at higher λ; this is especially true for larger Kuhn lengths that greatly exceed the average block sizes near λL. However, it is important to note that the metric of importance in this case is not the average size of the block, but rather the statistical distribution of block sizes.

151 Chapter 7

Figure 7.5 Simulation results of random copolymers as a function of chemical correlation and chain rigidity, comparing a semi-flexible chain with a rigid rod. Solid lines in the structure factor represent those predicted by the mean-field theory at χG=0 and χG=0.5 χsG, where χs is the critical interaction parameter at the spinodal. Points in the structure factor are those calculated from the simulated morphology and represent the structure factor in a single direction, and the red markers are those corresponding to structure in the highly-segregated regime(χG=25 χsG). Representative morphologies in the highly-segregated regime are shown in the inset simulation snapshots, where red and blue correspond to A and B species.

152 Chapter 7

It is important to stress that while the ODT may remain the same and independent of rigidity, the resulting morphologies can be completely different. Figure

7.5 highlights this effect via simulations of random copolymer melts in the heterogeneous phase and underscores the effects of both rigidity and chemical correlation on the phase-separated structure. While the mean-field theory described previously provides insight into the phase-transition behavior and characterstic interactions below the order-disorder transition, it fails to capture the proper physics near and above the order-disorder transition. The particle-based simulations inherently account for the instantaneous densities within the system and, as a result, offer a more adequate account of thermodynamic behavior past the phase-transition and are therefore critical to elucidating heterogeneous morphologies above the ODT. The inherent quenched disorder in these copolymers is accounted for by averaging the simulations over multiple realizations of the same holistic compositions, and the validity and proper sampling of these results are corroborated against the structure predicted by the analytical theory. We note that below the ODT, the simulation and theory share similar results for the structure factor. The simulation also corroborates the qualitative effects of chain rigidity and chemical correlation predicted by the theory, where the size of the phase-segregated domains is much more confined for the rigid rod compared to the semi-fleixble polymer. Furthermore, increasing the chemical correlation towards an ideal random copolymer results in larger and more disperse domain sizes.

153 Chapter 7

7.4 Experimental Implications

The pragmatic question that arises is the applicability of these results to not only interpreting and evaluating existing experimental data, but to establishing of materials design principles. The core parameters of chemical correlation/monomer reactivity ratio and chain stiffness in this study are the same parameters governing phase behavior in many of the random copolymer materials discussed in Chapter 3 and may provide additional insight into the results. Understanding the impact of these parameters can offer a different perspective for interpreting and understanding the empirical data.

For example, Hickner et al. reported a random copolymer benzyltrimethylammonium polysulfone via polycondensation of ionizable (tetramethyl bisphenol A) and non-ionizable (4-fluorophenyl sulfone and 4,4’-biphenol) monomers.[44] The theoretical composition is characterized by a sulfone monomer alternating with either an ionizable bisphenol A monomer or a non-ionizable biphenol monomer. Compared to a homopolymer with similar ion content, these random copolymers exhibited lower water uptake. The differences in macroscopic behavior were attributed to facilitated ion clustering in the random copolymer. Small angle X- ray scattering (SAXS) data for these samples detected no distinct local heterogeneities.

However, the scattering range was in the 0.1 nm-1 < q < 1 nm-1 range, corresponding to approximate domain sizes in between 6 and 60 nm. As understood by the theory, the local heterogeneities correspond to the statistical size of a block; as these polymers were synthesized from monomer constituents, it is feasible that the phase-separated domain sizes would be smaller than that probed in the SAXS experiment. In addition,

154 Chapter 7 polysulfones are known to be relatively flexible, with characteristic ratios ca. 2 owing to free rotation around their ether bonds. As predicted by the theory, increased flexibility at λ< λL results in higher critical χG; thus, it could also be reasonable that the χ parameter between the monomer components is insufficient for phase-separation, and the membrane exists in a mixed, homogeneous phase.

Watanabe et al. have prepared several different multiblock copolymers with alternating (λ=-1) hydrophobic and hydrophilic blocks.[47,133,194,195] In one report, they synthesized copolymers where using a fluorinated phenylene/biphenylene hydrophobic oligomer and quaternary ammonium-functionalized diphenyl ether or diphenyl sulfide as the hydrophilic oligomer. The morphologies of these materials were compared to a previous report in which the hydrophilic oligomer was composed of a bulky tetra-functional fluorenyl group as opposed to a diphenyl chalcogen group.

While both materials showed phase-separation under TEM, the authors noted that copolymers with fluorenyl blocks resulted in sharper phase boundaries and more defined segregation compared to those with diphenyl chalcogen blocks and attributed this to the bulkiness of the fluorenyl moitie. This result can partially be rationalized in the context of chain stiffness, where the use of a bulky fluorenyl group yields a stiffer backbone and consequently suppresses the ODT, thus yielding stronger segregation at similar chi interaction parameters.

Of course, the actual parameter space in evaluating the structure-property relationship of these materials is enormous, and the preceding discussion is a highly simplified account in an effort to highlight key parameters that can fundamentally affect their phase behavior. Indeed, these parameters are overlooked and not explicitly

155 Chapter 7 accounted for in the experimental design for these materials, yet are instrumental in determining the phase-separated morphologies critical for efficient ion transport. The ultimate goal is to corroborate the theory and simulations with experimental results and provide a robust set of guidelines for AEM materials design.

While we have not reached the point in which we can rigorously provide specific set of design principles, we can nevertheless assess current trends and approaches in the context of our results. Of particular interest is that much of the random multiblock copolymer AEM literature consists of perfectly alternating olgiomeric sequences, yet our theory and simulations suggest that higher chemical correlations leads to larger domain sizes. In light of this, nudging the chemistry towards more correlated reactivity (λ > -1) may prove beneficial in establishing a more efficient ion-transport morphology. In addition, the majority of these materials are comprised of bulky aromatic groups, which can impart rigidity to the overall polymer chain; yet, as we have demonstrated, rigidity confines polymer conformations, leading to energetic frustrations and locally trapped states that can impart dead-ends and a more tortuous morphology. Thus, a more flexible polymer may facilitate ion transport, although this recommendation must be optimized in light of decreased mechanical strength.

