Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Design and fabrication of electrospun nanofiber catalyst support for polymer electrolyte membrane fuel… Chan, Sophia Shuk Kwan 2018

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2018_november_chan_sophia.pdf [ 9.63MB ]
Metadata
JSON: 24-1.0372790.json
JSON-LD: 24-1.0372790-ld.json
RDF/XML (Pretty): 24-1.0372790-rdf.xml
RDF/JSON: 24-1.0372790-rdf.json
Turtle: 24-1.0372790-turtle.txt
N-Triples: 24-1.0372790-rdf-ntriples.txt
Original Record: 24-1.0372790-source.json
Full Text
24-1.0372790-fulltext.txt
Citation
24-1.0372790.ris

Full Text

  DESIGN AND FABRICATION OF ELECTROSPUN NANOFIBER CATALYST SUPPORT FOR POLYMER ELECTROLYTE MEMBRANE FUEL CELLS by  Sophia Shuk Kwan Chan  B.ASc., The University of British Columbia, 2011  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY  in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Materials Engineering)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  October 2018  © Sophia Shuk Kwan Chan, 2018    ii  The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoctoral Studies for acceptance, the dissertation entitled:  Design and Fabrication of Electrospun Nanofiber Catalyst Support For Polymer Electrolyte Membrane Fuel Cells  submitted by Sophia Shuk Kwan Chan   in partial fulfillment of the requirements for the degree of Doctor of Philosophy  in Materials Engineering   Examining Committee: Frank Ko, Materials Engineering Supervisor  Peyman Servati, Electrical and Computer Engineering  Supervisory Committee Member   Supervisory Committee Member David Wilkinson, Chemical and Biological Engineering University Examiner Tom Troczynski, Materials Engineering  University Examiner   Additional Supervisory Committee Members: Jasna Jankovic, Materials Science & Engineering Supervisory Committee Member Rizhi Wang, Materials Engineering Supervisory Committee Member iii  Abstract  Progress in fuel cell technology is impeded by the lack of understanding of the fundamental design characteristics necessary to improve the performance. Current catalyst design suffers from challenges related to platinum utilization, triple phase boundary, mass transport, and durability. Little improvement has been made in novel fuel cell catalyst designs, in particularly, there is a lack of microstructural optimization. As a result, a relationship between the catalyst layer microstructure and fuel cell performance has not been established. This study addresses this crucial problem by finding connections between controlled microstructure and key fuel cell performance factors. Specifically, electrospun nanofibers as a catalyst support offered a number of controllable structural parameters including: porosity, fiber diameter, fiber alignment, and layer thickness. The material and structural properties of carbonized nanofibers were optimized by factorial design for incorporation into a fuel cell membrane electrode assembly. Validation of the structural and material properties of the carbon nanofiber catalyst support was analyzed by electrochemical, physiochemical, and microscopy methods.   Carbon nanofiber were integrated into membrane electrode assemblies and tested in situ to develop structure-property-performance relationships pertaining to Pt loading, ionomer loading, substrate electrical properties, and fiber mesh geometry. Performance was characterized by cyclic voltammetry, polarization, and electrochemical impedance spectroscopy. Results confirm not only ionomer thickness and loading, but more importantly ionomer distribution influenced polarization losses. 150 µg cm-2 and 250 µg cm-2 Pt loading achieved the same maximum current density, suggesting reduced loading maintained satisfactory Pt utilization, decreasing the amount of Pt. Although the influence of fiber orientation on fuel cell performance was inconclusive, fiber iv  electrical conductivity, ionomer thickness, and Pt distribution were found to be essential for the development of efficient low cost catalyst layers. In conclusion, carbon nanofiber catalyst support revealed enhanced surface area, durability, Pt utilization, and efficiencies due to the porous mesh structure.  v  Lay Summary  Although fuel cells are considered an efficient means of converting chemical energy into renewable electrical energy, the technology faces challenges that prevent wide spread commercialization. The current Department of Energy goals propose a target cost of $40 kW/hour by 2020. Recent research progressed steadily in this direction through reducing material costs and improving efficiencies. The catalyst layer in a fuel cell system, which consists of expensive Pt metal supported on carbon powder, is one chief area for improvement. By designing a catalyst layer with controllable geometry and microstructure, the fuel cell performance can be correlated to the specific material design parameters. Using electrospun carbon nanofibers (CNF), the catalyst showed enhanced surface area, electrical conductivity, and porosity—leading to improved designs in the future. The CNF catalyst achieved comparable power density to published results using alternative carbonaceous materials while using less Pt, validating the potential of CNFs as a support material.  vi  Preface Unless otherwise stated below, the work presented in this thesis, including the literature review, research proposal preparation, experimental design and execution, data analysis and interpretation, and thesis preparation was completed by Sophia Shuk Kwan Chan, under the supervision of Dr. Frank Ko from the University of British Columbia and Dr. Jasna Jankovic from Automotive Fuel Cell Cooperation.     The publications arising from this work are listed:   A version of Chapter 4 has been published. Chan, S., Jankovic, J., Madhu, S., Susac, D., Tam, M., Yang, H., Ko, F., Electrospun carbon nanofiber catalyst layers for polymer electrolyte membrane fuel cells: Fabrication and optimization, Journal of Materials Science. 53(16) (2018) 11633-11647. The author conducted all the testing and was responsible for preparation of the publication and its content.   A version of Chapter 6 and 7 has been published. Chan, S., Jankovic, J., Madhu, S., Susac, D., Tam, M., Yang, H., Ko, F., Electrospun carbon nanofiber catalyst layers for polymer electrolyte membrane fuel cells: structure and performance, Journal of Power Sources. 392 (2018) 239-250. Dr. Madhu Saha and Dr. Jasna Jankovic operated the XRF for the Pt loading measurements. AFCC fuel cell operator was responsible for obtaining the in situ performance data (EIS and polarization data). S/TEM images/maps were acquired with the assistance of Dr. Jankovic. ALD Pt deposition was conducted by Dr. Philip Kubik at SFU 4D Labs. H2 furnace was operated by AFCC research staff. The author was responsible for all other testing and preparation of the publication and its content. vii   A remaining manuscript is in preparation regarding portions of Chapter 5 and 6: Chan, S., Susac, S., Jankovic, J., Madhu, S., Kremliakova, N., Yang, H., Ko, F., “Electrospun carbon nanofiber catalyst layers for polymer electrolyte membrane fuel cells: functionalization and surface chemistry analysis”. A portion of the orthogonally aligned nanofiber mats were fabricated by a volunteer student, Nerine Law, under the author’s direction. Dr. Darija Susac assisted the author with portions of the XPS data collection and interpretation. The author conducted all other measurements and is responsible for preparation of the publication and its content.  Research results from this thesis work have been presented by Sophia Chan at the following conferences:   Chan, S., Jankovic, J., Susac, D., Saha, M.S., Tam, M., Yang, H., Ko, F “Electrospun Carbon Nanofiber Based Catalyst Support for Proton Exchange Membrane Fuel Cell,” in: Pacific Centre for Advanced Materials and Microstructures, December 10, 2016, Vancouver, BC, Canada (poster)  Chan, S., Jankovic, J., Susac, D., Saha, M.S., Tam, M., Choi. J., Yang, H., Ko, F. 2016. “Electrospun carbon nanofiber based catalyst support for proton exchange membrane fuel cell”, in: SAMPE 2016, May 23 – 26 (Long Beach, California, USA), Society for the Advancement of Material and Process Engineering.  Chan, S., Jankovic, J., Susac, D., Saha, M.S., Tam, M., Yang, H., Ko, F “Geometrically controlled Nanofiber mesh,” AMPEL Synergy Day, May 29, 2015, Vancouver, BC, Canada  viii   Chan, S., Yip. G., Yang, H., & Ko, F. 2014. “Fabrication and characterization of electrospun poly(acrylonitrile) single nanofibers,” in: SAMPE 2014, June 2-5 (Seattle, Washington, USA), Society for the Advancement of Material and Process Engineering. ix  Table of Contents Abstract ......................................................................................................................................... iii Lay Summary .................................................................................................................................v Preface ........................................................................................................................................... vi Table of Contents ......................................................................................................................... ix List of Tables .............................................................................................................................. xiv List of Figures ............................................................................................................................ xvii List of Symbols ......................................................................................................................... xxvi List of Abbreviations ............................................................................................................. xxviii Acknowledgements .................................................................................................................. xxxi Dedication ............................................................................................................................... xxxiii  Introduction and Background ..............................................................................................1  Motivation ....................................................................................................................... 1  Fuel cells ......................................................................................................................... 6 1.2.1 Introduction and history of fuel cells .......................................................................... 6 1.2.2 Electrochemistry and basic principle .......................................................................... 8 1.2.3 Fuel cell design and components .............................................................................. 13 1.2.3.1 The fuel cell assembly ...................................................................................... 13 1.2.3.2 Gas diffusion layers .......................................................................................... 14 1.2.3.3 Catalyst layers ................................................................................................... 16 1.2.3.4 Proton conductive membrane and ionomers ..................................................... 18  Current catalyst layer materials .................................................................................... 19 1.3.1 Conventional carbon supports................................................................................... 19 x  1.3.2 Advanced material supports ...................................................................................... 20 1.3.3 Geometry of catalyst support .................................................................................... 23 1.3.4 Electrospinning in fuel cell applications ................................................................... 24  The electrospinning process .......................................................................................... 28  Carbon fiber fabrication ................................................................................................ 31  Research objectives ....................................................................................................... 34  Materials and Experimental Methods ...............................................................................38  Materials ....................................................................................................................... 38  Preparation of random and orthogonally aligned nanofibers ........................................ 38  Heat treatment ............................................................................................................... 40  Platinum deposition ...................................................................................................... 42  Ionomer deposition ....................................................................................................... 45  MEA assembly .............................................................................................................. 46  Characterization and analysis methods ......................................................................... 47 2.7.1 Thermal profile of PAN-co-MA ............................................................................... 47 2.7.2 SEM and TEM dimensional measurements .............................................................. 48 2.7.3 Mechanical testing .................................................................................................... 49 2.7.4 In-plane and through-plane electrical properties ...................................................... 49 2.7.5 Chemical characterization ......................................................................................... 52 2.7.6 Porosity measurement and calculations .................................................................... 52 2.7.7 X-Ray fluorescence ................................................................................................... 53 2.7.8 Ex situ cyclic voltammetry ........................................................................................ 53 2.7.9 In situ electrochemistry ............................................................................................. 54 xi  2.7.10 Electrochemical impedance spectroscopy ............................................................ 55  Optimization of Randomly Oriented CNF Layers ...........................................................56  Introduction ................................................................................................................... 56  Preliminary process screening ...................................................................................... 56  Two-level factorial design ............................................................................................ 62 3.3.1 Effect of process parameters on fiber diameter ........................................................ 64 3.3.2 Effect of process parameters on fiber electrical conductivity ................................... 66 3.3.3 Effect of process parameters on fiber mesh porosity ................................................ 68 3.3.4 Effect of process parameters on fiber young’s modulus ........................................... 70 3.3.5 Development and validation of factor models .......................................................... 72  Summary ....................................................................................................................... 78  Comparison Between RNL and OAN Properties .............................................................80  Introduction ................................................................................................................... 80  Structural and mechanical properties ............................................................................ 80  Through-plane and in-plane electrical properties ......................................................... 86  Surface carbon content via x-ray photoelectron spectroscopy ..................................... 87  Summary ....................................................................................................................... 91  Platinum and Ionomer Deposition on CNF .......................................................................94  Introduction ................................................................................................................... 94  Evaluation of Pt deposition process .............................................................................. 95 5.2.1 SEM/TEM evaluation of Pt nanowire coating .......................................................... 98 5.2.2 Chemical composition by XRD and XPS ............................................................... 100 5.2.2.1 Functionalization analysis ............................................................................... 101 xii  5.2.2.2 Chemical composition of platinized CNF....................................................... 105  Ex situ CV and effective platinum surface area .......................................................... 109 5.3.1 Electrochemical surface area of RNL ..................................................................... 110 5.3.2 Electrochemical surface area of OAN .................................................................... 112  Ionomer deposition and morphology .......................................................................... 113 5.4.1 Surface wettability of CNF catalyst layer ............................................................... 117  Summary ..................................................................................................................... 118  In situ Evaluation of CNF Based MEA ............................................................................120  Introduction ................................................................................................................. 120  In situ MAT cell performance (pre-screening of CNF) .............................................. 120  In situ CAT cell performance (full evaluation of CNF) ............................................. 132 6.3.1 Current and power density analysis ........................................................................ 132 6.3.2 Impedance and transport analysis ........................................................................... 137  Degradation post fuel cell test..................................................................................... 142 6.4.1 Fiber and Pt distribution evaluation ........................................................................ 142 6.4.2 Cross sectional evaluation....................................................................................... 146  Summary ..................................................................................................................... 148  Conclusions and Recommendations .................................................................................151  Development of nanofiber based catalyst layer .......................................................... 152  Research outcomes and contributions ......................................................................... 153 7.2.1 Effect of fiber orientation........................................................................................ 153 7.2.2 Effect of fiber electrical conductivity ..................................................................... 154 7.2.3 Effect of Pt deposition and loading ......................................................................... 155 xiii  7.2.4 Effect of ionomer distribution and loading ............................................................. 156  Health and safety concerns ......................................................................................... 157  Recommendations for future work ............................................................................. 159 Bibliography ...............................................................................................................................162 A Appendix .............................................................................................................................178 B Appendix .............................................................................................................................182 C Appendix .............................................................................................................................184 C.1 Raw data for factorial design experiments ....................................................................... 184 C.2 Regression modelling for process optimization ............................................................... 186 C.3 Response surface methodology principles ....................................................................... 188 D Appendix .............................................................................................................................195 D.1 Resolving oxygen transport resistance from limiting current data .................................. 195 D.2 Introduction to EIS theory................................................................................................ 196 D.3 Procedure for analyzing EIS data ..................................................................................... 201  xiv  List of Tables Table 1-1: Exchange current density (i0) for various metals in acid media [44] .......................... 13 Table 2-1: Electrospinning parameters for RNL and OAN .......................................................... 39 Table 2-2: Platinum plating bath chemical composition and concentrations ............................... 44 Table 2-3: Test station set-up and EIS testing protocols .............................................................. 55 Table 3-1: Electrospinning parameters used for spinning 12 wt% PAN-co-MA ......................... 