7.5 Conclusion and Outlook

Random copolymers represent the majority of alkaline exchange membranes, yet their fundamental phase behavior have not been well-studied. Guiding future experimental design therefore requires a better understanding of the parameters affecting phase- segregation and morphology in these materials. Mean-field theory and Monte Carlo

156 Chapter 7 simulations show that chain rigidity and compositional distribution in random copolymer melts can significantly influence their order-disorder transition and phase- separated morphologies. Efforts are underway to rigorously validate the simulation with the theory, and to corroborate these approaches with empirical results via scattering experiments. Ultimately, we aim to provide a platform for understanding the morphology of random copolymers and how that morphology translates to macroscopic performance and, in the process, establish a set of principles towards guiding the design and synthesis of AEM materials.

157 Chapter 8

Chapter 8: Conclusions

8.1 Summary

Although alkaline anion exchange membranes promise a low-temperature fuel cell solution free from precious metal catalysts, their poor stability and low performance must be addressed before they can become commercially viable. Indeed, the slower diffusion of the hydroxide anion compared to the proton presents an intrinsic limitation in the ion transport kinetics of alkaline exchange membrane fuel cells and an inherent barrier to displacing the popular proton exchange membrane fuel cell. While it is possible to offset the lower mobility of the ions by increasing the overall charge density, doing so results in excessive, omsmotic-presure driven water- uptake, resulting in poor mechanical durability and robustness. Moreover, compared to

Nafion and similar PEMs, typical AEMs lack a well-defined morphology for ion transport, resulting in slow, tortuous diffusion and high resistive losses.

I have presented different design rationales and synthetic approaches in addressing these challenges through modification of the popular quaternary ammonium polysulfone archetype. The overall experimental design of these materials is manifest in a structure-property motif, wherein I aim to effect better macroscopic properties through tuning molecular composition. In Chapter 5, I showed the ability to significantly reduce water uptake while preserving high charge concentrations in quaternary ammonium polysulfone by reinforcing the linear ionomer with an inert poly(styrene-co-dinvylbenzene) network. The resulting semi-interpenetrating network

158 Chapter 8 showed improved temperature and dimensional stability and offered promising device performance. The design rationale in Chapter 6 focused on increasing hydroxide transport efficiency in a quaternary ammonium polysulfone AEM by effecting a more efficient ion transport morphology via hydrophilic grafts. Specifically, I showed that grafting poly(ethylene glycol) sidechains along a quaternary ammonium polysulfone backbone enabled phase-separation and formation of hydrophilic, ion transport domains. As a result, the hydrophilic sidechains doubled the hydroxide transport efficiency and offered a significant increase in in-plane conductivity and device performance.

Finally, Chapter 7 describes collaborative efforts towards developing a theoretical framework for understanding phase-separation and morphology in random copolymers and its implications in the design of AEM materials. We showed that both the chain stiffness and stochastic distribution of chemical identities along a random copolymer backbone, parameters often overlooked in the experimental literature, can profoundly influence their phase-transition behavior and morphologies.

8.2 Outlook

We used quaternary ammonium polysulfone as both a starting material and benchmark in our materials design. However, synthetic protocols developed in this thesis are not limited to quaternary ammonium polysulfones and can be extended to other classes of aromatic polymer electrolyte materials. Indeed, it may be necessary to evaluate other classes of AEMs given the poor alkaline stability of both the quaternary ammonium headgroup and the functionalized polysulfone backbone, both of which have been shown to rapidly degrade under high pH conditions. While the approaches described

159 Chapter 8 here offer a modest improvement in alkaline stability over the base quaternary ammonium polysulfone membranes, the materials are nonetheless ill suited for long- term operation. It would be prudent to apply these same design principals with more stable backbones (e.g., poly(phenylene oxide)) and more stable cationic headgroups

(e.g., phosphonium or sulfonium).

The theory and simulation work continues to be refined, and we hope to corroborate the predicted results with experimental data. In particular, the structure factor from the theory and simulation work can naturally be juxtaposed against empirical scattering results. The current work offers a high-level treatment of random copolymers, and more complexity and nuances must be accounted for to realistically reflect a physical AEM material. For example, the current study focuses exclusively on a polymer melt, while real polymer electrolyte membranes are equilibrated with water, which may change the ultimate morphology of the material. Nevertheless, our current results provide a framework with which to assess the degree to which various physical parameters affects the phase behavior of a random copolymer. Ultimately, we hope that the results from further development and refinement can offer a robust set of guidelines for designing AEM materials. From a larger perspective, the fundamental understanding garnered from furthering this study can elucidate the structure-property relationships of general random copolymer systems.

160

Appendix A: Morphology of Semi-Interpenetrating Network

Anion Exchange Membranes

The materials discussed in this section refer to those reported in Chapter 5. Our initial synthesis approach involved attempts to thermally polymerize a solution of quaternary ammonium polysulfone and styrene and divinylbenzene monomers; however, this protocol resulted in macroscopic heterogeneities and demixing, indicating significant phase incompatibility between the charged polymer and the aromatic monomers. To produce a macroscopically uniform film, we adopted the protocol described in section

5.2.1, where we first produce a semi-IPN of the chloromethylated polysulfone and the styrene/divinylbenzene monomers and subject the membrane to a post-treatment process to introduce the quaternary ammonium charge groups. The resulting film was macroscopically uniform; however, we suspected that the incompatibility between the quaternary ammonium headgroups and the poly(styrene-co-divinylbezene) network may lead to mesoscale phase-separation between the linear ionomer and its reinforcing scaffold. This hypothesis is supported by the conductivity results, which suggested that the poly(styrene-co-divinylbenzene) network was sufficiently phase-separated from the ionically active portions of the material given that its presence had little influence on the apparent activation energy for ion transport.