59 Table 3-2: 12 wt% PAN-co-MA average fiber diameter after heat treatment process, fiber shrinkage of 30% .......................................................................................................................... 61 Table 3-3: Relationship between fiber in-plane electrical conductivity and carbonization temperature ................................................................................................................................... 62 Table 3-4: CNF design targets for MEA application based on recommendations from AFCC and industry standards ......................................................................................................................... 64 Table 3-5: Significance probability (P-value) and correlation coefficient of linear regression for response surface equations ............................................................................................................ 73 Table 3-6: Optimization targets and constraints for each response and factor ............................. 75 Table 3-7: Best predicted solution for optimized fabrication process .......................................... 75 Table 3-8: Material properties of CNF fabricated using predicted optimal parameters ............... 76 Table 4-1: Porosity measurement by densometry of random and aligned CNF after Platinum and Ionomer deposition ....................................................................................................................... 82 Table 4-2: Summary of OAN and RNL structural dimensions measured by SEM and densometry....................................................................................................................................................... 86 Table 4-3: Summary of electrical properties for OAN vs. RNL ................................................... 86 xv  Table 4-4: 1100 °C carbonized 10 wt% PAN-co-MA deconvoluted XPS peaks for C 1s, pass energy 20 eV ................................................................................................................................. 89 Table 4-5: 900 °C carbonized 10 wt% PAN-co-MA deconvoluted XPS peaks for C 1s, pass energy 20 eV ................................................................................................................................. 89 Table 5-1: Surface chemical composition denoted in atomic percentages ................................. 105 Table 5-2: ECSA for various RNL and OAN samples ............................................................... 112 Table 5-3: H2 cross over current and shorting resistance for RNL and OAN samples tested in CAT cell at 105% and 50% RH for 68 °C and 85 °C, respectively (*V50 was tested in 68 °C for both RH)...................................................................................................................................... 112 Table 5-4: Ionomer loading and I/C for CNF with similar Pt loading........................................ 115 Table 6-1: Operating conditions of single cell performance tests performed in MAT cell hardware ...................................................................................................................................... 122 Table 6-2: Summary of in situ MAT cell samples and their properties...................................... 123 Table 6-3: Summary of current densities at 825 and 100 mV for all samples measured in the MAT cell; 68 °C 105% RH and 85 °C 50% RH for NOC and HOT, respectively .................... 128 Table 6-4: Summary of CNF sample properties tested in CAT cell for polarization analysis ... 132 Table 6-5: Operating conditions of single cell performance tests performed in CAT cell hardware..................................................................................................................................................... 134 Table 6-6: The shorting resistance and the H2 cross-over currents measured with H2/N2 at 2.5 bara balanced inlet, 105% RH 68 °C and 50% RH 85 °C operating conditions; *measured at 68 °C 50% RH ................................................................................................................................. 137 Table 6-7: Total mass transport resistance and pressure independent transport resistance in CAT samples ........................................................................................................................................ 139 xvi  Table 6-8: Ohmic and ionic resistance measured by EIS for CAT samples under NOC and HOT operating conditions .................................................................................................................... 139 Table A-1: Summary of prior research related to CNF/CNT as catalyst supports in FC ........... 178 Table C-1: Summary of 4 factor 2 level factorial design............................................................ 184 Table C-2: ANOVA analysis for diameter, porosity, electrical conductivity, and young’s modulus regression models ......................................................................................................... 186 Table C-3: Response surface function predicted optimized solutions ranked by desirability .... 187  xvii  List of Figures Figure 1-1: Breakdown of the 2016 projected fuel cell stack cost at 500,000 units per year ......... 2 Figure 1-2: Fuel cell performance targets for year 2020 issued by U.S. Department of Energy [3]......................................................................................................................................................... 3 Figure 1-3: Example of basic electrochemical cell, current travels from anode to cathode ........... 8 Figure 1-4: Comparison of the theoretical reversible work as a function of operating temperature for heat engine and fuel cells [44] (reproduced with permission) ................................................ 10 Figure 1-5: Sources of overpotential and the resulting polarization curve [47] (reproduced with permission) .................................................................................................................................... 12 Figure 1-6: Schematic of membrane electrode assembly components [18] (reproduced with permission) .................................................................................................................................... 14 Figure 1-7: Cross section of MEA components [18] (reproduced with permission) .................... 15 Figure 1-8: Micrographs of commercial carbon fiber paper and carbon cloths (a: Toray Industries, b: SGL, and c: E-Tek) [18] (reproduced with permission) ......................................... 16 Figure 1-9: Schematic of agglomerate based microstructure and Triple Phase Boundary (TPB) 16 Figure 1-10: Chemical structure of Nafion (perfluorinated sulfonic acid) ................................... 18 Figure 1-11: Schematic illustration of electrospinning process .................................................... 29 Figure 1-12: Evolution of PAN chemical structure during the 3 steps of stabilization and carbonization (Top to bottom) [reproduced with permission] [120–122] .................................... 32 Figure 1-13: Diagram of experimental approach illustrating the progression of experiments; tasks were divided into three key experimental phases: Fabrication optimization, deposition of Pt and ionomer, and MEA in-situ testing ................................................................................................. 37 Figure 2-1: Conventional electrospinning apparatus .................................................................... 40 xviii  Figure 2-2: Modified electrospinning apparatus for orthogonally aligned fibers ......................... 40 Figure 2-3: Stainless steel frame for applying tension to as-spun mat during heat treatment ...... 41 Figure 2-4: Tube furnace supplied with Air/Argon gas for carbonization procedure (left); graphite sample holder plate for keeping CNF flat during heat treatment ................................... 41 Figure 2-5: Transition between (A) as-spun, (B) stabilized, and (C) carbonized fiber mat ......... 42 Figure 2-6: Plastic frame holding CNF, used during Pt deposition .............................................. 44 Figure 2-7: Ionomer deposition set-up .......................................................................................... 45 Figure 2-8: Exploded view of the MEA (CAT cell) components for hot pressing ...................... 46 Figure 2-9: Completed MEA for CAT cell testing ....................................................................... 46 Figure 2-10: MAT MEA sample: circular CNF sample hot pressed onto NRE-211 membrane and anode CL ....................................................................................................................................... 47 Figure 2-11: 2 cm gauge tensile test specimen ............................................................................. 49 Figure 2-12: Conventional multi-meter and specimen preparation with silver paint ................... 51 Figure 2-13: Through-plane electrical resistivity tool, measures resistivity as a function of pressure and thickness................................................................................................................... 51 Figure 2-14: CNF impregnated onto glassy carbon electrode for cyclic voltammetry (CNF geometric area was normalized for comparison) .......................................................................... 54 Figure 3-1: TGA spectra of PAN-co-MA (top) and pure PAN (bottom) in air (5 °C min-1) ....... 58 Figure 3-2: DSC curves on heating PAN-co-MA and homopolymer PAN in air at 1 °C min-1 [145] .............................................................................................................................................. 59 Figure 3-3: Fiber shrinkage after each stage of heat treatment (as-spun, stabilized, and carbonized) .................................................................................................................................... 60 xix  Figure 3-4: SEM micrograph of As-spun 12 wt% PAN-co-MA fiber mat (5 kV; 3Kx left, 10Kx right) .............................................................................................................................................. 61 Figure 3-5: SEM micrograph of stabilized 12 wt% PAN-co-MA fiber mat (5 kV; 3Kx left, 10Kx right) .............................................................................................................................................. 61 Figure 3-6: SEM micrograph of carbonized 12 wt% PAN-co-MA fiber mat (5 kV; 3Kx left, 10Kx right) .................................................................................................................................... 62 Figure 3-7: Factorial cube for fiber diameter recorded at 0 and 1 wt% CNT, red and blue circle indicates interaction ...................................................................................................................... 66 Figure 3-8: Design factor cube and interactions graph for electrical conductivity, interaction recorded at polymer concentration = 10 wt% and stabilization temperature = 230 °C ................ 68 Figure 3-9: Design factor cube recorded at 10 wt% polymer and interactions graph for % porosity; carbonization temperature setting: black square = 850 °C and red triangle = 1100 °C 70 Figure 3-10: Design factor cube recorded at 10 wt% polymer and interactions graph for elastic modulus; stabilization temperature setting: black square = 230 °C and red triangle = 260 ° ....... 72 Figure 3-11: Contour plot of predicted values for fiber diameter; optimal solution of 149 nm at carbonization temperature of 1100 °C and 0 wt% CNT ............................................................... 75 Figure 3-12: CNF morphology from optimized fabrication parameters (2 kV, secondary electron mode) ............................................................................................................................................ 76 Figure 4-1: RNL fiber diameter distribution using optimized fabrication process (CNF) ........... 81 Figure 4-2: OAN fiber diameter distribution using optimized fabrication process (CNF) ........... 82 Figure 4-3: Schematic of OAN fiber organization and through pores .......................................... 82 Figure 4-4: Orthogonally aligned carbon nanofiber mat fabricated by modified electrospinning method (SEM: 5 kV Secondary electron mode) ........................................................................... 84 xx  Figure 4-5: Orthogonally aligned (A and C) and random nanofiber (B and D) cross sectional thickness (15 kV, secondary electron mode) ................................................................................ 85 Figure 4-6: Ultimate tensile strength (left) and Young’s modulus (right) for OAN and RNL (average of 5 samples carbonized at 900 °C) ............................................................................... 86 Figure 4-7: Example of deconvoluted C 1s XPS spectrum for OAN (10 wt% PAN-co-MA carbonized at 900 °C), calibrated at 284.6 eV, pass energy: 20 eV .............................................. 90 Figure 4-8: OAN and RNL (carbonized at 900 °C) surface chemistry by deconvoluted C 1s XPS spectrum, pass energy 20 eV ........................................................................................................ 91 Figure 5-1: Pt deposition by various methods (5 and 10 kV, secondary electron mode) ............. 96 Figure 5-2: SEM micrograph of electroless plated Pt on CNF (top to bottom: low, medium, and high concentration Pt precursor) (1 kV, secondary electron mode) ............................................. 99 Figure 5-3: SEM image showing dense Pt nanowire clusters on CNF surface (120kx, 1 kV, secondary electron mode) ............................................................................................................. 99 Figure 5-4: TEM of Pt Nanowire crystal structure (left) and cross sectional STEM EDX map of Pt around each fiber (right) (STEM/EDX, 200 kV) ................................................................... 100 Figure 5-5: Comparison of nitrogen species before (left) and after (right) HNO3 chemical treatment for RNL carbonized at 700 °C, 900 °C, and 1100 °C ................................................. 102 Figure 5-6: Comparison of carbon species before (left) and after (right) HNO3 chemical treatment for RNL carbonized at 700 °C, 900 °C, and 1100 °C ................................................. 104 Figure 5-7: Comparison of nitrogen species before (left) and after (right) HNO3 chemical treatment for OAN carbonized at 900 °C ................................................................................... 104 Figure 5-8: Relationship between degree of graphitization and metallic Pt to oxidized Pt ratio 105 xxi  Figure 5-9: SEM images of CNF/Pt without HNO3 treatment (left) and with treatment (right); Pt surface area is higher after treatment .......................................................................................... 107 Figure 5-10: Nanowire growth on CNF surface with (left) limited anchoring sites and (right) many anchoring sites [176] ......................................................................................................... 107 Figure 5-11: Comparison of metallic to oxidized Pt ratio for RNL with and without HNO3 treatment ..................................................................................................................................... 108 Figure 5-12: X-ray diffraction diagram for Pt electroless plated CNF layers ............................ 108 Figure 5-13: EPSA was calculated based on the area under the Had oxidation curve ................ 109 Figure 5-14: Cyclic voltammogram of CNF/Pt catalysts with 3 different Pt loadings at 25 °C in N2-saturated 0.09M H2SO4, scan rate: 20 mV s−1 (1.25, 1.5, and 2.0 g L-1 precursor), normalized to geometric surface area of electrode ........................................................................................ 110 Figure 5-15: Relationship between Pt precursor concentration, ECSA and Pt loading.............. 111 Figure 5-16: Ionomer distribution between the pores and all around CNF/Pt surface ............... 115 Figure 5-17: TEM/EDX mapping of ionomer by fluorine signal (left), combined map showing Pt, carbon and ionomer (right) .................................................................................................... 116 Figure 5-18: TEM/EDX maps of combined Platinum (red) and Fluorine in ionomer (green) distribution (A-D) and only ionomer distribution (E-F): Low I/C sample (A and E), medium I/C with a part of NRE membrane shown (B and F), high I/C (C and G), and high I/C OAN revealing ionomer pockets (D and H). ........................................................................................................ 116 Figure 5-19: Ionomer thickness distribution for low, mid, high I/C on RNL (averaged over 200 readings) ...................................................................................................................................... 117 Figure 5-20: Wettability of catalyst layer before and after ionomer deposition; contact angle of 50° and 125° (A and B) ............................................................................................................... 118 xxii  Figure 5-21: Orientation of ionomer side chains towards hydrophilic Pt particles .................... 118 Figure 6-1: Regions of polarization losses in typical fuel cell. Voltages at which measurements were performed in this study are marked on the graph ............................................................... 123 Figure 6-2: High resolution STEM/EDX images of CNF/Pt and V50. (A) Electroless plating: nanowires showing (111) crystal planes (insert); (B) Cross section of CNF/Pt; (C) Pt sputter coated on fibers close to membrane; (D and E) STEM-EDX maps comparing carbon black V50 catalyst layer (0.25 mgPt cm-2), white rectangle selects a region visually comparable in morphology to the CNF region; (F) EDX map of Pt distribution (red) on fibers; Pt layer illustrates 0.25 mg cm-2 loading. Green color denotes Fluorine signal from ionomer, blue color denotes Carbon particles or fiber cross section, and red denotes Pt. .......................................... 124 Figure 6-3: Comparison between CNF samples in the ohmic loss region (85 °C, 50% RH, 600 mV and 68 °C, 105% RH, 650 mV conditions): high and low Pt loading; Pt deposition methods; low, medium, and high I/C; carbonization temperature; and orthogonal and random fiber mat.131 Figure 6-4: STEM/EDX mapping of OAN catalyst layer with Pt loading of 0.15 mg cm-2 (red: platinum, green: ionomer) ........................................................................................................... 132 Figure 6-5: Polarization characteristics of the MEAs fabricated with RNL and OAN cathode catalyst (0.15 and 0.25  mgPt cm−2) for H2/Air at 68 ◦C and 85 ◦C, Nafion 211 membrane, 2.5bara balanced anode and cathode inlet pressure. Dotted lines indicate iR-corrected data for comparison .................................................................................................................................. 135 Figure 6-6: Power density characteristics of the MEAs fabricated with RNL and OAN cathode catalyst (0.15 and 0.25  mgPt cm−2) for H2/Air at 68 ◦C and 85 ◦C, Nafion 211 membrane, 2.5 bara balanced anode and cathode inlet pressure ......................................................................... 136 xxiii  Figure 6-7: Nyquist plot of CAT samples in NOC conditions under H2/N2 conditions (right: zoom in of x-intercepts) .............................................................................................................. 139 Figure 6-8: Equivalent Circuit used to fit Nyquist plot for CAT samples .................................. 140 Figure 6-9: |Z| as a function of ω−1/2 for the cell in NOC conditions, slope represents a relationship for Rionic ................................................................................................................... 140 Figure 6-10: Before and after fuel cell testing; RNL with Pt loading of 167 µg cm-2 (left before, right after) (2 kV, secondary electron mode) .............................................................................. 144 Figure 6-11: Before and after fuel cell testing; RNL with Pt loading of 148 µg cm-2 (left before, right after) (2 kV, secondary electron mode) .............................................................................. 145 Figure 6-12: Before and after fuel cell testing; R150M with Pt loading of 157 µg cm-2 (left before, right after) (2 kV, secondary electron mode) .................................................................. 145 Figure 6-13: Distribution of Pt particle cluster size measured in the long axis post fuel cell testing..................................................................................................................................................... 145 Figure 6-14: Cross sectional evaluation post fuel cell testing: Pt shown on fibers (15 kV, secondary electron mode) ........................................................................................................... 146 Figure 6-15: Cross sectional evaluation post fuel cell testing: R150H cross section, Pt shown on fibers (15 kV, secondary electron mode) .................................................................................... 146 Figure 6-16: O150H OAN cross section post MAT FC testing (15 kV, secondary electron mode)..................................................................................................................................................... 146 Figure 6-17: Post fuel cell testing with sufficient Pt distribution and adhesion (left OAN O150H, right RNL R150M) (STEM EDX, 200 kV) ................................................................................ 147 Figure 6-18: Cross section of R150H (left) and R250-T1 (right) showing regions of delamination (15 kV, secondary electron mode) .............................................................................................. 147 xxiv  Figure 6-19: Cross sectional evaluation post fuel cell testing: O150H OAN (left) R150H RNL (right) (15 kV, secondary electron mode) ................................................................................... 147 Figure C-1: Histogram of average fiber diameter of Sample #10 (nm) ...................................... 185 Figure C-2: Fiber mat factorial design average fiber diameter (nm) .......................................... 185 Figure C-3: Fiber mat factorial design in-plane electrical conductivity (S cm-1) ....................... 185 Figure C-4: Fiber mat factorial design % porosity ..................................................................... 185 Figure C-5: Fiber mat factorial design results for elastic modulus (GPa) .................................. 185 Figure C-6: Fiber mat factorial design through-plane electrical resistance (mΩ∙cm2) ............... 185 Figure C-7: Example of half-normal plot in Design-Expert, here C and D are main effects as they fall far from the linear line .......................................................................................................... 189 Figure C-8: Pareto graph displaying the magnitude of the chosen effects, effects below the lower limit are insignificant .................................................................................................................. 190 Figure C-9: Residuals vs. run number; diagnostic method to validate the model ...................... 191 Figure C-10: Example contour plot for fiber diameter, factor B and C are fixed at 230 °C and 1100 °C, respectively .................................................................................................................. 193 Figure D-1: Limiting current as a function of dry mole fraction of oxygen for four different total pressures in a cell ........................................................................................................................ 195 Figure D-2: Dry transport resistance as a function of total pressure in a cell. The line represents the best linear fit to the four data points ..................................................................................... 196 Figure D-3: Current versus Voltage curve showing pseudo-linearity[226]................................ 197 Figure D-4: Presentation of an AC sinusoidal potential signal (E) with a resulting current response (I) [226] ........................................................................................................................ 197 Figure D-5: Nyquist Plot with Impedance Vector [226] ............................................................ 199 xxv  Figure D-6: Example of a Bode plot showing the relationship between log impedance, phase . 200 Figure D-7: Nyquist plots for (a) a capacitor, (b) a capacitor in series with a resistor, (c) a capacitor in parallel with a resistor, and (d) a resistor in series with a parallel RC-circuit (Reproduced from [227] with permission of The Royal Society of Chemistry) ..................................................................... 200 Figure D-8: Example of EIS data visualization in the ZView software, complex plot and bode plot are used to resolve impedances in the fuel cell components ............................................... 201 Figure D-9: Screenshot of model fitting using Randles circuit in ZView .................................. 202 Figure D-10: Example plot of |Z| (magnitude) as the function of ω−1/2 (frequency) for the cell; the slope can be used to determine Rionic .......................................................................................... 202 xxvi  List of Symbols  Greek symbols Symbol  Description Unit ε  Dielectric constant  ηCarnot  Efficiency of Carnot engine % ηElectrochem,max  Efficiency of electrochemical cell % ηs,a  Anodic activation overpotential V ηs,c  Cathodic activation overpotential V ηd,a  Anodic concentration overpotential V ηd,c,  Cathodic concentration overpotential V σ  Electrical conductivity S cm-1 R  Electrical resistance Ω l  Length between the electrical contacts cm A  Cross sectional area of the specimen cm2 ρ  Electrical resistivity Ωcm η  Intrinsic viscosity dL g-1 γ  Surface tension N m-1 ρc  Density of bulk carbon fiber g cm-2 Roman symbols Wout  Work output J Qin  Heat input J Q  Flow rate m3 s-1 QM  Charge associated with the formation of a monolayer mC cm-2 TL  System’s low temperature °K TH  System’s high temperature °K Ea  Electrical potential at the anode V Ec  Electrical potential at the cathode V E°cell  Equilibrium cell potential at standard conditions V Ecell  Equilibrium cell potential, non-standard conditions V xxvii  EOC  Open circuit voltage (OCV) V F  Faradays constant C mol−1 n  Number of moles of electrons  ΔGrxn  Change in Gibbs free energy J mol-1 ΔHrxn  Change in Enthalpy J mol-1 R  Universal gas constant J mol-1 K-1 pH₂  Partial pressure of hydrogen gas Pa pO₂  Partial pressure of oxygen gas Pa ΔVohm  Ohmic overpotential V i  Total current density A cm-2 ilim  Limiting current density A cm-2 i0  Exchange current density A cm-2 d  Fiber diameter nm I  Current carried by the nanofiber A X  Ratio of the initial jet length to the nozzle diameter  Vtot  Total bulk volume including pore volume m3 Vsolid  Volume of solids, minus pore volume m3 mCNF  Mass of CNF fibers g LCNF  Geometrical length of CNF mat cm WCNF  Geometrical width of CNF mat cm τCNF  Geometrical thickness of CNF mat cm Mdry  Mass of dry fiber mat g Mwet  Mass of wet fiber after submersion g Msub  Mass of fiber during submersion g APt  Electrochemically active surface area m2 gPt-1 RPI  Pressure independent resistance s m-1 RMT  Total mass transport resistance s m-1 Rohm  Total ohmic resistance of the MEA mΩ·cm2 Rionic  Proton resistance of the catalyst layer mΩ·cm2 Cd  Double layer capacitance F xxviii  List of Abbreviations AFCC = Automotive Fuel Cell Cooperation ALD = Atomic layer deposition CAT = In-house test cell with larger active area (5 cm2) CB = Carbon Black  CCM = Catalyst coated membrane CF = Carbon fiber CL = Catalyst layer  CNF = Carbon nanofiber  CNT = Carbon nanotube CV = Cyclic voltammetry  CVD = Chemical vapour deposition  DoE = Department of Energy  DOE = Design of Experiments  DMF = N,N’-Dimethylformamide ECSA = Electrochemically active surface area  EPSA = Effective platinum surface area  EDX = Energy-dispersive X-ray spectroscopy EIS = Electrochemical impedance spectroscopy FC = Fuel cells  FFP = Flow field plate  GDL = Gas diffusion layer HOT = Cell temperature is 85 °C and the inlet RH’s are at 50% xxix  HOR = Hydrogen oxidation reaction  I/C = Ratio between ionomer and carbon mass iLim = Limiting current density  MAT = In-house test cell with small active area (1.13 cm2) for performance screening MEA = Membrane electrode assembly  MPL = Microporous layer  MWCNT = Multiwall carbon nanotube NOC = Normal operating conditions cell temperature is 68 °C and the inlet RH’s are at 105% OAN = Orthogonally aligned nanofiber layer  OCV = Open circuit voltage ORR = Oxygen reduction reaction  PAN = Poly(acrylonitrile) PAN-Co-MA = Poly(acrylonitrile-co-methyl acrylate) PEMFC = Polymer electrolyte (Proton exchange) membrane fuel cell  PTFE = Polytetrafluoroethylene PVP = Polyvinylpyrrolidone  RH = Relative humidity  RHE = Reversible hydrogen electrode  RNL = Random nanofiber layer RSM = Response surface method  SEM = Scanning electron microscopy SHE = Standard Hydrogen Electrode STP = Standard temperature and pressure xxx  TEM = Transmission electron microscopy TGA = Thermogravimetric analysis TPB = Triple phase boundary  XPS = X-ray photoelectron spectroscopy XRD = X-ray powder diffraction XRF = X-ray fluorescence     xxxi  Acknowledgements First and foremost, I would like to express my sincere gratitude to my supervisor Prof. Frank Ko, for granting me the opportunity to pursue my PhD and for his continuous support and belief in me. I am thankful for the valuable guidance, advice, patience, and commitment Prof. Ko has given throughout my academic journey. Through our casual discussions together, Prof. Ko has instilled inspirational and wise words, nurturing my personal and professional growth.  Additionally, I am sincerely appreciative of the research team at Automotive Fuel Cell Cooperation (AFCC). In particular, I am thankful for the collaborative research initiated by Dr. Jasna Jankovic. I offer sincere gratitude to Dr. Jankovic for her continuous encouragement, suggestions, mentorship, and guidance. As my industry supervisor, Dr. Jankovic taught me invaluable knowledge about fuel cell technology and research. I would also like to express my sincere gratitude to Dr. Madhu Sudan Saha, Dr. Darija Susac, Dr. Natalia Kremliakova, and Mr. Mickey Tam. I could not have accomplished my research without their technical assistance and training. Special thanks to the AFCC team for sharing the facilities, equipment, and knowledge with me.        I would also like to thank my thesis advisory committee members Prof. Rizhi Wang and Prof. Peyman Servati for committing their valuable time and technical input related to my thesis work. Also, many thanks to the technical staff at UBC and SFU 4D Labs for training and sharing their experience and insight – Jacob Kabel, Gethin Owen, Philip Kubik, Nathanael Sieb, Saeid Soltanian, Xin Zhang, and Brad Ford.     Financial support from the Natural Sciences and Engineering Research Council of Canada (NSERC) and MITACS is gratefully acknowledged. I would also like to thank AFCC for access to the labs and the financial support that have been provided for this project. The University xxxii  of British Columbia should also be thanked for the generous financial awards and the outstanding support staff in the Materials Engineering Department.  To my supportive friends in the Advanced Fibrous Materials Laboratory, I am thankful for the positive energy and conversations we had. Thanks for offering help and encouragement whenever situations became seemingly hopeless and difficult. Thanks for the great memories, best wishes to everyone in the future. I am grateful for my volunteer student, Nerine Law, who was a helping hand in fabricating orthogonal samples. Special thanks to Dr. Heejae Yang, Dr. Jungsu Choi, Dr. Yuqin Wan for the valuable technical input and maintaining the lab tools. Many thanks to my brilliant friends who have contributed to the completion of this thesis, whether it is proof reading or producing illustrations.   Finally, I am grateful for my supportive parents and family members that help me in their own unique ways. Without their unconditional love, patience, and reassurance, I may not have the fortune to attend university and graduate school. I am truly blessed to have this remarkable family. And thanks to my loving partner for understanding my academic obligations and genuinely supporting my career goals.      xxxiii  Dedication      To my parents, Lisa and Peter, whose sacrifices have given me the opportunity to lead a fulfilling life. Their hardships as immigrants inspire me to always chase my dreams.   To my siblings, whom I never have a dull moment with.        1   Introduction and Background  Motivation  One of the most well-known Nobel laureates of our time, Richard Smalley, predicted ten of humanity’s most pressing problems for the next 50 years. Of the issues he listed, energy and environment placed first and fourth, respectively. Presently, 85% of world energy consumption comes from fossil fuels, a non-renewable source [1]. Because of the worldwide population and economic growth, demand for fossil fuels is rapidly increasing. Not only is the increasing demand for energy depleting oil reserves, fossil fuels emit large amounts of greenhouse gases into the environment, affecting global climate patterns. Of the total U.S. emissions in 2016 (6,511 Million Metric tons), transportation and electricity account for 28% each [2]. For this reason, there is a global effort in commercializing low carbon and renewable energy technologies such as: biomass, solar cells, batteries, nuclear, and fuel cells (FC). Of particular interest are fuel cells for large scale applications, such as transport vehicles, back-up generators, or stationary power generators [3]. In fact, commercialization of fuel cells is already underway. Globally, several countries have demonstration projects utilizing fuel cells in space vehicles, city buses, cars, and submarines. With global fuel cell investments totaling 355 million dollars from 2014 – 2016, fuel cells are gradually becoming an attractive option for green energy. Specifically, Polymer Electrolyte Membrane Fuel Cells (PEMFC) have enormous potential in reducing CO2 emissions from portable and transportation applications. The low operating temperature and compact design make this type of fuel cell more attractive, gaining earnest attention in research and academia. In addition to eliminating toxic emissions, fuel cell technology is two to three times more efficient than internal combustion engines because electricity is produced via electrochemical reactions. PEMFC also 2  offer the advantage of minimal maintenance since there are no moving parts in the fuel cell stack compared to combustion engines. Moreover, unlike battery powered vehicles which require several hours to fully charge, the fuelling process for fuel cells is similar to that of gasoline engines [4]. Although PEMFC has numerous advantages it remains challenging to commercialize. In order to reach wide spread commercialization, substantial research progress has been made in developing inexpensive catalysts, porous structures, hydrogen storage solutions, and hydrogen extraction methods. Moreover, significant effort has been made in building the infrastructure to accommodate fuel cell vehicles. The infrastructure and cost challenges are thoroughly discussed in the literature and demonstrates the complexity of scaling-up [5–9]. While there are several areas for research and development, many studies aim to reduce the cost of PEMFC by targeting the most expensive component – the catalyst layer (CL) which consists of precious metals [10–16]. A breakdown of the total manufacturing cost of the fuel cell components reveal that the catalyst layer accounts for 43% of the overall cost (Figure 1-1) [17]. Figure 1-2 illustrates the current status of the Department of Energy (DoE) fuel cell goals. As shown in Figure 1-2, the fuel cell cost and durability are the two targets the industry has yet to reach [3]. To meet the goals of $40/kW and 5000 operational hours by 2020, a more collaborative effort to develop the membrane electrode assembly (MEA), particularly, the catalyst layer is paramount.  Figure 1-1: Breakdown of the 2016 projected fuel cell stack cost at 500,000 units per year  3   Figure 1-2: Fuel cell performance targets for year 2020 issued by U.S. Department of Energy [3] a Ratio of direct current output energy to lower heating value of hydrogen  b Target includes fuel cell stack, BOP, and thermal system, and excludes hydrogen storage, battery, electric drive, and power electronics  c Based on Pt cost of $1,550/troy ounce and cost projected to high-volume production (500,000 fuel cell stacks per year) The catalyst layer’s main function include: catalysis of the anode and cathode reactions, transport of fuels to reactant sites, conduction of protons and electrons, and water and heat removal [18]. In a PEMFC, the cathode oxygen reduction reaction (ORR) (½O2 + 2H+ + 2e− → H2O) is torpid under the typical operating conditions, thus Pt catalyst is necessary to improve the kinetics of the cathode reaction. Much of the current research is related to developing new catalyst materials or metal alloys [10,13,19,20]. However, in addition to the cost of catalyst, there are several issues with the current catalyst layer including: catalyst activity, corrosion, lack of triple phase boundary, durability, and water removal. Research has shown that these challenges could be resolved by developing novel catalyst support structures that exhibit higher surface area, porosity, and conductivity [14,16,21–24].    4  Efficiency, performance, and durability of a PEMFC MEA depend largely on the microstructure and composition of the anode and cathode catalyst layers. A conventional PEMFC electrode catalyst layer comprises of catalyst Pt nanoparticles dispersed on an electrically conductive carbon black (CB) support arranged in connected agglomerated domains [25]. The porous CB structure, coated with a layer of ionomer (such as Nafion), provides a percolation path for electron and proton transport. The CB microstructure is irregular and often non-homogeneous, with highly tortuous transport paths for electrons, protons, and gas. Because of the inherent properties of CB, conductivity and gas diffusivity are insufficient, and portions of the Pt catalyst are buried within the agglomerate which becomes inaccessible for the reaction. The triple-phase boundary (TPB) is often not continuous and unavailable because of the agglomeration due to corrosion under operating conditions, reducing catalyst activity [26–28]. The concept of TPB is widely investigated in literature and states that ORR can only occur when proton conductive electrolyte, gas, and electrically connected catalyst are in contact with one another [29,30]. At high current densities cathode flooding (blocked pores) is an issue because product water does not quickly diffuse out of the MEA and causes reduction of gas diffusion. Furthermore, fabrication of these catalyst layers is not a fully controllable process; hence the properties of the layers and fuel cell performance cannot be easily refined. Due to the irregularity of the catalyst layer and inability to control the manufacturing process and layer parameters, there is a lack of understanding of the complex electrode structure and the factors that influence the catalyst activity and durability. Due to the disadvantages of carbon blacks, several studies focused on developing advanced materials with high surface area as the catalyst support material [31]. Recent research revealed the advantage of organized catalyst support structures and suggests that the geometry of macropores improved water management and mass transport [24,32,33]. Additionally, several types of carbon based 5  catalyst supports have shown promising results, including CNT and CNF [31,34]. Although studies have demonstrated performance improvement using novel catalyst structures, systematic research on the effect of geometry and organized microstructure on key fuel cell performance factors has been scarce. As a result correlations between microstructure and performance have not been established.  Realizing the challenges associated with the current catalyst supports, this thesis addresses the research opportunity regarding catalyst supports in the scope of material processing and fabrication. The development of highly controlled structures, coupled with full characterization of the material and structural properties, enables understanding of the relationships between fuel cell catalyst layer microstructure and its performance, knowledge crucial for further fuel cell technology development. Investigating the design of the catalyst support layer microstructure allows for fine tuning of key fuel cell performance factors, increasing fuel cell efficiency and reducing costs. This will lead to more compact devices, less materials usage, and more efficient manufacturing of MEA, hence lowering the cost of industrial production.   This thesis research aims to develop CNF based catalyst layers with optimized material and structural properties (porosity, electrical conductivity, fiber diameter, and mechanical strength) to investigate key fuel cell design parameters, such as support geometry, Platinum (Pt) loading, ionomer loading and distribution, Pt deposition methods, and substrate electrical conductivity. The study involves developing methods and techniques to optimize the fiber properties, Pt deposition, and ionomer morphology. This will enable understanding of the catalyst support layer and effect of various design parameters on fuel cell properties and performance. Although prior work has shown promising results using alternative carbonaceous materials in the catalyst, few have discussed the relationship between key fabrication parameters, material properties, and fuel cell 6  performance [22,35–41]. The current goal of the study is to compare the effect of orthogonally aligned and randomly oriented fibers on fuel cell polarization losses. Gaining a fundamental understanding of the catalyst structure could potentially improve Pt utilization, durability, water management, and mass transport, further improving fuel cell performance. Research in this unexplored area may improve current fuel cell designs, addressing DoE economic goals which lead to commercialization.   Fuel cells    1.2.1 Introduction and history of fuel cells Since the discovery of fuel cells in the early 1800s, several types of fuel cells are currently in different stages of development. Fuel cells are often categorized by the type of electrolyte used. The types of fuel cells include: alkaline fuel cell (AFC), direct methanol fuel cell (DMFC), solid oxide fuel cell (SOFC), molten carbonate fuel cell (MCFC), phosphoric acid fuel cell (PAFC), and polymer electrolyte membrane fuel cell (PEMFC) [3]. PEMFC is considered a promising candidate for next generation automobiles due to its high power density and lower operating temperature. PEMFC, the focus of this research, is a zero emission energy device that can reduce greenhouse gas emissions if commercialized.  