In light of this, we performed small angle X-Ray scattering (SAXS) measurements in order to probe the morphology of the semi-IPN membranes (Figure

A.1); radial averaging of the data was performed using the Irena software package for

Igor.[196]

161

Figure A.1 SAXS data for the semi-IPN anion exchange membranes.

The lack of a distinct scattering feature measured suggests that there is no well- defined heterogeneity or structure on the length-scales probed in the experiment. This is consistent with previous SAXS investigations of quaternary ammonium-based polysulfones and suggests that the semi-IPN architecture does not significantly influence polymer structure at this lengthscale.[44,140] An upturn at lower scattering vector q suggests disperse, large length-scale heterogeneity resulting from the polymer matrix.

On the other hand, we might have expected phase contrast arising from the local thermodynamic incompatibility between the charged, hydrophilic quaternary ammonium polysulfone and the nonpolar, hydrophobic poly(styrene-co- divinylbenzene). The lack of a distinct feature in the SAXS spectrum may be due to

162

several reasons. One reason could be that there is insufficient electron density contrast beween the two components. It is also possible that the characteristic domain size is outside of the range probed in the SAXS experiment. Moreover, it could be likely that the local phase-segregation exhibit a fuzzy phase-boundary and lack well-defined ordering, resulting in a broad and disperse scattering pattern.

In any case, preliminary dynamic mechanical analysis (DMA) data show two glass transition temperatures (determined by peaks in the loss modulus) in the semi-

IPN materials as a result of compositional heterogeneity, suggesting phase-separation

(Figure A.2). Moreover, the two glass transition temperatures are shifted from the individual Tg of the pure QA PSf baseline (240.56°C) and poly(styrene-co- divinylbenzene) (ca. 100°C to 150°C depending on relative composition), indicating partial miscibility between the two components. The Flory-Fox equation is a semi- empirical relation that allows determination of weight composition of different components in a miscible polymer blend and is given by:

1 � � = ! + ! �! �!,! �!,! where Tg is the shifted glass transition temperature of the blend, Tg,n is the glass transition temperature of pure component n, and wn is the weight fraction of component n. This relation was used to approximate the relative composition of the two phases observed in the DMA (Table A.1). Note that these results are preliminary, and more characterization and data are needed to properly elucidate phase behavior and quantify compositional heterogeneities; nevertheless, it is evident from the DMA data that there is some degree of phase separation as well as partial miscibility.

163

Table A.1 Phase composition of the DMA data as estimated by the Flory-Fox equation.

Sample Phase Tg (°C) wt.% PS-co-DVB wt.% QA PSf 1 177.88 58% 42% QA sIPN [90/7/3] 2 225.41 11% 89% 1 167.72 79% 21% QA sIPN [72/20/8] 2 203.96 29% 71%

250 QA PSf––––––– QA sIPNt [90/7/3]– – – – QA sIPN [72/20/8]––––– ·

203.96°C

200

150 240.56°C

225.41°C

100 Loss Modulus (MPa)

177.88°C

162.72°C 50

0 0 50 100 150 200 250 300 Temperature (°C) Universal V4.5A TA Instruments

Figure A.2 Loss modulus (G’’) as a function of temperature for the semi-IPN membranes and

QA PSf baseline. Peaks represent glass transition temperatures. Measurements performed on a

TA Instruments Q800 Dynamic Mechanical Analyzer in tension mode at fixed frequency (1

Hz) and strain (1%) with a 10°C/min temperature ramp.

164

Figure A.3 Phase contrast tapping mode AFM image of QA sIPN [90/7/3] membrane equilibrated under ambient conditions (22°C, ca. 35% RH). Image was recorded by Joseph

Troderman.

The phase separation suggested by DMA characterization is supported by tapping mode atomic force microscopy of QA sIPN [90/7/3], which shows distinct phase contrast with domains on the order of 10-100 nm (Figure A.3). A smaller phase angle indicates a more elastic response and stiffer material, and therefore reflects the harder, PS-co-DVB-rich phase; conversely a larger phase angle indicates softer regions that correspond to QA PSf-rich domains. The AFM image shows three distinct phase angles in the tapping mode study (ca. 0°, 15°, and >30°), suggesting intermixing and partial miscibility between the QA PSf and PS-co-DVB components. While all the morphological data are preliminary, all the evidence we have gathered so far indicate phase-separation within the semi-IPN materials. More thorough characterization need

165

to be conducted to examine how surface effects (e.g., roughness, thermodynamics) affects the structure probed by tapping mode AFM.

166

Appendix B: Alternative Semi-IPN Synthesis Protocol with

Thermal-Initiated Polymerization

The work discussed in Chapter 5 concerning semi-interpenetrating network development and characterization was performed on materials made via photopolymerization, wherein a chloromethylated polysulfone film swells to equilibrium in a solution of styrene and divinylbenzene monomers. This section describes a different protocol for a thermal initiated polymerization method that allows more direct control on the mass ratios of the constituent materials within in the semi-

IPN.

Specifically, a known quantity of chloromethylated polysulfone, styrene, divinylbenzene, and azoisobutrylnitrile (AIBN, 1 mol % of double bonds) are dissolved in dimethylformamide to 10 wt.% solids. The solution is then cast onto a petri dish and placed in an oven at 80°C for 18 hours, after which the semi-IPN film is removed. Note that chloromethylated polysulfone is used in lieu of quaternary ammonium polysulfone, as the charge groups introduce poor miscibility with the monomers and results in macroscopic de-mixing and heterogeneities (Figure B.1).

Because of this, the resulting film must be quaternized with trimethylamine to produce an ionomer. However, initial attempts to do so by soaking the film in a 4.2 M trimethylamine (TMA) in ethanol solution failed as the film was resistant to trimethylamine and ethanol uptake. In light of this, the TMA solution was mixed with

5 v/v% DMF as a co-solvent; the semi-IPN films were then swelled in this mixture for

24 hours to convert the chloromethyl groups into benzyltrimethylammonium moieties.