The first demonstration of fuel cells was in 1839 by Sir William Robert Grove. However, the modern day fuel cell concept was not developed until the 1950s and 1960s. The first ever PEMFC was used in NASAs 1965 Gemini space flights as an auxiliary power source [42]. During the NASA Gemini program, the polymeric membrane was a type of polystyrene with low durability and conductivity. General Electric also used fuel cells to provide drinking water for the crew during space missions. In 1968, DuPont developed proton conducting electrolyte called Nafion and tested the first Nafion based PEMFC in 1969 [43]. General Electric developed PEMFC 7  for submarines used in the U.S. Navy from 1974 – 1976 using the same ion exchange membrane developed by DuPont for the Gemini 5 mission [42]. Siemens also developed PEMFC for submarines which were used by the German Navy in 2002 [42]. In 1993, Ballard presented the first fuel cell powered bus and in 1997 Daimler-Benz and Toyota launched prototype fuel cell cars [42,44]. Ballard deployed 30 fuel cell buses in 10 European cities [45]. In 2017, Ballard becomes the first fuel cell company to power buses for more than 10 million cumulative kilometers of revenue service [45]. Several fuel cells were later developed for backup power, stationary power generators, and fork lifts [42]. Since then, an increasing number of research has focused on hydrogen storage and extraction, fuel infrastructure, and improving fuel cell vehicles in collaboration with auto manufacturers.  According to the DoE, 62,000 fuel cells were shipped worldwide in 2016 [46]. There are three main markets for fuel cell technology: stationary power, transportation, portable power. The impressive 65% growth from 2014 to 2015 was largely attributed to a growth in the stationary and transportation sector. In the stationary sector, utilities in Korea and the U.S. are installing large-scale fuel cell power parks, contributing to a large portion of the 20% growth in fuel cell shipments. The gradual growth in light duty vehicles in California, Europe, and Japan also account for the increase in shipments in the transportation sector. Electric vehicle sales from Hyundai and Toyota, along with stationary fuel cells from Korea and Japan, increased Asian shipments by double from 2014 to 2015 [42]. In 2016, China emerged as a leading customer for fuel cell buses, with more than 30 deployed [46]. Furthermore, in addition to government research support, private investors globally have invested $355.8 million into fuel cell companies between 2014 – 2016 [46]. As the potential for fuel cells is realized, governments will continue to support programs and create new policies and initiatives to encourage global development and commercialization of fuel cells. 8  Improving fuel cell technology could have a significant impact on the society and alleviate the dependency on oil, creating a more sustainable future.  1.2.2 Electrochemistry and basic principle In order to understand how the fuel cell and its components operate, a brief discussion about thermodynamics and electrochemistry is presented. There are two main pathways of converting chemical energy into useful work – heat engines and electrochemical devices. Heat engines typically burn gasoline, coal, or oil, converting heat energy into mechanical work. Based on the second law of thermodynamics, the efficiency of a Carnot engine is defined by:     𝜂𝐶𝑎𝑟𝑛𝑜𝑡 =  𝑊𝑜𝑢𝑡𝑄𝑖𝑛= 1 −  𝑇𝐿𝑇𝐻 (1.1)  Where ηcarnot is Carnot efficiency, Wout is work output, Qin is heat input, TL is the system’s absolute low temperature, and TH is the heat engines absolute high temperature. This means, to achieve maximum efficiency, TH should be as high as possible, restricted by flame temperature and material properties (Figure 1-4). Carnot engine could never reach 100% efficiency because of the presence of wasted heat.     Figure 1-3: Example of basic electrochemical cell, current travels from anode to cathode 9  In comparison, electrochemical devices like batteries or fuel cells directly convert chemical energy into electrical energy, avoiding conversion losses present in heat engines (Figure 1-3). An electrochemical cell, such as a battery, consists of two half-cells. As shown in Figure 1-3, each half-cell consists of an electrolyte and electrode. A spontaneous redox reaction, where one half-cell loses electrons (oxidation) and one half-cell gains electrons (reduction) from their electrode, generates an electric current which is carried by an external circuit. This constant flow of electrons from the electrode produces a direct current which can be used to do work. One way to define electrochemical efficiency is through maximum thermodynamic efficiency, given by:  𝜂𝑒𝑙𝑒𝑐𝑡𝑟𝑜𝑐ℎ𝑒𝑚,𝑚𝑎𝑥 =ΔG𝑟𝑥𝑛 ΔH𝑟𝑥𝑛= −𝑛𝐹𝐸𝑐𝑒𝑙𝑙ΔH𝑟𝑥𝑛× 100% (1.2) Where ΔHrxn (J mol-1) is the change in enthalpy, ΔGrxn (J mol -1) is the change in Gibbs free energy, n is the number of moles of electrons, F is Faraday’s constant, and Ecell is the electrical potential. Actual efficiency is calculated based on operating cell voltage, rather than ideal equilibrium cell voltage. Fuel cells are often compared to efficiencies of Carnot heat engines. Below is an illustration comparing the two device’s theoretical efficiency with regards to operating temperature (Figure 1-4). The graph clearly shows the advantages of fuel cells at lower operating temperatures compared to heat engines. Although the graph represents theoretical values, the difference in energy output at low temperatures is quite substantial. This means that fuel cells can operate at a much lower temperature while maintaining high efficiencies, ideal for automotive applications.    10   Figure 1-4: Comparison of the theoretical reversible work as a function of operating temperature for heat engine and fuel cells [44] (reproduced with permission)  E°cell = E°c - E°a (1.3) 2H2  4H+ + 4e- E°= 0 V vs. SHE (1.4) O2 + 4H+ + 4e-  2H2O E°= 1.23 V vs. SHE (1.5) O2 + 2H2 2H2O E°= 1.23 V (1.6) 𝐸𝑐𝑒𝑙𝑙 =  𝐸𝑐𝑒𝑙𝑙° −  𝑅𝑇𝑛𝐹𝑙𝑛 (1𝑝𝐻2𝑝𝑂21/2) Nernst Equation (1.7) In the case of standard temperature and pressure (STP), theoretical cell potential is given as the difference between the standard half-cell potential of the cathode and anode (1.3). Thus, given the half-cell potentials for the H2/O2 redox reaction, E°cell of a PEMFC is 1.23 V (1.6). However when the cell is operating at non-standard conditions, the equilibrium cell potential is calculated from the Nernst equation. The Nernst equation can be expressed in several ways, a general expression for a H2/O2 fuel cell is shown in (1.7). This expression gives the equilibrium cell potential when no current is flowing, or open circuit voltage (OCV). However, in reality, the OCV is much less than theoretical values. In addition, the voltage falls when more current is drawn from the fuel cell. A typical performance diagram for fuel cells is shown in Figure 1-5. The shape of the 11  graph can be explained by four major irreversibilities (overpotential): (1) activation losses, (2) fuel crossover and internal currents, (3) ohmic losses, and (4) mass transport or concentration losses. Activation loss is caused by the slow kinetics of the reaction at the electrode surface. The transfer of electrons to or from the electrode uses a portion of the generated voltage. Fuel crossover is energy loss in fuel passing through the electrolyte, affecting the OCV of low-temperature cells. Ohmic losses occur because of various resistances associated with interconnects and materials inside the MEA, as well as resistance to the flow of ions through the electrolyte. The voltage drop is linear, proportional to current density. Concentration loss occurs when insufficient fuel is transported to the electrode surface. Because mass transport is responsible for the reduction of fuel concentration at the electrode surface, it is also termed mass transport loss. Therefore, the actual fuel cell voltage can be expressed by subtracting all the irreversibilities: 𝑬 = 𝑬𝑶𝑪 − |𝜼𝒔,𝒂| − |𝜼𝒔,𝒄| − |𝜼𝒅,𝒂| − |𝜼𝒅,𝒄| − 𝚫𝑽𝒐𝒉𝒎 (1.8)  Where E is actual cell voltage, EOC is open circuit voltage (OCV), ηs,a is anodic activation (surface) overpotential, ηs,c is cathodic activation overpotential, ηd,a is anodic concentration (diffusion) overpotential, ηd,c is cathodic concentration overpotential, and ΔVohm = iR is the ohmic overpotential (i is current density, R is area-specific resistance). The mathematical expressions for each overpotential has been defined and detailed in literature and cited text (appendix) [18,44].  12   Figure 1-5: Sources of overpotential and the resulting polarization curve [47] (reproduced with permission) Considering the sources of overpotential, there are a number of ways to improve the performance or decrease the losses. Since activation loss is governed by the i0 exchange current density (i.e. the constant forwards and backwards flow of electrons to and from electrode at equilibrium). The current density is important to fuel cell performance and it is crucial to make this current density as high as possible. Some suggested ways to improve exchange current density include: using effective catalyst, increasing electrode surface roughness, increasing reactant concentration,  and increasing input gas pressure [44]. The advantage of using an effective catalyst is clearly shown in Table 1-1, where Pt and Pd display the highest i0. Generally, Pt is more suitable in acidic media and Pd in alkaline media due to instability and corrosion [15,19]. Additionally, increasing roughness of the electrode increases the available surface area per geometric area, thus allowing available surface for the catalyzed reaction to occur. Similarly, methods to decrease ohmic resistance within the stack include employing: electrodes with higher conductivity, thinner electrolytes, and appropriate materials and design. To decrease concentration losses, fuel cell design should be wary of partial pressure drops due to the consumption of the fuel at the surface 13  and electrode flooding. Insufficient water removal can plug the pores and slow down reactant gas diffusion [44].      Table 1-1: Exchange current density (i0) for various metals in acid media [44] Metal i0 (Acm-2) Pb 2.5 x 10-13 Zn 3 x 10-11 Ag 4 x 10-7 Ni 6 x 10-6 Pt 5 x 10-4 Pd 4 x 10-3    1.2.3 Fuel cell design and components 1.2.3.1 The fuel cell assembly  A typical fuel cell stack contains a crucial arrangement called the Membrane Electrode Assembly (MEA).  This is the core of the fuel cell, where redox reactions take place to generate electricity. The basic components of the MEA include: gas diffusion layers with microporous layer, cathode and anode catalyst layers, and a proton conductive electrolyte membrane (Figure 1-6).  Typically, the anode or cathode will be no more than 10 µm thick while the ionomer membrane ranges from 10 - 25 µm [48].  The redox reactions occur at the electrodes and they depend on the type of the fuel cell. In a PEMFC, the hydrogen oxidation reaction (HOR) occurs at the anode, releasing electrons and protons (H+), as shown in (1.4). Protons are conducted through a proton conductive electrolyte membrane to the cathode, where the oxygen reduction reaction (ORR) occurs (1.5). The overall reaction is given by (1.6). Electrons released at the anode flow to the cathode through an external circuit to generate direct current.   14    Figure 1-6: Schematic of membrane electrode assembly components [18] (reproduced with permission) The MEA assembly is about a few hundred microns in thickness and can generate power densities up to about 1 W cm-2 of electrode area when supplied with air and hydrogen [18].  As mentioned above, the PEMFC operates at a comparatively lower temperature than other fuel cells because it is limited by the necessity to humidify the electrolyte to effectively transport protons. The following sections describe the structure and function of each component in the MEA.    1.2.3.2 Gas diffusion layers  In a fuel cell, membrane and the two catalyst layers are sandwiched between two gas diffusion layers (GDL), usually ~200 µm thick. Beside mechanical support, GDL provides an electrically conductive pathway for current collection between the flow field plate (FFP) and the electrode and enables reactant gas transport to the electrodes and liquid waste water removal. Figure 1-7 shows the supply of reactant gas through the gas channels on the FFP.  The gas diffuses through the GDL uniformly, contacting the entire catalyzed surface area. The reaction is initiated by the catalyst and the electrons are released generating a current (1.4). Only the landing areas can conduct electric current, therefore the GDL transports the current laterally to the contact points [18]. It is typically H2 in  15  fabricated using carbon cloth or paper for its porous properties (Figure 1-8) [18]. The GDL has a number of functions and roles in the MEA: (i) it must be both thermally and electrically conductive through-plane and in-plane to transport electrons to the FFP, (ii) it must be sufficiently porous to allow waste water removal and transport of reactant gas, and (iii) it must provide structural support while maintaining ample flexibility to ensure adequate electrical contacts. Since the flux through the GDL is in both directions, it is important to ensure that the small carbon particles from the conventional catalyst layer do not migrate through; thus small pores (20 – 300 nm) are required to eliminate this concern [49,50]. This is usually addressed by the use of a Micro Porous Layer (MPL), a non-catalytic layer consisting of carbon particles and a hydrophobic binder, Polytetrafluoroethylene (PTFE). Because MPL plays a critical role in the MEA, substantial research has been done to understand and improve the interfacial properties with regards to material design [51,52]. The MPL and PTFE ensures better electrical contact to the electrodes while maintaining moisture levels in the membrane, essential for achieving high proton conductivity [49]. GDL layers are fabricated to possess hydrophobicity to avoid flooding and enable rapid gas diffusion [18,49]. Traditionally, MEAs are manufactured by stacking the catalyst layer onto the GDL, then hot-pressing the GDL to the polymer membrane [49].     Figure 1-7: Cross section of MEA components [18] (reproduced with permission)  16   Figure 1-8: Micrographs of commercial carbon fiber paper and carbon cloths (a: Toray Industries, b: SGL, and c: E-Tek) [18] (reproduced with permission) 1.2.3.3 Catalyst layers  The efficiency and performance of a membrane electrode assembly (MEA) depends largely on the composition and microstructure of the two catalyst layers, especially the cathode, where the kinetically slow reaction of oxygen reduction takes place. The conventional material used for commercial applications consists of agglomerated domains of carbon blacks which form an electronically conductive structure by providing a percolation path for electron transfer. Typically, a thin catalyst layer of less than 0.5 mgPt cm-2 is desirable to maintain efficiency while keeping costs low [25]. The catalyst layer consists of dispersed nanoparticles of noble Pt catalyst (2 - 4 nm) on the carbon support and a coating of ionomer (such as Nafion) as the proton conducting network [27]. Together, the proximity of these three materials form an interface called the triple-phase boundary (TPB). This is where the electronic and protonic conductive phases and gas come into contact with the catalyst (Figure 1-9) [29]. One of the challenges in conventional fuel cell is attributed to the mesoporous carbon structure since carbon degrades during start up and shut down procedures.  This leads to carbon agglomeration and buried Pt particles which limit the TPB availability [53]. Figure 1-9: Schematic of agglomerate based microstructure and Triple Phase Boundary (TPB) 17  The degradation of carbon support and Pt are exacerbated by the presence of one another. Generally, the presence of Pt catalyzes the oxidation of carbon (1.9), while the oxidation of carbon accelerates Pt sintering/agglomeration.  The agglomerated or large Pt particles end up having lower surface area and limited access to the gas, reducing the catalyst activity and ultimately the fuel cell efficiency. Cathode flooding is also a concern since the gas cannot diffuse through the pores flooded with the product water at the cathode. Additionally, carbon monoxide (CO) concentrations as low as 10 ppm can significantly decrease the performance. CO poisoning occurs because CO adsorbs strongly to Pt surface causing the active catalyst surface area to decrease [25]. This thin layer of CO prevents the adsorption of H2 onto the Pt surface, therefore the reaction cannot take place, reducing the fuel cell efficiency. Similarly, CO2 can produce CO by the reverse water-shift reaction, leading to the same CO poisoning effect (1.9). With the diverse issues associated with catalyst active surface area, it is essential to gain a fundamental understanding of the catalyst layer.    C + 2H2O  CO2 + 4H+ + 4e- CO + H2O  CO2 + H2 E°= 0.207 V vs. RHE (1.9) A number of studies have been focused on improving the catalyst layer by introducing advanced materials that offer high surface area and stability. Several groups constructed novel supports to improve the Pt deposition and loading challenges [16,22,31,39,54,55]. Some groups manufactured nanowires, platinum coated whiskers, ordered carbon supports, and carbon nanotube supports with promising results [19,38,56–58]. Unfortunately, none of the studies show a significant reduction in catalyst loading while maintaining high level of fuel cell performance and satisfactory durability. To remain economically viable, much research is needed to improve the stability and efficiency of PEMFCs. In this study we will investigate the relationship between the catalyst layer’s material properties and the overall fuel cell performance.  18  1.2.3.4 Proton conductive membrane and ionomers According to the chemical reactions happening at the anode catalyst layer shown in (1.4), the hydrogen is split into protons and thus need to be transported across the proton conducting electrolyte membrane and recombine with oxygen to produce water at the cathode (1.5). In addition, a proton conductive network of ionomer must be present in the electrodes in order to carry the protons required for the redox reaction. Kreuer and Grotthuss proposed two main mechanisms for proton migration through the membrane and the ionomer network: (1) proton diffuses through the electrolyte medium as a “vehicle,” in this case as H3O+; and (2) “proton-hopping”, wherein protons are transferred from one vehicle to the next by simultaneous reorganization and reorientation of the species [59,60].   The first types of polymer used as a membrane material were perfluroinated polymers; however it was not feasible for prolonged usage because the durability was poor under oxidizing environments. Soon after, the industry discovered perfluorinated sulfonic acid (Nafion) membranes and ionomers, consisting of a polymerized tetrafluoroethylene (PTFE) backbone, giving it reliable chemical inertness. The cation exchange is realized by the sulfonic acid side chain groups –SO3H. The acidity or exchange capacity of this polymer is measured by equivalent weight (EW)—lower EW denotes higher conductivity [18]. Nafion is currently the preferred ionomer membrane for fuel cell applications [18]. One limitation to Nafion, is that it is strongly dependent on the level of moisture. Beyond 100 °C there is insufficient hydration in the membrane and the conductivity drops dramatically.       Figure 1-10: Chemical structure of Nafion (perfluorinated sulfonic acid) 19   Current catalyst layer materials  Conventional support material will be discussed in detail to highlight key challenges and disadvantages in catalyst performance. Because of current material challenges, an increasing number of research is focused on developing the support material to improve performance and durability in the fuel cell. Development of new carbon based catalyst support structures show advantages in terms of durability and catalyst activity compared to conventional materials. The fabrication and characteristic of the new carbon materials will be discussed.  1.3.1 Conventional carbon supports  The catalyst support has a crucial function to the fuel cell performance. The support must have high surface area, porosity, electrical conductivity, and corrosion resistance. Thus, low cost carbon materials are traditionally used as the Pt support in low temperature PEMFC. Carbon black, typically produced by pyrolysis of hydrocarbons such as natural gas or oil fractions from petroleum processing, are widely used as catalyst support in PEMFC. Carbon black is a polycrystalline or amorphous form of carbon consisting of spherical particles of graphite below 50 nm in diameter and up to 250 nm when coalesced into aggregates [34]. The morphology and particle size distribution varies depending on the raw material and the thermal decomposition process. One example of an oil-furnace carbon black is Vulcan XC-72. This low cost and abundant material is produced by burning raw materials at about 1400 °C. The surface area of the support material is of utmost importance since it is related to Pt active surface area. Typically, the surface area of Vulcan material is about 250 m2 g-1, however with additional activation steps, the BET surface area could increase up to 800 – 1200 m2 g-1 depending on the degree of activation [34]. CB can be activated by chemical (oxidative treatment) or physical activation (thermal treatment). Some known chemical activation methods include: addition of ZnCl2 or H3PO4 to the carbon precursor 20  or oxidative treatments using HNO3, H2O2, O2 or O3 [22,34]. Activation of CB increases the microporosity (pores < 2 nm), thereby improving Pt dispersion and utilization [22]. Thermal treatment involves heating the carbon under inert atmosphere up to 800 – 1100 °C or in air/steam at 400 – 500 °C. The process removes surface impurities, thereby increasing the active surface area as shown by an increase in capacitive current during CV measurements [22].   Because PEMFC operates in extreme conditions, high humidity, low pH (< 1), and strongly reducing or oxidizing atmospheres (at the anode and cathode), there are often issues with catalyst and carbon stability. In addition, the path for electron, proton, and gas transport is erratic and sluggish because the microstructure is irregular and non-homogenous. Some other persistent problems associated with CBs include: (i) presence of trace organic sulphur compounds residual from petroleum (raw material) [16] and (ii) deep depressions that submerge the catalyst nanoparticles making them inaccessible to reactants leading to poor catalytic activity [16].  Additionally, CB is electrochemically unstable under certain fuel cell conditions, causing corrosion and thinning of the catalyst layer, which leads to the loss of active surface area and performance of the fuel cell [16]. Pt active surface area decreases because carbon degradation causes Pt particles to coalesce and form large aggregates or dissolve into the ionomer phase. 1.3.2 Advanced material supports  There are several types of carbon structures that were studied for fuel cell applications. Antolini gives a comprehensive review of new carbon materials for the application of fuel cells and their performance and attributes compared to traditional materials [22]. Some of the mentioned carbon structures include mesoporous carbons, carbon nanotubes, carbon nanohorns and nanocoils, activated graphite fibers, and activated carbon nanofibers.  21  Because of the complex structure of conventional catalyst layers and the lack of knowledge about the factors that influence the catalyst layer efficiency, alternative structures such as carbon nanofibers and carbon nanotubes were investigated to improve catalyst efficiency and increase electrochemical surface area (ECSA) [35,41,54,56,61–64]. Owing to its large aspect ratio, Pt can be thinly distributed over the fiber surface; thus reducing costs and maintaining high efficiency. The continuous network of fibers provides high porosity improving fuel diffusivity and electrical conductivity by having an organized and straightforward path. The most traditional methods of fabricating CNFs are catalytic chemical vapor deposition or chemical vapor deposition with variants like thermal and plasma-assistance.  The first CNFs were produced by decomposition of hydrocarbons over metal particles. Unlike CNT that are generally few tens of nanometers in diameter and few micrometers in length, CNF may go up to 500 nm in diameter and span several millimeters in length. The main differences between CNF and CNT are their size, morphology, processing technique, and price. Additionally, CNFs can be fabricated in several ways resulting in different structures and dimensions. In general, CNF can be categorized into three types: (i) ribbon like CNF, (ii) platelet CNF, and (iii) herringbone (or stacked-cup) CNF [44]. Each have their own unique set of properties and characteristics, however, herringbone is considered to be a balance between ribbon and platelet types, yielding higher catalytic activity than ribbon and better durability than platelet forms [16]. Since CNF have a peculiar structure, they generally do not need to be pre-treated for functionalization. CNF prepared in this manner can achieve a surface area of up to 120 m2 g-1, diameter of about 100 – 150 nm and length of about 5 – 50 um [65]. Overall, CNF produced by chemical vapor deposition (CVD) improved catalytic activity in the fuel cell and obtained comparable ECSA to CNT at a fraction of the cost [66].  22  One of the earliest studies of CNF as catalyst support was reported by Rodriquez et al. for the support of Fe-Cu particles [67]. Since then, CNF has become a popular research topic as fuel cell catalyst support materials [68]. The CNF were generally fabricated by CVD techniques. Much of the reported uses of CNF are related to applications associated with direct methanol fuel cell (DMFC). Although the reports found positive and promising results, the DMFC assembly is beyond the scope of this project [69–72]. Of the CNFs reported for use in PEMFC, Li et al. prepared Pt nanoparticles (2 – 4 nm) supported on stacked-cup (SC) CNFs using polyol process.  The Pt/SC-CNFs combination MEA displayed higher PEMFC performance than conventional CB supports (from 750 to 900 mW cm-2). According to the study, the high aspect ratio CNF formed a continuous conductive network within the Nafion matrix which attributed to improved fuel cell performance [63]. Zheng et al. also reported the effects of CNF microstructure on oxygen reduction reaction (ORR) activity [73–75]. They fabricated platelet (p-CNF), tubular (t-CNF), and fish bone (f-CNF) type CNF and conducted cyclic voltammetry analysis. The findings suggest that the ORR process on p-CNF, f-CNF and graphite are controlled by diffusion, whereas ORR on t-CNF is mutually controlled by surface reaction and diffusion [73,74]. Álvarez et al. synthesized CNF by decomposition of methane and deposited Pt by formic acid method. Single cell performance was conducted for CNF/Pt, Pt/f-CNF (functionalized), and Pt/CB. The ORR results found that CNF/Pt performed 1.5 times higher than Pt/CB at 600 mV [76]. The authors conclude that functionalization of the CNF surface negatively impacted the fuel cell performance because of the hydrophilic surface. The CNF was shown to be more resistant against electrochemical corrosion than CB under experimental conditions, possibly due to the difference in microstructure. The study suggests that improvement in ORR performance is attributed to the difference in textural properties and smaller number of micropores in the CNF support, thus increasing electroactive surface area and mass 23  transport compared to CB [76]. A brief table of notable CNF and CNT catalyst layers synthesized by different methods is presented in the appendix for further reading (Table A-1).    1.3.3 Geometry of catalyst support  Traditional CB have a non-homogenous and irregular microstructure that lack controllable parameters. Some research has been done to fabricate organized and ordered catalyst structures to improve the Pt utilization and mass transport. One study used aligned CNT (5 – 20 nm diameter and 10 µm length) prepared by chemical vapor deposition as a cathode catalyst substrate. The aligned CNT decorated with nano Pt was synthesized with tunable nanopore diameter and structure, providing efficient channel for gas transport between individual CNTs [56]. Compared to commercial ink based MEA (470 mW cm-2), the aligned CNT achieved a peak power density of 860 mW cm-2 at 0.4 V in oxygen atmosphere. The authors concluded that the substantial improvement in power density is due to better mass transfer of fuel and conductance between MEA and current collector by having aligned CNT structure and uniformly ordered nanopores [56]. A few studies have also fabricated ordered mesoporous carbon as electrocatalysts for fuel cells [32,55,77,78]. Kim et al. reported a novel platinum electrocatalyst for PEMFC application [24]. The unique 3D ordered macroporous inverse opal structure exhibited large porosity and low tortuosity with interconnected macropore network. The Pt (wall thickness 10 – 15 nm) electrode thickness was 1.5 – 2 µm thick, decreasing diffusion distance and increased conductivity of reactants and ions. The 3D ordered structure achieved a current density of 440 mA cm-2 at 0.6 V under fully humidified operating conditions [24]. Polarization results revealed enhanced mass transfer and water management compared to ink based catalyst layers in the high limiting current region because of the desirable 3D ordered geometry. The study concluded that 3D macroporous inverse opal structures had superior mass transport and water management as a result of the large 24  effective porosity and interconnected macropores, yielding short diffusion pathways and low tortuosity [24]. Xia et al. created a bio inspired hierarchical nanoarray decorated with Pt nanoparticles as a cathode layer [58]. The highly oriented nanowire array provided smooth mass transport channels facilitating fast charge and mass transport. The well-defined microstructure surpassed DoE goals achieving peak power density of 0.778 W cm-2 with a low Pt loading of 0.065 mg cm-2. When compared to conventional catalyst with higher Pt loading (198 µg cm-2), the nanoarray demonstrated excellent Pt utilization and ECSA, achieving similar OCV [58]. Overall, the authors concluded the threefold power density improvement compared to conventional CB is due to the finely oriented nanowire arrays that provide smooth mass transport channels, enhancing Pt utilization and eventually reducing fuel cell costs [58]. The current literature establishes the remarkable improvements in PEMFC using ordered catalyst structures, addressing issues such as water management, mass transport loss, and Pt utilization.       1.3.4 Electrospinning in fuel cell applications   More recently, electrospinning has become a popular way to produce continuous CNFs due to the relative simplicity of the fabrication technique and controllable fiber properties [79]. Because of the relatively low fabrication cost and tailorable processing parameters, there is a growing number of research utilizing electrospun nanofibers as electrode materials [80–84]. The fabrication process involves electrospinning a precursor material (Polyacrylonitrile, Lignin, Polyacrylic Acid) which is later carbonized at elevated temperatures to produce carbon fibers (Chapter 1.5). CNFs made by electrospinning are beneficial because the fibers are continuous and the dimensions and porosity can be tailored [85–91]. Fiber diameter will vary depending on the precursor material and electrospinning conditions used. According to the review by Cavaliere, CNF diameters for catalyst support layers are typically between 50 – 700 nm. Electrospun CNFs 25  have been tested in DMFC and PEMFC fuel cells using Pt, PtRu, Pd, and PtRh catalysts [84]. Typical methods of depositing catalyst onto the CNF layer include electrospinning a suspension of Pt powder, electroless plating, electrodeposited, wet impregnation, magnetron sputtering, and spraying Pt powder onto the carbon layer [75,84,92].  One of the pioneer work conducted by Mathew et al., demonstrated the use of electrospinning to fabricate electrodes using three commercial catalysts [93]. Nanofiber mats were fabricated using a spinning dope consisting of a commercial catalyst (Vulcan and Ketjen Black), Nafion solution, and Poly(acrylic acid). The Poly(acrylic acid) was used as a carrier polymer to prevent droplets and form continuous Nafion fibers [93]. The study found nanofiber electrode based MEAs achieved power densities of up to 470 mW cm-2 at 0.65 V under moderate operating conditions. The study also concluded that the performance of the MEA was insensitive to changes in the fiber ionomer content and did not change significantly with average fiber diameter in the range 200 – 500 nm [93]. Another two studies published by the same group showed the performance was improved because of the nanofiber morphology allowing lower Pt loadings and rapid expulsion of water [94,95]. In another study, Polyacrylonitrile (PAN) based electrospun CNF sprayed with 20 wt% Pt slurry was characterized in PEMFC [35]. The results showed that the surface area before Pt deposited and after deposition on electrospun CNF did not change (~300 m2 g-1). This is an important observation because it demonstrates that the Pt is unable to penetrate through the shallow pores on the fiber surface, maintaining the large surface area.  Furthermore, electrospun CNF was shown to possess much higher Pt utilization than Vulcan XC-72R (carbon black), 69% compared to 35% [35]. In a similar study, Pt nanoparticles were deposited onto PAN CNF for MEA testing[96]. The results demonstrated that functionalized CNF showed maximum Pt utilization when compared to other Vulcan and other variations of CNF. The in situ fuel cell 26  tests revealed decent maximum power density of 294.7 mW cm-2 for functionalized CNF [96]. Instead of electrospinning a support layer, Sung et al. electrospun Pt nanowires and mixed them with commercial Pt/C catalyst to form the cathode electrode [97]. The authors report the Pt nanowires catalyst blend may have improved the fuel cell performance due to the increase in exchange current density [97]. The results also demonstrated the importance of Nafion concentration as it relates to the catalyst utilization and fuel cell performance. The authors also mentioned future investigations should focus on the ionomer distribution and whether ionomer preferentially bonds to Pt nanowires or Pt/C [97]. Another study demonstrated that Pt nanowires could be deposited onto electrospun CNF derived from Polyvinylpyrrolidone (PVP). Pt nanowires (2 – 5 nm diameter and 100 nm length) on CNF created a hierarchal structure, credited for improving the electrochemical properties [57]. The authors concluded that the flexible carbon fiber film demonstrated high surface area (33.5 m2 gPt-1), good gas diffusion, and electron transport compared to carbon black catalyst [57]. Zamani et al. fabricated Iron-Polyaniline-Polyacrylonitrile nanofibers as non-precious metal catalysts for ORR. The study demonstrated the most active non-precious ORR nanofiber catalyst prepared by electrospinning to date. Using graphene doped PAN/Polyvinylident fluoride to fabricate an electrospun catalyst layer, Wei et al. demonstrated an improvement in ECSA as a result of the porous nanofiber structure. The unique electrode structure achieved higher peak power density (0.88 W cm-2 for Pt loading of 0.2 mg cm-2) compared to conventional electrodes (0.73 W cm-2 for Pt loading of 0.4 mg cm-2) [98].    Instead of using CNF as a catalyst layer, Duan et al. successfully produced CNF microporous layer by electrospinning PAN and has achieved excellent three dimensional porosities, improved gas diffusion and exhibited excellent hydrophobicity. In his work, Pt/C (0.4 mgPt cm-2) was sprayed onto a Nafion membrane and hot pressed onto a CNF gas diffusion layer 27  [99]. Overall it was shown that CNF is a feasible material for fuel cell applications owing to the enhanced electrical conductivity compared to commercial Vulcan XC-72R (> 10 S cm-1 for CNF) [35,79,83,94,99,100]. Another attractive aspect of electrospun nanofibers is the potential to fabricate highly aligned and precise patterned nanofibers mats [101]. Nanofibrous polymer electrolytes have also been widely published in the literature [102–108]. The polymer membranes are usually fabricated as electrospun composite nanofibers containing Nafion and a carrier or uncharged polymer [102,103,109]. The electrospinning method offers higher network ion exchange capability, which is crucial for proton transport [109]. Evidently, considerable interest is shown in fabricating electrospun CNF for the application of fuel cell electrodes and catalyst support layers.  Currently, electrospun fibers are gradually becoming attractive for various fuel cell components. Electrospinning also has the ability to form different mesh structures. Electrospinning allows simple tailoring of key characteristics such as porosity, fiber diameter, layer thickness, mechanical, and electrical properties. Improvement in mechanical or electrical properties could also be fulfilled by incorporation of nanofillers in the spinning dope, such as CNT [110]. More importantly, fiber orientation and geometry can be controlled using electrospinning. For example, aligned fiber bundles, twisted yarns, unidirectional, orthogonally aligned, and coaxial fibers have been reported in literature [101,111–113]. Since geometry has the potential to improve mass transport and overall fuel cell performance (Chapter 1.3.3), electrospinning is a strong candidate to address the current catalyst layer challenges. The literature review suggests that a systematic study to tailor the fabrication process and design parameters for electrospun CNF has not been thoroughly studied for the application of catalyst support layers. For example, the effect of fiber 28  orientation on fuel cell performance, which is a unique contribution of this thesis, has not been explored in the literature.    The electrospinning process  Electrospinning is a simple process that can produce polymeric fibers in the nanometer to micrometer range by using an electric field. Ultrafine fibers can be produced by applying high voltage to the polymer solution. The process involves applying a sufficiently large voltage to a small droplet of polymer. The droplet becomes charged under the electric field and undergoes electrostatic repulsion. When the electrostatic repulsion becomes large enough to overcome the surface tension of the polymer solution, the droplet stretches, and at a critical point a jet of polymer erupts from the surface creating a Taylor cone [114]. While traveling from the spinneret nozzle to the ground collector, the polymer solution jet undergoes a continuous whipping motion caused by the electrostatic repulsion initiated at small bends in the polymer jet. Within this instability region where whipping and elongation occur, the solvents are evaporated leaving solidified nanofibers on the collector plate (Figure 1-11). The process of whipping, thereby elongating the fiber, is the key factor in the formation of continuous and uniform nanoscale polymer fibers [114].  29   Figure 1-11: Schematic illustration of electrospinning process Electrospinning is an attractive method because the processing parameters which directly affect the fiber morphology can be controlled. For example, fiber diameter can be sufficiently controlled by varying polymer concentration, molecular weight of the polymer, spinning distance, voltage, and orientation or speed of the collector plate. Typically, smaller diameters can be achieved by decreasing the polymer concentration in the mixture, and vice versa for fabricating fibers with large diameters. The electrospinning process produces semi-crystalline nanofibers where some polymer chains form crystalline lamellae and others form an amorphous phase. The amount of lamellar structure determines the crystallinity of the fibers [110]. The shear force during the electrospinning process aligns the polymer chains along the fiber length. The resulting nanofiber deposits onto a collector plate creating a randomly deposited non-woven fabric membrane.   The electrospinning process is governed by a variety of forces which include Coulomb force between the charge particles on the jet surface, viscoelastic force due to the viscosity of the solution, the electrostatic force due to the external electric field, surface tension force, gravity 30  force, and drag force due to air friction [114]. The different stages of electrospinning have been investigated extensively through modeling and theoretical analysis [115,116]. Because the electrospinning process parameters are a function of fluid properties, the relationship between spinnability and polymer molecular structure is critical. The convoluted interplay of these processing parameters make it impractical to establish a universal formula to include all polymer-solvent systems. An approximation was created, called the Berry number (Be), which combines concentration and viscosity parameters [117]. Developed to determine spinnability and predict fiber diameter, the dimensionless number is defined as the product of the polymer concentration (c) and the solution intrinsic viscosity ([η]) (1.10) [87,88]. The fiber diameter can be expressed as a power law equation where α and β are experimental coefficients (1.11) [87]. The coefficient α is correlated to molecular weight, radius of gyration of molecular chains, and entanglement of the molecular chains. Whereas coefficient β is related to crystallinity. The relationship shows that diameter is influenced by molecular conformation, molecular weight, and concentration of the polymer. As a result, fiber diameter increases as polymer concentration increases and the rate of increase is greater at higher molecular weights. A Berry number between 1 and 2.7 predicts a fiber diameter in the range of 100 – 500 nm with sufficient molecular entanglement to form continuous fibers [87].        𝐵𝑒 = 𝑐[𝜂] (1.10) 𝑑 = 𝛼(𝐵𝑒)𝛽 (1.11) 𝑑 = [𝛾𝜀𝑄2𝐼22𝜋(2 ln 𝜒 − 3)]1/3 (1.12) Another set of parameters that control electrospinning parameters include: spinning distance, flow rate, applied voltage, and spinneret diameter. According to Rutledge et al. [118], 31  the diameter of electrospun fibers is governed by equation (1.12), where d is fiber diameter, γ the surface tension, ε the dielectric constant, Q the flow rate, I the current carried by the fiber, and χ the ratio of the initial jet length to the nozzle diameter. In essence, these physical parameters relate back to the forces acting on the polymer droplet during the formation of the Taylor cone. Because of the accessibility to process a wide range of polymers, electrospinning has become a popular method to synthesize nanofiber based materials for various applications including filters, batteries, and medical scaffolds.    Carbon fiber fabrication  Carbon fiber (CF) are synthesized from precursor material such as rayon, petroleum pitch, and PAN. Generally, the precursor fiber is subjected to three stages of heat treatments—stabilization, carbonization, and graphitization. The first stage involves stretching the fiber in an oxidizing atmosphere at 200 – 300 °C under tension. During this step, the thermoplastic is converted into a non-plastic ladder structure. During carbonization, the fibers undergo another transformation at about 1000 °C in inert atmosphere. To improve the degree of graphitization, the fibers are further heated at 1500 – 3000 °C. The entire process is a complex series of chemical reactions detailed in literature [119]. There are a number of parameters which affect the final material properties, including: precursor material, ramp rate, heating temperature, and heating duration. Since 90% of CFs are produced by PAN precursor, a brief review of the heat treatment process and resulting material properties is provided [119].         