167

Initial results from the thermal-polymerization approach are promising, with membranes exhibiting hydroxide conductivities ca. 40 mS cm-1 at room temperature and fully hydrated conditions. In contrast to the photopolymerization method described in Chapter 5, the synthesis protocol described here allows direct control of the mass ratios of the various polymer components. However, it is likely that the thermal polymerization and processing method may lead to different properties than the materials reported in Chapter 5 (even at similar compositions) due to differences in reaction kinetics and thermodynamic conditions. A comparison between the two methods may provide valuable structure-property-processing insight, and further tuning of this protocol may lead to enhanced bulk properties.

168

References

[1] “Types of Fuel Cells | Department of Energy,” can be found under http://energy.gov/eere/fuelcells/types-fuel-cells, 2015.

[2] J. Spendelow, J. Marcinkoski, DOE Fuel Cell Technologies Office: Fuel Cell System Cost (2013), 2014.

[3] J. W. Desmond Ng, Y. Gorlin, T. Hatsukade, T. F. Jaramillo, Adv. Energy Mater. 2013, 3, 1545.

[4] S. Lu, J. Pan, A. Huang, L. Zhuang, J. Lu, Proc. Natl. Acad. Sci. 2008, 105, 20611.

[5] M. Warshay, P. R. Prokopius, J. Power Sources 1990, 29, 193.

[6] N. Agmon, Chem. Phys. Lett. 2000, 319, 247.

[7] M. E. Tuckerman, D. Marx, M. Parrinello, Nature 2002, 417, 925.

[8] D. Marx, A. Chandra, M. E. Tuckerman, Chem. Rev. 2010, 110, 2174.

[9] A. Botti, F. Bruni, S. Imberti, M. A. Ricci, A. K. Soper, J. Chem. Phys. 2003, 119, 5001.

[10] A. Botti, F. Bruni, S. Imberti, M. A. Ricci, A. K. Soper, J. Chem. Phys. 2004, 120, 10154.

[11] D. Asthagiri, L. R. Pratt, J. D. Kress, M. A. Gomez, Proc. Natl. Acad. Sci. U. S. A. 2004, 101, 7229.

[12] D. Asthagiri, L. R. Pratt, J. . Kress, M. . Gomez, Chem. Phys. Lett. 2003, 380, 530.

[13] D. H. Lee, C. H. Choi, T. H. Choi, B. J. Sung, H. Kang, J. Phys. Chem. Lett. 2014, 5, 2568.

[14] P. Choi, N. H. Jalani, R. Datta, J. Electrochem. Soc. 2005, 152, E123.

[15] K. N. Grew, W. K. S. Chiu, J. Electrochem. Soc. 2010, 157, B327.

[16] K. M. Beers, D. T. Hallinan, X. Wang, J. A. Pople, N. P. Balsara, Macromolecules 2011, 44, 8866.

169

[17] Y. Ye, Y. A. Elabd, Polymer (Guildf). 2011, 52, 1309.

[18] K. Schmidt-Rohr, Q. Chen, Nat. Mater. 2008, 7, 75.

[19] H. Long, K. Kim, B. S. Pivovar, J. Phys. Chem. C 2012, 116, 9419.

[20] J. R. Varcoe, P. Atanassov, D. R. Dekel, A. M. Herring, M. A. Hickner, P. A. Kohl, A. R. Kucernak, W. E. Mustain, K. Nijmeijer, K. Scott, T. Xu, L. Zhuang, Energy Environ. Sci. 2014, 7, 3135.

[21] S. Chempath, B. R. Einsla, L. R. Pratt, C. S. Macomber, J. M. Boncella, J. A. Rau, B. S. Pivovar, J. Phys. Chem. C 2008, 112, 3179.

[22] S. A. Nuñez, M. A. Hickner, ACS Macro Lett. 2013, 2, 49.

[23] M. Tomoi, K. Yamaguchi, R. Ando, Y. Kantake, Y. Aosaki, H. Kubota, J. Appl. Polym. Sci. 1997, 64, 1161.

[24] C. Fujimoto, D.-S. Kim, M. Hibbs, D. Wrobleski, Y. S. Kim, J. Memb. Sci. 2012, 423-424, 438.

[25] C. G. Arges, V. Ramani, Proc. Natl. Acad. Sci. U. S. A. 2013, 110, 2490.

[26] J. Parrondo, V. Ramani, J. Electrochem. Soc. 2014, 161, F1015.

[27] H. Yanagi, K. Fukuta, in ECS Trans., ECS, 2008, pp. 257–262.

[28] J. Kizewski, N. Mudri, R. Zeng, S. Poynton, R. C. T. Slade, J. R. Varcoe, in ECS Trans., The Electrochemical Society, 2010, pp. 27–35.

[29] Y. Luo, J. Guo, C. Wang, D. Chu, ChemSusChem 2011, 4, 1557.

[30] P. Zschocke, D. Quellmalz, J. Memb. Sci. 1985, 22, 325.

[31] C. G. Arges, M.-S. Jung, G. Johnson, J. Parrondo, V. Ramani, in ECS Trans., 2011, pp. 1795–1816.

[32] C. G. Arges, J. Parrondo, G. Johnson, A. Nadhan, V. Ramani, J. Mater. Chem. 2012, 22, 3733.

[33] J. Pan, S. Lu, Y. Li, A. Huang, L. Zhuang, J. Lu, Adv. Funct. Mater. 2010, 20, 312.

[34] G. Wang, Y. Weng, D. Chu, R. Chen, D. Xie, J. Memb. Sci. 2009, 332, 63.

170

[35] E. N. Komkova, D. F. Stamatialis, H. Strathmann, M. Wessling, J. Memb. Sci. 2004, 244, 25.

[36] J. Pan, Y. Li, L. Zhuang, J. Lu, Chem. Commun. 2010, 46, 8597.

[37] J. Wang, J. Wang, S. Li, S. Zhang, J. Memb. Sci. 2011, 368, 246.