32        Figure 1-12: Evolution of PAN chemical structure during the 3 steps of stabilization and carbonization (Top to bottom) [reproduced with permission] [120–122]  The mechanical properties of CF highly depend on the successful execution of the thermal treatment. Among the three processes mentioned, stabilization is the most complex and crucial stage as it sets the structure of the formation of CF. It is proposed that the stabilization process involves cyclization, dehydrogenation, aromatization, oxidation and crosslinking, eventually forming the ladder structure [123,124]. Optimal stabilization temperature is important because incomplete reaction will produce carbon fibers with poor properties. Generally, it has been argued that the optimal stabilization temperature for PAN is either 270 °C, 300 °C, or above 400 °C [119]. The two chemical changes during stabilization are dehydrogenation and cyclization which occur PAN nanofiber layer Step 1: Stabilization (oxidation in air)Step 2: Carbonization (heat treatment in Nitrogen or Argon)Step 3: Graphitization (heat treatment in Argon) Carbon nanofibers 33  in oxidizing atmosphere. Dehydrogenation forms double bonds by oxidation and elimination of water, as shown in Figure 1-12. This reaction is beneficial because the double bond increases the  thermal stability of the polymer and reduces chain scission during carbonization [125]. Next, the most important reaction involves rearrangement of nitrile groups with adjacent groups to form a stable ladder structure. The formation of ladder structure holds the molecules together and increases stiffness [126]. Cyclization involves conversion of triple bond to double bond (C≡N to C=N) in oxidizing or inert atmosphere, resulting in three hexagons laterally bound by nitrogen atom. Carbonization typically occurs between 800 – 1000 °C [127], where lower temperatures of ~1000 °C produce low modulus type fibers and 1500 °C produces intermediate modulus type fibers [128]. The process occurs in either nitrogen or argon rich atmosphere at a heating rate of 5 °C /min because of PAN structure instability. Higher heating rate causes a large amount of shrinkage, therefore it is beneficial to stretch and apply tension to the fibers beyond 600 °C in order to preserve orientation and overall mechanical properties [129,130]. Very high modulus type fibers can be produced by further graphitization beyond 3000 °C, achieving 99% carbon conversion from PAN [131].  Typically carbon nanofibers with higher Young’s modulus can be achieved with an optimized stabilization temperature compared to fibers that have not been well stabilized [128]. A significant study by Yamashita et al. demonstrated that the fiber diameter of electrospun Poly(acrylonitrile-co-methyl acrylate) (PAN-co-MA) decreases after heat treatment [132]. They further observed that the stabilized fiber mats achieve higher Young’s Modulus with increasing stabilization temperature. Consistent with Ko’s results and other published results for PAN based CNF, the surface electrical resistance decreases as heat treatment temperature increases [132,133]. In Yamashita’s study, FT-IR results demonstrated that the ladder structure for PAN-co-MA was 34  achieved at 300 °C [132].  However, it was also dependent on retention time and heating rate. Yamashita noted that beyond 280 oC, the fiber morphology started to fuse and change shape. It was also believed that the color change, from white to yellow to black, during this stage is due to the formation of the ladder ring structures [134]. Several groups have formed yarns or twisted bundles of CNF to improve the mechanical properties [82,85]. One study electrospun PAN onto a rapid rotating disc to form aligned fiber bundles. Tensile and electrical properties were analyzed for a series of heat treatment temperatures. Zhou et al. revealed CNF bundles carbonized at 1000 °C achieved 325 MPa and 40 GPa for tensile strength and Young’s Modulus, respectively [135]. For fibers carbonized at 2200 °C, tensile strength and Young’s Modulus was 542 MPa and 58 GPa, respectively [135]. Similarly, electrical conductivity for CNF carbonized at 1000 °C was 180 S cm-1 in the parallel direction, while fibers carbonized at 2000 °C achieved 840 S cm-1. The authors concluded the improvement in electrical and tensile property is due to the high degree of graphitization and orientation [135]. The authors also mention improved properties which could be achieved by using PAN copolymer because the sudden evolution of heat during stabilization of PAN homopolymer may cause chain scission, making final CNF mechanically weak [135,136]. Not only do comonomers prevent thermal shock, they also act as plasticizer to break up the structure and allow polymer to be easily dissolved in the spinning solvent, improving the quality of spinning and fiber morphology [136]. Additionally, the applied tension should be optimized during heat treatment to achieve high degree of orientation [135].        Research objectives  This thesis is a contribution to the field of fuel cell development, specifically it addresses the fabrication and optimization of a novel catalyst support material for PEMFCs. The predominant purpose of this work was to develop a catalyst support, with predictable microstructure and 35  material properties, through a tailorable fabrication process to establish the relationship between the material and geometrical properties and overall fuel cell performance. The hypothesis of this work was that geometrically orientated CNF produced by electrospinning would provide an ideal microstructure to improve mass transport, proton transport, durability, and Pt utilization. The effect of Pt loading, Pt deposition method, ionomer distribution, fiber orientation, and fiber electrical conductivity on fuel cell polarization was explored. To evaluate the advantages of tailorable CNF as catalyst support material, full characterization using electrochemical and physicochemical methods was proposed. Specifically, the predominant objectives of this thesis was to understand the fuel cell’s performance in the context of the following material and architectural characteristics:  I. Effect of fiber orientation (random vs. orthogonal), II. Effect of fiber electrical conductivity,  III. Effect of Pt deposition method and Pt loading, and IV. Effect of ionomer loading and distribution  Upon completion of this project, information about the catalyst support microstructure and properties would assist in the future design of optimized MEAs that could significantly improve fuel cell cost and efficiencies. This research may potentially stimulate new fuel cell designs and uncover a new research field in advanced materials. If successful, this research may lead to revolutionary designs not only in fuel cells, but also in filtration devices, supercapacitors, and lithium ion battery electrodes and separators.    In order to establish these structure-property correlations, a systematic approach was proposed as follows: 36  1. Select a polymer precursor and develop fabrication method to produce nanoporous carbon fiber mesh structure with controllable geometry and material properties. Optimize the fabrication process to achieve electrode target criteria such as high electrical conductivity, robust mechanical properties, small fiber diameter, and large porosity (Chapter 3 and 4).    2. Investigate a simple one step catalyst deposition method to achieve platinization of CNF for catalyst layer formation (Chapter 5).   3. Improve Pt and ionomer deposition by functionalization of CNF surface. Techniques include incorporation of surfactant polymers and acid treatment (Chapter 5)  4. Characterize microstructure, measure ex situ properties and in situ performance of CNF based fuel cell catalyst layers. Variables of interest are: ionomer and Pt loading, Pt deposition method, fiber orientation, and fiber electrical conductivity (Chapter 6).    Figure 1-13 provides an overview of the organization of this thesis. The research tasks are divided into three distinct phases that correspond to the research approach listed above. The flowchart illustrates the validation of each fabrication step to ensure the material properties sufficiently meet the design requirements.    37   Figure 1-13: Diagram of experimental approach illustrating the progression of experiments; tasks were divided into three key experimental phases: Fabrication optimization, deposition of Pt and ionomer, and MEA in-situ testing   38   Materials and Experimental Methods  The materials, experimental methods, and characterization methods presented in this thesis are described in their respective sub sections.   Materials  An electrospinning dope was prepared by mixing Poly-acrylonitrile-co-methyl acrylate (PAN-Co-MA) and dimethylformamide (DMF).  PAN-Co-MA with average Mw of 100,000, was purchased from Scientific Polymer Products. 7 wt% and 10 wt% PAN-Co-MA solution was made by dissolving the polymer powder in DMF (Sigma-Aldrich) solvent. Carboxyl functionalized MWCNT (COOH-MWCNT) purchased from Cheap Tubes Inc. with 10 – 30 µm length and < 8 nm diameter was used as received. The electroless plating solution for platinum deposition on nanofiber mats consists of Hydrogen hexachloroplatinate (IV) hexahydrate (H2PtCl6, 99.9 %, Alfa Aesar), formic acid (CH2O2, >95.0 %, Sigma-Aldrich), deionized water, and Polyvinylpyrrolidone PVP ((C6H9NO)n, Sigma-Aldrich). Aquivion® D79-25BS, Solvay Specialty Polymers) was diluted to 5 wt% and 10 wt% by a 3:1 volume ratio of N-Propanol and deionized water for the ionomer deposition solution. Electrolyte membrane, GDL, and anode CL was supplied by AFCC and used as is.    Preparation of random and orthogonally aligned nanofibers    The electrospinning process (NEU, Kato-Tech Co. Ltd.) was used to fabricate the as-spun nanofiber mat. PAN-Co-Ma was dissolved in DMF by weight percent. 7 wt% and 10 wt% PAN-Co-MA was dissolved in DMF and mechanically stirred on a hot plate at 85 °C for 12 hours. For spinning dopes requiring MWCNT, 1 wt% of the original polymer weight was measured and dispersed in the DMF first. The DMF and CNT suspension were bath sonicated for 30 minutes to encourage sufficient dispersion. PAN-co-MA was subsequently added to the suspension and 39  stirred in the same manner for 12 hours. Once the polymer solution was cooled to room temperature, it was transferred to a 10 mL plastic syringe with an 18G needle (0.838 mm diameter). A schematic drawing of the electrospinning apparatus is shown in Figure 2-1. The electrospinning parameters are listed in Table 2-1. The random fiber mat was collected on a rotating collector drum at ambient temperature with the relative humidity adjusted to roughly 30%. After spinning, the mat was stored in the oven (70 °C) to fully evaporate the solvent before further post processing.  Table 2-1: Electrospinning parameters for RNL and OAN  Random Orthogonal Spinning Distance 15 cm 10 cm Voltage 17 kV 17 kV Pump speed 0.80 mm/min 0.80 mm/min Rotation speed 1 m/min N/A Traverse speed 10 cm/min N/A  Using the prepared polymer dope, orthogonally aligned fiber mat was fabricated using a parallel electrode method as described in the literature [137]. The ground wire was mounted to one pair of parallel plates and the distance between the parallel copper plates was maintained at a distance of about 7 cm. As shown in Figure 2-2, the orthogonal fibers were fabricated on a desktop set-up with the collector and syringe needle horizontally parallel to each other. A 10 mL plastic syringe with an 18G needle (0.838 mm diameter) was used. The voltage was held at 17 kV, with a spinning distance of 10 cm. Relative humidity (RH) was maintained at 30% at ambient temperature. By alternating the repulsion voltage between the two orthogonal directions after every 30 minutes a substantial mat was collected on the copper plates. The fibers were transferred onto a stainless steel frame and set aside in the 70 °C oven for drying before further heat treatment.  40    Figure 2-1: Conventional electrospinning apparatus    Figure 2-2: Modified electrospinning apparatus for orthogonally aligned fibers  Heat treatment  The heat treatment process consists of a series of intricate chemical reactions in oxidative and inert atmospheres to produce graphitic carbon. The first stage was to oxidize/stabilize the polymer to form a stable structure that does not degrade during higher temperature treatment. Samples were restrained between two square stainless steel frames to reduce fiber shrinkage and wrinkling (Figure 2-3). The samples were placed on a graphite boat and inserted into the high temperature furnace for two hours at 230 °C or 260 °C in air according to the experimental design (Figure 2-4). The temperature ramp rate was set at 1°C min-1 as higher rates appear to produce mats that were prone to cracks due to severe shrinkage. Subsequently, the samples were taken out of the furnace and the frames were removed for the carbonization step. Since the carbonization step induces significant fiber and mat shrinkage (~30% reduction), applying tension on the sample causes cracks and defects. Instead, the stabilized fibers were sandwiched between carbon paper and restrained by a metal mesh to keep the mat flat and wrinkle free. The samples underwent a one hour heat treatment in argon at 850 °C or 1100 °C, according to the design of experiments (Chapter 3.3). The ramp rate again was kept at 3 °C min-1 to prevent shrinkage and damage to the sample. 41  Figure 2-5 illustrates the transition between as-spun, stabilized, and carbonized sample. The furnace was allowed to cool down to ambient temperature without a pre-set cooling rate. Carbonized fiber mat was then cut to the desired dimensions for Pt deposition.   Figure 2-3: Stainless steel frame for applying tension to as-spun mat during heat treatment       Figure 2-4: Tube furnace supplied with Air/Argon gas for carbonization procedure (left); graphite sample holder plate for keeping CNF flat during heat treatment  42   Figure 2-5: Transition between (A) as-spun, (B) stabilized, and (C) carbonized fiber mat    Platinum deposition  Several deposition methods discussed in literature were explored. In this study the different deposition methods are expected to yield unique morphology and dispersion, thus investigating the simplest and repeatable deposition method was necessary. Common deposition methods were evaluated including: Atomic Layer Deposition (ALD), hydrogen reduction, acetic acid reduction, and formic acid reduction. ALD deposition was performed in the 4D labs at SFU using the Cambridge NanoTech Fiji F200. The film was deposited under a pressure of about 5 x 10-3 mbar. MeCpPtMe3 was used as a Pt precursor with the pulse time of 20 s and the air pulse time of 20 s. Purge times used in the experiment were 3 seconds. Oxygen was pulsed into the reactor with a flow rate of 30 standard cubic centimeters per minute (sccm). Thin film was deposited at 270 °C under 188 cycles.  Sputter coater was used to deposit a thin layer of Pt onto one side of the CNF mesh. Samples were mounted on 15 mm aluminum stubs with adhesive carbon tabs (Ted Pella) and the edges and sides of the glass were painted with colloidal silver to ensure electrical connection between the edge of the sample and the stub. The region of interest was sputter coated with a 20 nm layer of Pt (EM MED 020 fitted with a quartz thickness monitor, Leica Microsystems, Concord, ON, Canada) For the Pt deposition using wet chemistry, several pre-treatment methods were investigated to identify the most suitable method for CNF material. With a few exceptions, the methods closely  43  followed literature with reduced submersion time because CNF surface is less inert compared to CNT [138]. One variation, the carbon nanofibers were pre-treated in methanol for 30 minutes before submerging in plating solution [139].  Second variation, CNF were submerged in a mixture of 0.1 M citric acid for 1 hour and subsequently in a solution mixture of 1 mL (3-Aminopropyl) triethoxysilane and 50 mL ethanol for 30 minutes and dried in the fume hood [140]. Pre-treated and dried CNF were then submerged in Pt plating bath solutions described below. In the first method, CNF were submerged into an acetone solution containing H2PtCl6 and evaporated completely in the fume hood. The amount of dissolved H2PtCl6 was calculated based on the desired Pt loading and estimated surface area. For example, the surface area of a 25 cm2 CNF is estimated to be 5000 cm2 cm-2MEA (based on the calculations shown in chapter 2.7.6). To achieve 0.15 mg cm-2MEA, the amount of Pt needed would be 750 mg. Following the stoichiometric conversions for molar mass of H2PtCl6 and elemental Pt, a rough estimate of 1.6 g of H2PtCl6 should be dissolved in 40 mL acetone. The dried sample was reduced at 350 °C in 10% Hydrogen/Argon atmosphere furnace for 1 hour at a ramp rate of 10 °C min-1.  In the second approach, reduction by acetic acid was performed according to method proposed by Kamavaram et al. [138]. 25 mL acetone was used to dissolve 45 mg of Platinum (II) Acetylacetonate (Pt(II)acac). Using a condenser under the fume hood, the solution was heated in an oil bath at ~100 °C for 2 hours. 10 mL glacial acetic acid was slowly added to the solution to reduce the Pt2+ ions. The solution was heated for another hour using a condenser, then allowed to fully evaporate overnight. The solution changes color from yellow to dark brown and finally black upon complete evaporation of the solvent.  44   Figure 2-6: Plastic frame holding CNF, used during Pt deposition Table 2-2: Platinum plating bath chemical composition and concentrations No. Components  Amount 1 H2PtCl6  25, 30, and 40 mg 2 Polyvinylpyrrolidone  2 mg 3 Deionized water   20  mL 4 CH2O2  1 mL Plating temperature 20-25 °C Time for plating 72 hours  H2PtCl6 + 2HCOOH  Pt + 6Cl- + 6H+ + 2CO2 ↑ (2.1) In the third approach, reduction of Chloroplatinic acid by formic acid was performed according to the method outlined by Sun et al. [141]. Since CNF were thin and delicate, a plastic frame was used to add support during Pt deposition and subsequent steps (Figure 2-6). Because of the extremely sensitive reduction technique, it is crucial to have clean tools and glassware for successful deposition onto the CNF. Therefore, thorough cleaning with 1:1 mixture of H2SO4 and HNO3 is recommended. The concentrations of the plating bath are shown in Table 2-2. Chloroplatinic acid and PVP was dissolved in 20 mL of DI water into a mixing beaker. The mixed solution was bath sonicated for 20 minutes and transferred to petri-dishes for plating. The CNF 45  mats were placed in the petri-dish and submerged completely in the plating bath solution. The petri-dish was wrapped with para-film and undisturbed for 72 hours at room temperature. The mechanism of Pt reduction is shown in Equation (2.1). The solution becomes dark after plating, indicating metallic Pt was formed. Once the plating is complete, the samples were carefully rinsed with deionized water.  The resulting CNF/Pt mat was dried in an oven at 100 °C and stored in a desiccator for later use.    Ionomer deposition   Figure 2-7: Ionomer deposition set-up Ionomer (Aquivion® D79-25BS, Solvay Specialty Polymers) was diluted to 5 wt% and 10 wt% by a 3:1 volume ratio of N-Propanol and deionized water. The solution was stirred slowly on the mechanical stir plate for < 4 hours to enhance deposition viscosity. The stirred solution was then poured through a funnel and filter paper using a vacuum pump. The set-up, shown in Figure 2-7, allows the ionomer to evenly deposit through the fiber mat without blocking the pores. Another method, dip coating, was used to compare the dispersion of the ionomer and study the effect of concentration on pore blockage, but proved to give less desirable results. The coated fiber mat was annealed in a convection oven at 140 °C for 15 minutes.   46   MEA assembly  Plastic frames from the previous fabrication step were removed by cutting with scissors or razor blade. Anode half catalyst coated membrane (CCM) (Pt loading 0.05 mg cm-2) with NRE-211 membrane was hot pressed onto the CNF sample and trimmed to the desired size for FC testing. The heat press was activated for 2.5 minutes under 9.8 kN at 150 °C. The GDL for cathode (Freudenberg, H23C10) and anode (Freudenberg, H14C10) were attached using an adhesive thermoset sheet and heat pressed together for an additional three minutes (Figure 2-8). The completed MEA has an active area of 5 cm2 (CAT cell, Figure 2-9) and 1.13 cm2 (MAT cell, Figure 2-10), suitable for small scale fuel cell test station.    Figure 2-8: Exploded view of the MEA (CAT cell) components for hot pressing  Figure 2-9: Completed MEA for CAT cell testing 47    Figure 2-10: MAT MEA sample: circular CNF sample hot pressed onto NRE-211 membrane and anode CL Baseline CCM (V50) was prepared by AFCC staff using a conventional decal method. In the conventional decal method, the catalyst ink slurry was made by dispersing catalyst (50 wt% Pt on Vulcan XC-72 carbon black) with a mixture of 5 wt% Nafion ionomer solution, isopropyl alcohol, and DI water. The weight ratio of the ionomer (dry basis) to carbon (I/C) was maintained at 1.1. The catalyst inks were coated on polytetrafluoroethylene (PTFE) sheets that subsequently served as the decal substrates using metering bars (Meyer bars) under a motorized draw down table. Target Pt loadings for the cathode and the anode were 0.15 and 0.05 mgPt cm-2, respectively and were measured on dry decal by XRF for at least 2 locations per coating to gauge uniformity. Freshly coated decals were dried on a hot plate set at 70 °C in the fume hood (to reduce cracks). After annealing at 150 °C in convection oven for 30 minutes, the catalyst layers were transferred onto the membrane (NRE-211) by hot-pressing to form the CCM. The GDL (Freudenberg) was identical to the ones used in the CNF MEA.   Characterization and analysis methods  2.7.1 Thermal profile of PAN-co-MA  To determine the range of suitable heat treatment temperatures for PAN-co-MA, TGA was conducted to observe the onset of oxidation. The nanofiber mat was dried in 70 °C oven overnight 48  to ensure all moisture is evaporated. The specimen was rolled into a ball and placed in a clean Pt TGA pan. The data was collected from room temperature to 600 °C at a ramp rate of 5 °C min-1 in air atmosphere. Comparing the onset of oxidation between PAN and PAN-co-MA can give information about suitable thermal stabilization temperatures.  