[38] J. Wang, Z. Zhao, F. Gong, S. Li, S. Zhang, Macromolecules 2009, 42, 8711.

[39] J. Zhou, M. Unlu, J. A. Vega, P. A. Kohl, J. Power Sources 2009, 190, 285.

[40] J. J. Fang, J. Memb. Sci. 2006, 285, 317.

[41] L. Li, Y. Wang, J. Memb. Sci. 2005, 262, 1.

[42] X. Li, G. Nie, J. Tao, W. Wu, L. Wang, S. Liao, ACS Appl. Mater. Interfaces 2014, 6, 7585.

[43] G. Nie, X. Li, J. Tao, W. Wu, S. Liao, J. Memb. Sci. 2015, 474, 187.

[44] J. Yan, M. A. Hickner, Macromolecules 2010, 43, 2349.

[45] N. Li, Q. Zhang, C. Wang, Y. M. Lee, M. D. Guiver, Macromolecules 2012, 45, 2411.

[46] J. Choi, Y.-J. Byun, S. Y. Lee, J. H. Jang, D. Henkensmeier, S. J. Yoo, S.-A. Hong, H.-J. Kim, Y.-E. Sung, J.-S. Park, Int. J. Hydrogen Energy 2014, 39, 21223.

[47] M. Tanaka, M. Koike, K. Miyatake, M. Watanabe, Macromolecules 2010, 43, 2657.

[48] A. Amel, L. Zhu, M. Hickner, Y. Ein-Eli, J. Electrochem. Soc. 2014, 161, F615.

[49] Y. Li, T. Xu, M. Gong, J. Memb. Sci. 2006, 279, 200.

[50] X. Tongwen, Y. Weihua, J. Memb. Sci. 2001, 190, 159.

[51] L. Wu, T. Xu, W. Yang, J. Memb. Sci. 2006, 286, 185.

[52] L. Wu, T. Xu, D. Wu, X. Zheng, J. Memb. Sci. 2008, 310, 577.

[53] L. Wu, T. Xu, J. Memb. Sci. 2008, 322, 286.

171

[54] K. H. Gopi, S. G. Peera, S. D. Bhat, P. Sridhar, S. Pitchumani, Int. J. Hydrogen Energy 2014, 39, 2659.

[55] N. Li, M. D. Guiver, W. H. Binder, ChemSusChem 2013, 6, 1376.

[56] X. Yan, G. He, S. Gu, X. Wu, L. Du, H. Zhang, J. Memb. Sci. 2011, 375, 204.

[57] X. Wu, W. Chen, X. Yan, G. He, J. Wang, Y. Zhang, X. Zhu, J. Mater. Chem. A 2014, 2, 12222.

[58] J. Wang, G. He, X. Wu, X. Yan, Y. Zhang, Y. Wang, L. Du, J. Memb. Sci. 2014, 459, 86.

[59] Y. Xiong, J. Fang, Q. H. Zeng, Q. L. Liu, J. Memb. Sci. 2008, 311, 319.

[60] A. M. Maes, T. P. Pandey, M. A. Vandiver, L. K. Lundquist, Y. Yang, J. L. Horan, A. Krosovsky, M. W. Liberatore, S. Seifert, A. M. Herring, Electrochim. Acta 2013, 110, 260.

[61] Y. Wan, B. Peppley, K. A. M. Creber, V. T. Bui, E. Halliop, J. Power Sources 2008, 185, 183.

[62] Y. Wan, B. Peppley, K. A. M. Creber, V. T. Bui, J. Power Sources 2010, 195, 3785.

[63] Y. Xiong, Q. L. Liu, Q. G. Zhang, A. M. Zhu, J. Power Sources 2008, 183, 447.

[64] T. N. Danks, R. C. T. Slade, J. R. Varcoe, J. Mater. Chem. 2003, 13, 712.

[65] J. VARCOE, R. SLADE, Electrochem. commun. 2006, 8, 839.

[66] T. J. Clark, N. J. Robertson, H. a Kostalik, E. B. Lobkovsky, P. F. Mutolo, H. D. Abruña, G. W. Coates, J. Am. Chem. Soc. 2009, 131, 12888.

[67] H. A. Kostalik, T. J. Clark, N. J. Robertson, P. F. Mutolo, J. M. Longo, H. D. Abruña, G. W. Coates, Macromolecules 2010, 43, 7147.

[68] N. J. Robertson, H. A. Kostalik, T. J. Clark, P. F. Mutolo, H. D. Abruña, G. W. Coates, J. Am. Chem. Soc. 2010, 132, 3400.

[69] K. N. Grew, D. Chu, W. K. S. Chiu, J. Electrochem. Soc. 2010, 157, B1024.

[70] M. R. Hibbs, M. A. Hickner, T. M. Alam, S. K. McIntyre, C. H. Fujimoto, C. J. Cornelius, Chem. Mater. 2008, 20, 2566.

172

[71] S. S. Gu, R. Cai, T. Luo, Z. Chen, M. Sun, Y. Liu, G. He, Y. Yan, Angew. Chem. Int. Ed. Engl. 2009, 48, 6499.

[72] S. Gu, R. Cai, T. Luo, K. Jensen, C. Contreras, Y. Yan, ChemSusChem 2010, 3, 555.

[73] H. R. Hays, R. G. Laughlin, J. Org. Chem. 1967, 32, 1060.

[74] R. U. Pagilagan, W. E. McEwen, Chem. Commun. 1966, 652.

[75] S. Gu, R. Cai, Y. Yan, Chem. Commun. (Camb). 2011, 47, 2856.

[76] X. Yan, S. Gu, G. He, X. Wu, W. Zheng, X. Ruan, J. Memb. Sci. 2014, 466, 220.

[77] X. Yan, S. Gu, G. He, X. Wu, J. Benziger, J. Power Sources 2014, 250, 90.

[78] L. Jiang, X. Lin, J. Ran, C. Li, L. Wu, T. Xu, Chinese J. Chem. 2012, 30, 2241.

[79] K. J. T. Noonan, K. M. Hugar, H. A. Kostalik, E. B. Lobkovsky, H. D. Abruña, G. W. Coates, J. Am. Chem. Soc. 2012, 134, 18161.

[80] T. J. Clark, N. J. Robertson, H. A. Kostalik, E. B. Lobkovsky, P. F. Mutolo, H. D. Abruña, G. W. Coates, J. Am. Chem. Soc. 2009, 131, 12888.

[81] N. J. Robertson, H. A. Kostalik, T. J. Clark, P. F. Mutolo, H. D. Abruña, G. W. Coates, J. Am. Chem. Soc. 2010, 132, 3400.

[82] O. I. Deavin, S. Murphy, A. L. Ong, S. D. Poynton, R. Zeng, H. Herman, J. R. Varcoe, Energy Environ. Sci. 2012, 5, 8584.

[83] Y. Liu, J. Wang, Y. Yang, T. M. Brenner, S. Seifert, Y. Yan, M. W. Liberatore, A. M. Herring, J. Phys. Chem. C 2014, 118, 15136.

[84] B. Qiu, B. Lin, L. Qiu, F. Yan, J. Mater. Chem. 2012, 22, 1040.

[85] M. Guo, J. Fang, H. Xu, W. Li, X. Lu, C. Lan, K. Li, J. Memb. Sci. 2010, 362, 97.

[86] B. Lin, L. Qiu, J. Lu, F. Yan, Chem. Mater. 2010, 22, 6718.

[87] F. Zhang, H. Zhang, C. Qu, J. Mater. Chem. 2011, 21, 12744.

[88] A. H. N. Rao, R. L. Thankamony, H.-J. Kim, S. Nam, T.-H. Kim, Polymer (Guildf). 2013, 54, 111.

173

[89] Y. Ye, Y. A. Elabd, Macromolecules 2011, 44, 8494.

[90] D. Chen, M. A. Hickner, ACS Appl. Mater. Interfaces 2012, 4, 5775.

[91] O. M. M. Page, S. D. Poynton, S. Murphy, A. Lien Ong, D. M. Hillman, C. A. Hancock, M. G. Hale, D. C. Apperley, J. R. Varcoe, RSC Adv. 2013, 3, 579.

[92] B. Lin, H. Dong, Y. Li, Z. Si, F. Gu, F. Yan, Chem. Mater. 2013, 25, 1858.

[93] S. C. Price, K. S. Williams, F. L. Beyer, ACS Macro Lett. 2014, 3, 160.

[94] D. Henkensmeier, H.-R. Cho, H.-J. Kim, C. Nunes Kirchner, J. Leppin, A. Dyck, J. H. Jang, E. Cho, S.-W. Nam, T.-H. Lim, Polym. Degrad. Stab. 2012, 97, 264.

[95] D. Henkensmeier, H.-J. Kim, H.-J. Lee, D. H. Lee, I.-H. Oh, S.-A. Hong, S.-W. Nam, T.-H. Lim, Macromol. Mater. Eng. 2011, 296, 899.

[96] O. D. Thomas, K. J. W. Y. Soo, T. J. Peckham, M. P. Kulkarni, S. Holdcroft, J. Am. Chem. Soc. 2012, 134, 10753.

[97] L. Jheng, S. L. Hsu, B. Lin, Y. Hsu, J. Memb. Sci. 2014, 460, 160.

[98] J. Wang, S. Li, S. Zhang, Macromolecules 2010, 43, 3890.

[99] Q. Zhang, S. Li, S. Zhang, Chem. Commun. 2010, 46, 7495.

[100] L. Liu, Q. Li, J. Dai, H. Wang, B. Jin, R. Bai, J. Memb. Sci. 2014, 453, 52.

[101] J. Qiao, J. Fu, L. Liu, J. Zhang, J. Xie, G. Li, Solid State Ionics 2012, 214, 6.

[102] F. Gu, H. Dong, Y. Li, Z. Sun, F. Yan, Macromolecules 2014, 47, 6740.

[103] B. Zhang, S. Gu, J. Wang, Y. Liu, A. M. Herring, Y. Yan, RSC Adv. 2012, 2, 12683.

[104] Y. Zha, M. L. Disabb-Miller, Z. D. Johnson, M. A. Hickner, G. N. Tew, J. Am. Chem. Soc. 2012, 134, 4493.

[105] M. L. Disabb-Miller, Y. Zha, A. J. DeCarlo, M. Pawar, G. N. Tew, M. A. Hickner, Macromolecules 2013, 46, 9279.

[106] Y.-C. Cao, K. Scott, X. Wang, Int. J. Hydrogen Energy 2012, 37, 12688.

174

[107] X. Wang, M. Li, B. T. Golding, M. Sadeghi, Y. Cao, E. H. Yu, K. Scott, Int. J. Hydrogen Energy 2011, 36, 10022.

[108] J. Hu, D. Wan, W. Zhu, L. Huang, S. Tan, X. Cai, X. Zhang, ACS Appl. Mater. Interfaces 2014, 6, 4720.

[109] Y. Zhao, H. Yu, F. Xie, Y. Liu, Z. Shao, B. Yi, J. Power Sources 2014, 269, 1.

[110] Y. Zhao, J. Pan, H. Yu, D. Yang, J. Li, L. Zhuang, Z. Shao, B. Yi, Int. J. Hydrogen Energy 2013, 38, 1983.

[111] G. Li, J. Pan, J. Han, C. Chen, J. Lu, L. Zhuang, J. Mater. Chem. A 2013, 1, 12497.

[112] T. Y. Guo, Q. H. Zeng, C. H. Zhao, Q. L. Liu, A. M. Zhu, I. Broadwell, J. Memb. Sci. 2011, 371, 268.

[113] A. M. Park, P. N. Pintauro, in ECS Trans., The Electrochemical Society, 2011, pp. 1817–1826.