2.7.2 SEM and TEM dimensional measurements   Nanofiber diameter and morphology were observed by Hitachi S-3500N and FEI Helios NanoLab 650 scanning electron microscope (SEM). As-spun and stabilized samples required sputtering to provide a conductive surface, which reduces charging and improves image quality. A thin coating of Au was sputter coated on the sample surface for 60 seconds at 40 amps using Edwards S150A sputter coater. Images were taken in several quadrants of the fiber mat and 100 diameter measurements were averaged using Image J software (U.S. National Institutes of Health). Fiber cross section, Pt particle size and ionomer distribution and morphology was observed using FEI Helios. The Pt particle size was measured using Image J and averaged over 30 measurements. Fiber mat cross sectional thickness was also measured using Image J and later confirmed using digital thickness gauge.  Samples post-fuel cell testing were embedded in resin and polished to a mirror finish for cross sectional imaging in SEM. Transmission electron microscope (TEM) and energy dispersive X-ray (EDX) maps of the samples were collected in order to observe Pt and ionomer distribution and film thickness. The ionomer thickness (including bridges) was measured using Image J and averaged over 200 measurements. Cross sections of the samples for TEM imaging were prepared by embedding small pieces of tested MEAs (with GDL removed) in TTE resin (trimethylolpropane triglycidyl ether and 4,4′- methylenebis (2-methylcyclohexylamine)). Thin (∼100-200 nm) TEM slices were microtomed using a Leica Ultracut UCT ultramicrotome (Leica Mikrosystem, Wien, 49  Austria), and deposited on 100-mesh TEM copper grids. STEM/EDX imaging was performed on FEI Tecnai Osiris (Scanning) Transmission Electron Microscope, with 200 kV accelerating beam voltage. 2.7.3 Mechanical testing  Figure 2-11: 2 cm gauge tensile test specimen Mechanical tensile testing of nanofiber mats were conducted using the KES-G1 Kawabata microtensile testing system at the elongation rate of 0.01 cm s-1. For random nanofiber mats, sample strips of 3 cm x 0.5 cm were glued on a paper frame and tested. The gauge length for tensile test is 2 cm (Figure 2-11). The measured load in grams was converted to specific stress (g tex-1) via the formula in Equation (2.2). Areal density represents the weight of the specimen divided by the area, in m2 (length x width). The specific stress in g tex-1 was then converted to N tex-1 by multiplying by 9.81 m s-2 and further to MPa by multiplying the bulk density of carbon. Stress (g/tex) = 𝐹𝑜𝑟𝑐𝑒 (𝑔)𝑊𝑖𝑑𝑡ℎ (𝑚𝑚)×𝐴𝑟𝑒𝑎𝑙 𝐷𝑒𝑛𝑠𝑖𝑡𝑦 (𝑔𝑠𝑞∙𝑚) (2.2) 2.7.4 In-plane and through-plane electrical properties  The random and orthogonally aligned CNF samples were characterized for electrical conductivity. The specimen were cut into small squares or strips and glued onto a glass slide using silver paint. Using a narrow brush, a thin and even layer was applied to the two edges that were 50  used to contact the electrical probes (Figure 2-12). The specimen dimension were measured with a micrometer and digital calipers to compute the conductivity per length. It was crucial that the specimen lays flat on the glass slide because any wrinkling would cause error in the calculations. Electrical resistance was measured with a two probe multi-meter (Agilent U1272A). Equation (2.3) was used for computing the resistivity and conductivity of the samples. ρ is the electrical resistivity measured in ohm-centimeter, A is the cross sectional area, l is the length between the silver paint and R is the electrical resistance in ohms. Conductivity (σ), measured in S cm-1, was found by reciprocation of the resistivity (ρ). At least 4 replicates were prepared and measured for each sample and the results were reported as an average with one standard deviation. An in-house apparatus for measuring through-plane resistance, mat thickness, and compression strength was used to mimic operational FC conditions (Figure 2-13). The measurements were recorded over a 25 s dwell at each pressure while fixing the current at 1 A. A series of electrical resistivity (Ωcm) were recorded at pressures ranging from 500 kPa to 3000 kPa. The final through-plane conductivity (S cm-1) was calculated by reciprocating the resistivity at 1500 kPa (roughly FC operational pressure).    51     Figure 2-12: Conventional multi-meter and specimen preparation with silver paint   Figure 2-13: Through-plane electrical resistivity tool, measures resistivity as a function of pressure and thickness               (2.3)  52  2.7.5 Chemical characterization   CNF were cut into small squares and mounted on a quartz single crystal holder using clear double sided tape. The chemical phase composition and crystal structure was confirmed by X-ray diffraction (XRD, Rigaku Multiflex, Japan) using Cu Kα radiation at 40 kV and 20 mA.  Diffraction data were collected from 2θ = 30° to 90° with a step size of 0.5°. Pt loading was examined by X-ray fluorescence (XRF) (Thermo Scientific, Niton XL3t). Fiber surface chemistry was examined by X-ray photoelectron spectroscopy (XPS) (Axis Ultra DLD, Kratos Analytical) using monochromatic Al Kα radiation at a power of 150 W. The pass energy was set at 20 eV, and the C 1s line at 284.6 eV was used as a reference. Because nanofiber sample tends to outgas, samples were cut into small squares and adhered onto the sample holder using minimal carbon tape. It is crucial to keep the sample dry and clean to avoid detecting contamination on the surface.  2.7.6 Porosity measurement and calculations  Porosity was calculated by estimating the fiber spacing and amount of fibers in a given volume based on measured weight and density of carbon. The equation below was used to approximate the % porosity in the fiber mats. Porosity of the fiber mats was calculated based on their mass (mCNF), carbon bulk density ρ = 1.9 g cm-3 and measured fiber mats dimensions, length (LCNF), width (WCNF) and thickness (CNF) (2.4). The calculated porosity was later confirmed by physical measurement methods.  𝑴𝒆𝒔𝒉 𝒑𝒐𝒓𝒐𝒔𝒊𝒕𝒚 (%) =𝑽𝒕𝒐𝒕 − 𝑽𝒔𝒐𝒍𝒊𝒅𝑽𝒕𝒐𝒕=𝑳𝑪𝑵𝑭 ∙ 𝑾𝑪𝑵𝑭 ∙ 𝝉𝑪𝑵𝑭 − 𝒎𝑪𝑵𝑭 ∙ 𝝆𝑪𝑳𝑪𝑵𝑭 ∙ 𝑾𝑪𝑵𝑭 ∙ 𝝉𝑪𝑵𝑭  (2.4) Porosity of fiber mat was measured by using Archimedes principle in Equation (2.5). A rig was installed onto the digital balance to measure the mass of the fibers submerged in water or ethanol. It was crucial to use a wetting agent that can fully wet the carbon nanofibers, thus ethanol 53  was chosen. The dry mass of the fiber was measured before submersion, denoted Mdry. The mass after submersion with excess solution gently patted dry was measured, denoted Mwet. The mass of the sample fully submerged is denoted Msub. The % porosity was calculated by the following equation:  𝑴𝒆𝒔𝒉 𝒑𝒐𝒓𝒐𝒔𝒊𝒕𝒚 (%) =𝑴𝒘𝒆𝒕 − 𝑴𝒅𝒓𝒚𝑴𝒘𝒆𝒕 − 𝑴𝒔𝒖𝒃  (2.5) 2.7.7 X-Ray fluorescence Pt loading was validated by X-ray fluorescence calibrated to a standardized platinum block. The average Pt loading was measured across the CNF sample in 3 locations. The measured area was 1 cm2. Front side and back side of the CNF sample was measured to ensure Pt deposited evenly through-thickness.   2.7.8 Ex situ cyclic voltammetry Cyclic voltammetry measurements were conducted at room temperature in a three-compartment electrochemical cell.  The electrolyte was 0.09M H2SO4, the working electrode was a circular piece of CNF/Pt sample impregnated onto  the mirror finish glassy carbon disk electrode (area = 0.196 cm2) by adding a 5 L drop of diluted Nafion solution (water/5% Nafion ratio was 100:1) on it and dried under a heat lamp. Princeton Applied Research model 263A potentiostat/galvanostat was used, together with the CorrWare and CView system software for data acquisition and analysis. The electrolyte solution was deaerated by bubbling Nitrogen through it for 30 minutes prior to electrochemical measurements. The electrode was activated by cycling the potential between 0.025 and 1.2 V vs. Standard Hydrogen Electrode (SHE) at a scan rate of 20 mV s-1. The potential limits were chosen to be near the H2 and O2 evolution potentials, respectively.  54  Three cycles were performed to reduce instabilities, the last cycle which was the most stable voltammogram was recorded for analysis.  Figure 2-14: CNF impregnated onto glassy carbon electrode for cyclic voltammetry (CNF geometric area was normalized for comparison)   2.7.9 In situ electrochemistry  Performance testing of the CNF and baseline MEAs was done using an AFCC in-house design rapid screening test apparatus (MAT cell), which utilizes straight channel cathode flow and radial anode flow with adjustable compression and no gasket sealing. For performance evaluation, the samples were tested in NOC (Normal Operating Condition) and HOT (High Operating Temperature) operating condition under voltage control mode with 100% H2 on the anode, air on the cathode, and 2.5 bara balanced inlet pressures. For NOC, the cell temperature was 68 °C and both fuel and oxidant inlet RHs were at 105%. For HOT, the cell temperature was 85 °C and the inlet RH’s were at 50%. Following the standard protocols used by AFCC, the overall performance was assessed by current density measured at 650 mV in NOC and 600 mV in HOT, while kinetic performance was assessed at 825 mV for both conditions, and oxygen transport assessment was done via the limiting current measured at 0.1 V in NOC. All performance measurements were repeated at least three times for each sample to ensure data quality.  For the full polarization analysis, the MEAs were tested in a single-cell hardware (CAT cell) with straight channels, and operated at 10.2 stoic (anode and cathode) to achieve iso-55  conditions in the cell. Prior to obtaining current–voltage (IV) polarization data, the cell was conditioned at 68 °C and 100% RH in the potential range between 0.9 and 0.3 V in 0.05 V steps for 10 hours. Polarization data were collected by scanning the cell voltage from open circuit voltage (OCV) to 0.2 V in 0.05 V steps (3 minutes at each potential point). The HOT and NOC operating conditions were identical as the MAT cell measurements. Scans were repeated two times to ensure reproducibility within each MEA. 2.7.10 Electrochemical impedance spectroscopy  Table 2-3: Test station set-up and EIS testing protocols   Parameter Units Value  Temperature °C 68, 85 Cathode     Pressure bara 2.00  Flow rate per cell SLPM 4.00  Composition % 100% N2  RH % 40, 50, 65, 80, and 100 Anode     Pressure bara 2.00  Flow rate per cell SLPM 1.00  Composition % 100% H2  RH % 40, 50, 65, 80, and 100 The impedance spectra were obtained by applying an AC voltage of 10.0 mV over the frequency range from 1 Hz to 50 kHz with the CAT fuel cell test statio. EIS data were analyzed using a Randles equivalent circuit and fitted using ZView software (version 3.1c Scribner Associates, Inc.). Initially, cells were conditioned in H2/Air at 100 % RH and 1.5 A cm-2 for 2 hours. Then the RH was brought to 20% until equilibrium was reached (~ 1 hour); after that the MEA was conditioned at the desired RH for another 20 minutes in H2/N2. All EIS measurements were recorded under H2/N2 atmosphere and in the order of lowest to highest RH (Table 2-3). Once the measurements were completed for one condition, the RH was increased and the sequence repeated.   56   Optimization of Randomly Oriented CNF Layers   Introduction  In order to study the factors affecting structure and properties of electrospun CNF, a factorial experiment was designed. The processing parameters were optimized by way of the response surface method (RSM). PAN-co-MA was selected as a suitable precursor polymer because the thermal profile indicated a steady and controlled stabilization reaction, improving the formation of ladder structures. Preliminary experiments were conducted to approximate the appropriate range for the independent variables (factors). The four factors in the experiment were: polymer concentration, CNT concentration, stabilization temperature, and carbonization temperature. The systematic approach demonstrated the main effects, in addition to which fabrication parameters played a significant role on the four responses (fiber diameter, porosity, Young’s modulus, and electrical conductivity). The factorial design indicated the most significant factors and interaction terms which was used to navigate the response surface. Contrary to the hypothesis and literature, addition of CNT in the polymer dope did not improve electrical properties and was removed from the process. There appears to be discrepancy between the validation experiments and the predicted model values; however, the trends were applicable as a starting point for the fabrication of catalyst layers with sufficient properties. The challenges and errors in the model are explained and discussed; suggestions are also made for future work. Overall, the RSM predictions generated a set of fabrication parameters which produced fibers that met the design criteria of a catalyst support layer.      Preliminary process screening  From the TGA spectra shown in Figure 3-1, it can be seen that both PAN and PAN-co-MA experience a rapid weight loss around 300 °C. Onset temperature was measured at the inflection 57  point, indicating the initiation of decomposition. PAN-co-MA revealed a decomposition temperature of 274 °C, whereas pure PAN had a decomposition temperature of 280 °C (Figure 3-1). It was previously reported that increasing the amount of the comonomer decreases the decomposition temperature [142,143]. The onset of rapid weight loss can also indicate the initiation of cyclization, a component of the stabilization reaction. Aside from TGA indications, DSC is also thoroughly reported for the stabilization of PAN copolymers. It is widely accepted that pure PAN is not a suitable candidate for the synthesis of carbon fiber because of the sudden and rapid evolution of heat [132,136,142,144,145]. As described in earlier chapters, the thermal shock can result in chain scission, yielding poor properties in the carbon fiber [136]. Comparing the DSC spectra of PAN and PAN-Co-MA, it is clear that the onset of cyclization occurs at a lower temperature, along with a slow and steady release of heat [145]. The authors proved that PAN-co-MA released less heat and underwent two distinct reactions, as opposed to pure PAN, where a large amount of heat is suddenly released rapidly (Figure 3-2) [145]. The gradual release of heat helps to maintain good chain structure during cyclization. The similarity of the TGA behaviour of the polymer used in this study and the reported literature suggests the polymer would behave similarly under heat treatment processing. It is necessary to choose an appropriate temperature range to optimize the heat treatment because the stabilization reaction occurs over a range of temperatures. Gupta proposed the stabilization temperature range for the PAN-Co-MA in his experiment occurs from 230 – 267 °C [145]. Gupta suggests that stabilization initiates from the amorphous phase at lower temperatures and goes to completion at higher temperatures in the crystalline phase. Because fabrication of nanofibers by electrospinning alters the crystallinity of the polymer, the amount of amorphous and crystalline phases vary [146]. Using these temperatures 58  as a starting point, a factorial design of experiments can be conducted to narrow down the optimal stabilization temperature.   Figure 3-1: TGA spectra of PAN-co-MA (top) and pure PAN (bottom) in air (5 °C min-1) 59   Figure 3-2: DSC curves on heating PAN-co-MA and homopolymer PAN in air at 1 °C min-1 [145] Table 3-1: Electrospinning parameters used for spinning 12 wt% PAN-co-MA  Electrospinning Parameters Spinning Distance  15 cm Needle gauge 18G Voltage 17.5 kV Pump speed 0.80 mm min-1 Collector rotation speed  1 m min-1 The preliminary results about the fiber diameter and electrospinning conditions can be used as a marker towards building a suitable factorial design. The conventional electrospinning apparatus for randomly oriented nanofiber layers was used for the following experiments. Using the electrospinning parameters shown in Table 3-1, 12 wt% PAN-co-MA was used to produce fiber mats. The fiber mats were stabilized and carbonized at various temperatures to determine the relationship between electrical properties and degree of graphitization. Fiber diameter after each stage of the heat treatment is shown in Table 3-2.   60   Figure 3-3: Fiber shrinkage after each stage of heat treatment (as-spun, stabilized, and carbonized) From the diameters measured by Image J, the average diameter decreases after carbonization step. The bulk fiber mat also shows shrinkage after carbonization (Figure 3-3). Because of the non-uniform shrinkage, a metallic frame was used to sandwich the fiber mat which applied tension to the sample during heat treatment. The frame kept the sample flat and applied enough tension to prevent deformation. Applying tension or stretching during heat treatment also improves crystallinity and mechanical properties [119]. From the SEM images, it is clear that the electrospinning produced uniform continuous nanofibers without defect or beading. This reveals that the spinning parameters used are suitable for fabricating continuous nanofibers.  Additionally, in-plane electrical conductivity was measured after carbonization to investigate the effect of increasing carbonization temperature. The stabilization temperature was kept constant for all samples while carbonization temperature ranged from 700 °C, 750 °C, 800 °C, to 900 °C. Electrical resistivity was measured for the strip samples using a multi-meter (Agilent U1272A). As can be seen in Table 3-3, the resistivity decreased dramatically at 900 °C. Electrical conductivity was calculated from Equation (2.3) as described in previous sections. The results reveal that the minimum temperature needed to produce satisfactory electrical conductivity (> 5 S cm-1) is around 900 °C. Above 900 °C, CNF conductivity will improve proportionally. The mechanical strength is also expected to improve at higher carbonization temperatures due to the formation of graphitic structures. However, increasing the Young’s modulus causes the CNF 61  samples to become brittle and fragile to handle. Drawing from the preliminary results, clearly a set of experiments should be conducted to determine the optimized processing conditions to meet all design targets. Comparing the target design criterion to the preliminary results, it appears that the fiber diameter is too large while the electrical and mechanical properties needs improvement.    Table 3-2: 12 wt% PAN-co-MA average fiber diameter after heat treatment process, fiber shrinkage of 30% (um) As spun Stabilized Carbonized  % change (As spun to Carbonized) Mean 0.458 0.355 0.309 32.53% SD 0.036 0.069 0.039  Min 0.412 0.236 0.256 37.86% Max 0.529 0.459 0.39 26.28%   Figure 3-4: SEM micrograph of As-spun 12 wt% PAN-co-MA fiber mat (5 kV; 3Kx left, 10Kx right)   Figure 3-5: SEM micrograph of stabilized 12 wt% PAN-co-MA fiber mat (5 kV; 3Kx left, 10Kx right) 62   Figure 3-6: SEM micrograph of carbonized 12 wt% PAN-co-MA fiber mat (5 kV; 3Kx left, 10Kx right) Table 3-3: Relationship between fiber in-plane electrical conductivity and carbonization temperature Carbonization Temperature (°C) Resistivity (Ωmm) Conductivity (S cm-1) 700 320 0.034 750 25 0.40 800 6.3 1.6 900 0.58 17    Two-level factorial design  A set of experiments was designed to optimize the fabrication process and produce fibers that meet design targets (Table 3-4). Fiber diameter, porosity, and layer thickness are essential parameters to control and observe. In addition to studying the microstructure, the catalyst support layer must have the basic functionalities such as being mechanically robust and electrically conductive to transport electrons.  A factorial experiment was designed using four factors to systematically develop an optimized fabrication process. The design aims to control the material properties essential in achieving a high efficiency catalyst layer. The surface area, porosity, electrical properties and mechanical robustness are of significant importance in the catalyst support design. The factors chosen for the experiment: (1) polymer precursor concentration, (2) fiber stabilization temperature, 63  (3) fiber carbonization temperature, and (4) concentration of the conductive filler material, were shown in prior studies to directly influence these criteria. It is well known that polymer concentration has a great influence on fiber diameter and spinnability [88], which would affect the final product surface area and porosity. Heat treatment temperature influences the electrical conductivity and mechanical integrity of the porous structure [82,119,147]. Addition of conductive filler material such as carbon nanotubes (CNT) is also shown to improve electrical properties of composite structures [140]. The two-level DOE produced 16 experimental runs as shown in Table C-1. Low and high level of polymer concentration was 7 and 10 wt%. Fiber diameters measured at 12 wt% already had fairly large diameters (Table 3-2), thus a much lower concentration was chosen as the low level to produce fibers with smaller diameter. CNT concentration of 0 and 1 wt% was chosen for the low and high level because higher concentrations would present challenges in dispersion [140]. Stabilization temperature was 230 °C and 260 °C for low and high level based on current TGA data and published results [145]. Ramp rate and durations were also found through the exploratory work. Ramping too fast caused fiber mats to crack and shrink, in addition to fiber fusion. Selecting a rate of 1°C min-1 ensured fibers experienced less shrinkage stress and maintained the fibrous structure. For complete stabilization, fibers should be calcined for at least one hour, some authors found two hours to be sufficient [132,144,145]. Carbonization temperature was 850 °C and 1100 °C for low and high levels based on the electrical properties seen in the exploratory work. Beyond 800 °C the electrical conductivity increased dramatically, reaching 17 S cm-1 at a carbonization temperature of 900 °C (Table 3-3). Experiments were carried out in standard order to reduce block effect and other environmental variations.  