[114] A. M. Park, P. N. Pintauro, Electrochem. Solid-State Lett. 2012, 15, B27.

[115] A. M. Park, F. E. Turley, R. J. Wycisk, P. N. Pintauro, Macromolecules 2014, 47, 227.

[116] A. M. Park, F. E. Turley, R. Wycisk, P. N. Pintauro, ECS Trans. 2013, 58, 405.

[117] S. Roddecha, Z. Dong, Y. Wu, M. Anthamatten, J. Memb. Sci. 2012, 389, 478.

[118] E. D. Wang, T. S. Zhao, W. W. Yang, Int. J. Hydrogen Energy 2010, 35, 2183.

[119] C.-C. Yang, S.-J. Chiu, W.-C. Chien, S.-S. Chiu, J. Power Sources 2010, 195, 2212.

[120] X. Li, Y. Yu, Y. Meng, ACS Appl. Mater. Interfaces 2013, 5, 1414.

[121] X. Li, J. Tao, G. Nie, L. Wang, L. Li, S. Liao, RSC Adv. 2014, 4, 41398.

[122] T.-H. Tsai, S. P. Ertem, A. M. Maes, S. Seifert, A. M. Herring, E. B. Coughlin, Macromolecules 2015, 48, 655.

[123] N. Li, T. Yan, Z. Li, T. Thurn-Albrecht, W. H. Binder, Energy Environ. Sci. 2012, 5, 7888.

175

[124] N. Li, Y. Leng, M. A. Hickner, C.-Y. Wang, J. Am. Chem. Soc. 2013, 135, 10124.

[125] N. Li, L. Wang, M. Hickner, Chem. Commun. (Camb). 2014, 50, 4092.

[126] A. N. Lai, L. S. Wang, C. X. Lin, Y. Z. Zhuo, Q. G. Zhang, A. M. Zhu, Q. L. Liu, J. Memb. Sci. 2015, 481, 9.

[127] J. Pan, C. Chen, Y. Li, L. Wang, L. Tan, G. Li, X. Tang, L. Xiao, J. Lu, L. Zhuang, Energy Environ. Sci. 2014, 7, 354.

[128] H.-S. Dang, E. A. Weiber, P. Jannasch, J. Mater. Chem. A 2015, 3, 5280.

[129] Q. Li, L. Liu, Q. Miao, B. Jin, R. Bai, Chem. Commun. (Camb). 2014, 50, 2791.

[130] J. Ran, L. Wu, T. Xu, Polym. Chem. 2013, 4, 4612.

[131] J. Ran, L. Wu, B. Wei, Y. Chen, T. Xu, Sci. Rep. 2014, 4, 6486.

[132] M. A. Hossain, Y. Lim, S. Lee, H. Jang, S. Choi, Y. Jeon, J. Lim, W. G. Kim, Int. J. Hydrogen Energy 2014, 39, 2731.

[133] M. Tanaka, K. Fukasawa, E. Nishino, S. Yamaguchi, K. Yamada, H. Tanaka, B. Bae, K. Miyatake, M. Watanabe, J. Am. Chem. Soc. 2011, 133, 10646.

[134] N. Yokota, M. Shimada, H. Ono, R. Akiyama, E. Nishino, K. Asazawa, J. Miyake, M. Watanabe, K. Miyatake, Macromolecules 2014, 47, 8238.

[135] X. Li, Y. Yu, Q. Liu, Y. Meng, J. Memb. Sci. 2013, 436, 202.

[136] Q. Zeng, Q. Liu, I. Broadwell, A. Zhu, Y. Xiong, X. Tu, J. Memb. Sci. 2009, DOI 10.1016/j.memsci.2009.11.051.

[137] S. S. Jang, V. Molinero, T. Çaǧın, W. A. Goddard, J. Phys. Chem. B 2004, 108, 3149.

[138] Y.-L. S. Tse, A. M. Herring, K. Kim, G. A. Voth, J. Phys. Chem. C 2013, 117, 8079.

[139] B. V. Merinov, W. A. Goddard, J. Memb. Sci. 2013, 431, 79.

[140] S. S. He, C. W. Frank, J. Mater. Chem. A 2014, 2, 16489.

[141] S. S. He, A. L. Strickler, C. W. Frank, ChemSusChem 2015, 8, 1472.

176

[142] E. Avram, E. Butuc, C. Luca, I. Druta, J. Macromol. Sci. Part A 1997, 34, 1701.

[143] J. Ilavsky, J. Appl. Crystallogr. 2012, 45, 324.

[144] G. Merle, M. Wessling, K. Nijmeijer, J. Memb. Sci. 2011, 377, 1.

[145] N. Agmon, Chem. Phys. Lett. 2000, 319, 247.

[146] M. A. Hickner, A. M. Herring, E. B. Coughlin, J. Polym. Sci. Part B Polym. Phys. 2013, 51, 1727.

[147] J. R. Varcoe, R. C. T. Slade, Fuel Cells 2005, 5, 187.

[148] P. P. Kundu, B. T. Kim, J. E. Ahn, H. S. Han, Y. G. Shul, J. Power Sources 2007, 171, 86.

[149] K.-Y. Cho, H.-Y. Jung, S.-S. Shin, N.-S. Choi, S.-J. Sung, J.-K. Park, J.-H. Choi, K.-W. Park, Y.-E. Sung, Electrochim. Acta 2004, 50, 588.

[150] T. Li, G. Zhong, R. Fu, Y. Yang, J. Memb. Sci. 2010, 354, 189.

[151] Y.-J. Choi, M.-S. Kang, S.-H. Moon, J. Appl. Polym. Sci. 2003, 88, 1488.

[152] L. Chikh, V. Delhorbe, O. Fichet, J. Memb. Sci. 2011, 368, 1.

[153] J. Wang, R. He, Q. Che, J. Colloid Interface Sci. 2011, 361, 219.

[154] C. Iojoiu, J.-Y. J.-Y. Sanchez, High Perform. Polym. 2009, 21, 673.

[155] H. Nakagawa, S. Tsuge, Macromolecules 1985, 18, 2068.

[156] J. Pan, S. Lu, Y. Li, A. Huang, L. Zhuang, J. Lu, Adv. Funct. Mater. 2010, 20, 312.

[157] C. Wohlfarth, in Data Extr. from Landolt-Börnstein VIII/6D2 Polym. Polym. Solut. Phys. Prop. Their Relations I (Thermodynamic Prop. PVT-Data Misc. Prop. Polym. Solut. (Eds.: M.D. Lechner, K.F. Arndt), Springer-Verlag Berlin Heidelberg, 2010.

[158] F. Bai, X. Yang, W. Huang, Macromolecules 2004, 37, 9746.

[159] Y. Miwa, O. Urakawa, A. Doi, K. Yamamoto, S. Nobukawa, J. Phys. Chem. B 2012, 116, 1282.

[160] N. Bach Van, C. Noel, J. Polym. Sci. Polym. Chem. Ed. 1976, 14, 1627.

177

[161] H. Schott, J. Macromol. Sci. Part B 1992, 31, 1.

[162] DuPont Nafion PFSA Membranes (Data Sheet), 2009.

[163] E. C. Muniz, G. Geuskens, Macromolecules 2001, 34, 4480.

[164] W.-F. Lee, Y.-J. Chen, J. Appl. Polym. Sci. 2001, 82, 2487.

[165] Yuesheng Ye, Yossef A. Elabd, Polymers for Energy Storage and Delivery: Polyelectrolytes for Batteries and Fuel Cells, American Chemical Society, Washington, DC, 2012.

[166] M. R. Sturgeon, C. S. Macomber, C. Engtrakul, B. S. Pivovar, ECS Trans. 2014, 64, 1201.

[167] M. Zhiani, S. Majidi, M. M. Taghiabadi, Fuel Cells 2013, 13, 946.

[168] N. T. Rebeck, Y. Li, D. M. Knauss, J. Polym. Sci. Part B Polym. Phys. 2013, 51, 1770.

[169] H. Zarrin, J. Wu, M. Fowler, Z. Chen, J. Memb. Sci. 2012, 394-395, 193.

[170] Z. Zhao, J. Wang, S. Li, S. Zhang, J. Power Sources 2011, 196, 4445.

[171] N. Li, M. D. Guiver, Macromolecules 2014, 47, 2175.

[172] B. Yameen, A. Kaltbeitzel, A. Langer, F. Müller, U. Gösele, W. Knoll, O. Azzaroni, Angew. Chem. Int. Ed. Engl. 2009, 48, 3124.

[173] E. Bortel, A. Kochanowski, Die Makromol. Chemie 1984, 185, 1409.

[174] J. Kang, W. Li, X. Wang, Y. Lin, X. Xiao, S. Fang, Electrochim. Acta 2003, 48, 2487.

[175] J. Y. Park, M. H. Acar, A. Akthakul, W. Kuhlman, A. M. Mayes, Biomaterials 2006, 27, 856.

[176] J. M. Serpico, S. G. Ehrenberg, J. J. Fontanella, X. Jiao, D. Perahia, K. A. McGrady, E. H. Sanders, G. E. Kellogg, G. E. Wnek, Macromolecules 2002, 35, 5916.

[177] M. Yoonessi, H. Heinz, T. D. Dang, Z. Bai, Polymer (Guildf). 2011, 52, 5615.

[178] M.-P. Nieh, M. D. Guiver, D. S. Kim, J. Ding, T. Norsten, Macromolecules 2008, 41, 6176.

178

[179] B. Steinmüller, M. Müller, K. R. Hambrecht, G. D. Smith, D. Bedrov, Macromolecules 2012, 45, 1107.

[180] S. H. Chen, S. L. Chang, R. Strey, J. Appl. Crystallogr. 1991, 24, 721.

[181] L. García-Tarrés, E. Guàrdia, J. Phys. Chem. B 1998, 102, 7448.

[182] S. Lüsse, K. Arnold, Macromolecules 1996, 29, 4251.

[183] G. H. Fredrickson, S. T. Milner, L. Leibler, Macromolecules 1992, 25, 6341.

[184] G. Fredrickson, S. Milner, Phys. Rev. Lett. 1991, 67, 835.

[185] A. von der Heydt, M. Müller, A. Zippelius, Macromolecules 2010, 43, 3161.

[186] A. A. Gavrilov, Y. V. Kudryavtsev, P. G. Khalatur, A. V. Chertovich, Chem. Phys. Lett. 2011, 503, 277.

[187] A. J. Spakowitz, Z.-G. Wang, Macromolecules 2004, 37, 5814.

[188] E. F. Koslover, A. J. Spakowitz, Macromolecules 2013, 46, 2003.

[189] S. Mehraeen, B. Sudhanshu, E. F. Koslover, A. J. Spakowitz, Phys. Rev. E 2008, 77, 061803.

[190] A. J. Spakowitz, Europhys. Lett. 2006, 73, 684.

[191] A. Spakowitz, Z.-G. Wang, Phys. Rev. Lett. 2003, 91, 166102.

[192] S. Mao, S. S. He, M. Eissien, E. Coletta, C. W. Frank, A. J. Spakowitz, Impact of Conformational and Chemical Correlations on Microphase Segregation in Random Copolymers (in Preparation), 2015.

[193] F. A. Detcheverry, D. Q. Pike, U. Nagpal, P. F. Nealey, J. J. de Pablo, Soft Matter 2009, 5, 4858.

[194] N. Yokota, H. Ono, J. Miyake, E. Nishino, K. Asazawa, M. Watanabe, K. Miyatake, ACS Appl. Mater. Interfaces 2014, 6, 17044.

[195] M. Watanabe, H. Uchida, Y. Seki, M. Emori, P. Stonehart, J. Electrochem. Soc. 1996, 143, 3847.

[196] J. Ilavsky, P. R. Jemian, J. Appl. Crystallogr. 2009, 42, 347.

179