The 16 experimental results were inputted into the Design-Expert software for statistical analysis. Each factor was fitted to a regression model and verified for statistical significance by 64  ANOVA. The graphical representation of the factors and responses was used to identify the presence of factor interaction. Contour plots illustrate predicted values, not experimental values, from the design. Summary of the regression analysis and model constraints can be found in the appendix (Table C-2).  Table 3-4: CNF design targets for MEA application based on recommendations from AFCC and industry standards Parameter Target Unit Mechanical strength (compression without breaking) > 700 (4.8) psi (MPa) In-plane electrical conductivity > 5 S cm-1 Through-plane electrical conductivity > 0.05 @ 218 psi ( 1.5 MPa) compression at 10 µm S cm-1 Catalyst layer thickness < 10 µm Geometrical surface area > 50 cm2cm-2geom. Fiber diameter < 200 nm Mesh porosity > 40 % Area weight < 1.15 mg cm-2 3.3.1 Effect of process parameters on fiber diameter Using the Design-Expert software for model fitting, the main effects were computed.   Verifying the model using ANOVA method, it was found that the significant coded model terms were A, C, AD, BC, and ABC.  This means that fiber diameter was influenced by a combination of the main effects polymer concentration and carbonization temperature. And was also influenced by the interaction between polymer concentration/CNT concentration, stabilization/carbonization temperature, and polymer concentration/stabilization/carbonization temperature.  The results obtained from the model are consistent with published work which discuss that nanofiber diameter is directly related to polymer concentration. The diameter was also affected by carbonization temperature since the nanofiber experienced shrinkage during the heat treatment process. In fact, the fiber shrunk up to ~30% after carbonization (Table 3-2). Because of the 65  complex interaction terms, it is impossible to find the optimal fabrication parameters by simply looking at the design factors cube. For example in Figure 3-7, the cube shows the results for 0% CNT and 1 wt % CNT. The cube demonstrates that there was an interaction between polymer/CNT concentrations, since some vertices show a decrease of fiber diameter with decreasing CNT concentration, while others do not.  When this counteracting trend is observed, there is a strong interaction between two factors. Therefore, a prediction of the optimal process conditions is useful for responses that involve multiple main effects and interaction terms. Chapter 3.3.5 explains the model fitting and optimization of the factors based on RSM.  The fiber diameters from the experiment ranged from 159 nm to 372 nm. A graphical representation of the fiber diameter and distribution can be found in the appendix (Figure C-1 and Figure C-2). The fiber diameters were normally distributed and showed narrow standard deviation.   66    Figure 3-7: Factorial cube for fiber diameter recorded at 0 and 1 wt% CNT, red and blue circle indicates interaction   3.3.2 Effect of process parameters on fiber electrical conductivity  The main effects of fiber electrical conductivity were computed using the Design-Expert software. Verifying the model using ANOVA method, it was found that the significant coded model terms were C and D. Therefore, the electrical conductivity of the sample was simply related to carbonization temperature and CNT concentration. This result appears to be consistent with the 67  previously cited literature. In general, the designs factor cube and interactions curve revealed that there were no interactions. A strong interaction occurs when the response is different depending on the settings of two factors. When interaction curves intersect, it indicates that the effect of one factor depends on the level of the other. The interaction curve shows that D-1 and D+1 never crosses, this is a clear indication that interactions were not involved and the only main effects were carbonization temperature and CNT concentration.  A graphical representation of the electrical conductivities for each sample can be found in the appendix (Figure C-3). Contrary to the original hypothesis and published literature, the electrical conductivity was not improved by addition of CNT. The discrepancy can be attributed to the lack of dispersion in the spinning dope. During mixing, CNT could form agglomerates and become non-homogenously distributed in the fiber. Thus, there was no significant enhancement seen when CNT was added. The agglomerates obstruct the electric current from travelling through the fiber, the electrons must jump between CNT agglomerates instead of smoothly across a streamlined and straight connected network. Several studies have mentioned that the electrical properties depend greatly on dispersion and percolation limit [140,148]. Increasing the carbonization temperature from 850 °C to 1100 °C led to an order of magnitude increase in in-plane conductivity, from 3.2 to 43.3 S cm-1. Because the in-plane electrical conductivity surpassed the minimal requirement desired for the current design, the addition of CNT was not a necessary step for the optimized fabrication process.   68     Figure 3-8: Design factor cube and interactions graph for electrical conductivity, interaction recorded at polymer concentration = 10 wt% and stabilization temperature = 230 °C 3.3.3 Effect of process parameters on fiber mesh porosity  The main effects of mesh porosity were computed using the Design-Expert software. Verifying the model using ANOVA method, it was found that the significant coded model terms were D and interaction BC. This means the main effect was CNT concentration and there was an interaction between stabilization/carbonization temperatures. As can be seen in the design factor 69  cube, the porosity was increased with increasing CNT concentration (Figure 3-9). Interactions curve demonstrates that there was a strong interaction between BC because the C-1 and C+1 intersects. Again, this signifies that the two effects were co-dependent. The observed trend, related to CNT concentration, may be explained by the electrospinning process theory. The electrospinning process is governed by a variety of forces (viscoelastic and electrostatic force) [44]. The addition of CNT increases both the solution viscosity and electrical conductivity. The increase in solution electrical conductivity was shown to decrease fiber diameter and improve spinnability [45]. Moreover, increasing the electrical conductivity of the solution has an effect on the formation of the Taylor cone and straight jet length [45]. Decreasing the length of the straight jet portion allows for increased instability portion, which is responsible for the whipping and reduction of fiber diameter. Thus, altering the surface charge and the tangential electric field along the fluid surface resulted in lower fiber diameters leading to higher mat porosity. In summary, the porosity met the design targets for catalyst support layer. However, the porosity used in this model was estimated by calculating the surface area, a physical characterization method was used later to verify the assumptions (Chapter 4.2). A graphical representation of the porosity can be found in the appendix (Figure C-4).  70    Figure 3-9: Design factor cube recorded at 10 wt% polymer and interactions graph for % porosity; carbonization temperature setting: black square = 850 °C and red triangle = 1100 °C  3.3.4 Effect of process parameters on fiber young’s modulus   Main effects related to Young’s modulus was determined using Design-Expert software. Verifying the model using ANOVA method, it was found that the significant coded model terms were A, D, AB, and BCD. Thus, the main effects were polymer and CNT concentration. The response was also influenced by the interaction between polymer concentration/stabilization 71  temperature and stabilization temperature/carbonization temperature/CNT concentration. Interactions curve demonstrates that there was a strong interaction between AB since the B-1 and B+1 intersects. Again, this means that the effect of one factor depends on the level of the other. From the graphical representation of the Young’s moduli recorded for the experiments, the range was very broad, from approximately 16 GPa to 3 GPa (appendix Figure C-5). Although clear trends cannot be observed in the data set, it appears that modulus increased with decreasing polymer concentration while increasing CNT decreased the modulus. This observation might be explained by the poor CNT dispersion in the nanofiber, as previously mentioned. It was observed in prior work that poor electrical properties caused by insufficient dispersion of CNT, also led to poor mechanical properties [148]. Essentially, the agglomerated CNT forms stress concentrations inside the fiber causing it to break prematurely. Because lower polymer concentration typically produces smaller diameter fibers, it was expected that the fibers would have enhanced mechanical properties [88]. Briefly, the classical Weibull size effect theory explains the probability of having a small defect would decrease with decreasing cross sectional area [149]. Because the defects are smaller and fewer, fracture propagation caused by stress concentration decreases, yielding enhanced mechanical properties. To conclude, the modulus achieved for most of the samples were satisfactory for the current application, revealing that CNF is a mechanically robust and suitable material for this application.     72    Figure 3-10: Design factor cube recorded at 10 wt% polymer and interactions graph for elastic modulus; stabilization temperature setting: black square = 230 °C and red triangle = 260 ° 3.3.5 Development and validation of factor models  To model and predict the optimal processing parameters for fabrication of CNF, the experimental responses were developed into a first order function. The response equation is a function of the main effects and interactions, as mentioned in previous sections Equation (3.1). Representing the equation in terms of coded factors instead of actual values simply shows the 73  magnitude of each effect. To verify the significance of the model, key statistical values were studied. P-value, which measures the models statistical significance, should be less than the significance level of 0.05. As shown in Table 3-5, the P-value for all 4 responses satisfied this requirement, validating adequacy of these models. R2 is a measure of the percent of the responses being represented by the variables. Table 3-5 shows R2 and adjusted R2 for the four responses. Although higher values would be desirable, the models show adequate signal as measured by the signal to noise ratio – adequate precision. All modeled equations demonstrate a ratio greater than 4, the desired minimum ratio. The model predicts variabilities (adjusted R2) of 82%, 91%, 66% and 82% in new data for diameter, conductivity, porosity, and modulus, respectively. Ideally, the difference between predicted R2 and adjusted R2 should be less than 0.2 [150]. Larger differences, shown in the diameter and modulus models, may indicate a large block effect or issues with the input data and/or model (Table 3-5). Additional statistical fit parameters are presented in the appendix (Table C-2). 𝑦𝐷 =  242 +  18𝐴 + 3𝐵 − 16.75𝐶 + 10.12𝐷 + 3𝐴𝐵 − 3.25𝐴𝐶 + 27.63𝐴𝐷 + 25𝐵𝐶 + 22𝐴𝐵𝐶 (3.1) ln 𝑦𝐶 =  2.3 + 0.96𝐶 − 0.37𝐷 𝑦𝑃 = 80.94 + 0.31𝐵 + 0.19𝐶 + 4.19𝐷 − 2.19𝐵𝐶 𝑙𝑛 𝑦𝑀 = 1.75 − 0.22𝐴 − 0.13𝐵 +  0.046𝐶 − 0.14𝐷 + 0.22𝐴𝐵 − 0.13𝐵𝐶 − 0.13𝐵𝐷 − 0.020𝐶𝐷 + 0.26𝐵𝐶𝐷 Response polynomial equation in coded factors, for yD diameter, yC conductivity, yP porosity, and yM Young’s modulus Table 3-5: Significance probability (P-value) and correlation coefficient of linear regression for response surface equations Response P-value R2 Adjusted R2 Predicted R2 Adequate Precision Diameter 0.0085 0.9273 0.8181 0.4827 11.345 Conductivity < 0.0001 0.9260 0.9146 0.8878 19.036 Porosity 0.0023 0.7532 0.6635 0.4779 7.315 Modulus 0.0086 0.9270 0.8175 0.2016 10.173 Since the model equations show statistical significance, it is suitable for navigating the design space. To find the optimal processing parameters for this set of equations, a set of targets 74  and constraints on the variables were implemented. The highest priority was to minimize the fiber diameter, the experimental results showed a range between 159 and 372 nm. Modulus and conductivity were medium priority thus, achieving a value between 5 and 52 S cm-1 was set as the target. Young’s modulus was optimized in the same manner, targeting a modulus between 5 – 16 GPa. Lastly, since nanofiber mats are well known for large porosities, the optimization for porosity was set as the lowest priority [151]. The target was to maximize the porosity, which was between 71 and 95% from experimental results. A summary of the values and constraint used for optimization is presented in Table 3-6.  A list of solutions were generated by the software with the best solution having a desirability factor of 0.883. A table showing all possible solutions is presented in the appendix (Table C-3). From the list of predictions, the best possible fabrication process is shown in Table 3-7. From the results, the prediction removes agglomerated CNT from the process. This is plausible since non-uniformly dispersed CNT caused negative effects, such as decreased mechanical strength and electrical properties. As expected, the model predicts a high carbonization temperature (1100 °C) since this was the main effect for electrical conductivity. An example of the contour plots for predicted diameter is illustrated in Figure 3-11 . Although the plot indicates that smaller diameter could be achieved in the blue region, it is not a feasible solution since stabilization occurs at higher temperatures. Sufficient stabilization is necessary to ensure the chemical structure is well aligned in preparation for carbonization step [119]. The red labels on the plot show the design points which were part of the 16 experiments. Lastly, the optimal predicted solution, producing a diameter of 149 nm, was revealed to be 10 wt% PAN-co-MA, 230 °C stabilization, 1100 °C carbonization, and no addition of CNT. Using the optimized fabrication parameters, CNF were prepared and characterized again to verify the RSM solution (Figure 3-12).  75  Table 3-6: Optimization targets and constraints for each response and factor Name Goal Limit Limit Weight Weight Importance A:Polymer concentration  is in range  7 10 1 1 3 B:Stabilization T  is in range  230 260 1 1 3 C:Carbonization T  is in range  850 1100 1 1 3 D:CNT concentration  minimize  0 1 1 1 3 Diameter  minimize  159 372 1 1 5 Conductivity  is in range  5 52.28 1 1 3 Porosity  maximize  71 95 1 1 1 Modulus  is in range  5 16.558 1 1 3 Table 3-7: Best predicted solution for optimized fabrication process Polymer Concentration (wt %) 10.00 Stabilization T (°C) 230 Carbonization T (°C) 1100 CNT Concentration (wt %) 0.00 Diameter (nm) 149.25 In-plane conductivity (S cm-1) 37.96 Porosity (%) 78.81 Modulus (GPa) 6.79 Desirability  0.88  Figure 3-11: Contour plot of predicted values for fiber diameter; optimal solution of 149 nm at carbonization temperature of 1100 °C and 0 wt% CNT 76   Table 3-8: Material properties of CNF fabricated using predicted optimal parameters Fiber diameter (nm) 333 ± 26 In-plane conductivity (S cm-1) 31.94 ± 2.95 Porosity (%) (measured) 79.82  Elastic modulus (GPa) 18.76 ± 1.22 Ultimate Tensile Strength (MPa) 72.69 ± 6.67 Through-plane conductivity (S cm-1) 0.37 ± 0.86 Compression strength (MPa) > 3MPa   Figure 3-12: CNF morphology from optimized fabrication parameters (2 kV, secondary electron mode) Although the validation experimental results reported in Table 3-8 differed from predicted model, the majority of the design targets were met. The assumption is that the trends and effects predicted by the RSM model are valid and can be taken as guidance for future investigation of CNF fabrication. It is important to note that process optimization by RSM is an iterative procedure [152]. To reduce variance and improve confidence levels, the DOE should be repeated using the predicted values from the initial optimization. Moreover, in the current study there were no repeats for each run, one sample was fabricated and tested for each of the 16 experiments. To improve the dataset, an average of three samples per run is recommended for future optimizations. Deviation from the model may be due to some outliers in the data or a block effect. Blocking essentially 77  happens when a variability cannot be overcome (e.g. needing two batches of raw material to produce one container of a chemical). Some of the variation in the data may be due to different batch spinning dopes or variations in the environment over several days. It is unclear whether temperature and relative humidity affects the electrospinning process [153]. The samples were prepared on separate occasions with maintenance on the tube furnace, thus the reproducibility of the data may be affected. Initially, tensile specimen were cut into strips after the carbonization step, this may have induce defects as the specimen are brittle after carbonization. The handleability may explain the large deviation between the 16 specimens. Subsequent tensile specimen, the optimized batch, were cut before the carbonization step to minimize the amount of defects, lowering the SD and increasing modulus (Table 3-8). Significant effort was made to perfect the sample preparation process, however variation between each specimen over the course of the 16 runs is plausible and probable. Outliers were discarded, however, the reported standard deviations across the samples should be considered (see appendix).   Variability in the results may also arise from the measurement methods used. For instance, the variance was relatively large for fiber diameter measurements (Figure C-2). The deviation was mostly a result of manually measuring the diameter using visual methods and Image J. Mechanical properties of the fibers also showed large deviations due to the nature of sample preparation. Because CNF were brittle and formed micro defects during sample preparation, the measured modulus and ultimate tensile strength revealed significant variance (Figure C-5). Experimental error caused by inaccurate measurement systems would decrease the significance of the predictors. As a result, the variability and errors from the experimental runs were transmitted to the RSM model. Ultimately, the predicted trend from the model was still applicable and satisfied design criteria. Except for fiber diameter, all responses showed accurate prediction within the 78  experimental error. In order to improve future models, reiteration of the optimization using additional sets of data would be required to refine the predictions and narrow the deviation. Presently, it is sufficient to demonstrate that the electrospinning and heat treatment process can produce CNF that possesses the material properties desired for MEA fabrication. Assembly of the MEA with the optimized CNF is discussed in Chapter 6.    Summary  Electrospinning method was used to fabricate non-woven nanofibrous layers for the application of fuel cells. As fuel cell catalyst supports require high surface area, porosity, durability, and electrical conductivity, a factorial design was implemented to optimize the material properties. Because of the requirement for high electrical conductivity, PAN and PAN-co-MA precursor were selected as possible candidates to produce carbon nanofiber. TGA result suggested that PAN-co-MA was favorable for high temperature carbonization and was selected for the factorial design. A factorial design using four factors and four responses was conducted to optimize the fabrication of carbon nanofibers. The four factors selected (polymer concentration, CNT concentration, stabilization temperature, and carbonization temperature) were previously shown to influence key fiber properties, such as, fiber diameter, porosity, electrical conductivity, and mechanical strength. From the 16 experiments conducted, a polynomial model was fitted to predict the most desirable fabrication parameters. The optimized variables were: 10 wt% PAN-co-MA, 0 wt% CNT, 230 °C stabilization temperature, and 1100 °C carbonization temperature. The predicted responses for this model were: ~150 nm diameter, ~38 S cm-1 in-plane conductivity, ~78% porosity, and ~6.8 GPa Young’s modulus. However, in validation experimental results, the properties deviated from the model, although still satisfying the overall design criteria of the CNF 79  layer. The current study has demonstrated the feasibility of using electrospinning to fabricate catalyst supports that met industry design targets.  80   Comparison Between RNL and OAN Properties   Introduction  The conventional electrospinning method was modified to spin unidirectional fibers with the same optimized processing parameters established in the previous chapter. The electrospinning method used two pairs of copper plates and two high voltage power sources. A schematic of the modified electrospinning apparatus is shown in Figure 2-2. OAN were fully characterized to investigate the effect of orientation on fiber diameter, layer thickness, porosity, electrical conductivity, and surface chemistry. Material properties were evaluated and compared against RNL to ensure design criteria were satisfactorily met. Surface chemistry was verified to demonstrate carbon content and surface groups were the same as RNL post-heat treatment. The majority of design targets were met, however OAN had inherent weaknesses due to the modified electrospinning process.    Structural and mechanical properties  The optimized parameters generated from the RSM were used to fabricate RNL and OAN samples. For the remainder of the research work RNL and OAN carbonization temperature was lowered to 900 °C instead of 1100 °C because samples carbonized at 900 °C showed better handleability while maintaining sufficient electrical properties. RNL was fabricated using traditional electrospinning method, while OAN was fabricated using modified electrospinning set-up. Since the electrospinning method was modified, the fibers were characterized to ensure they satisfy the design targets for MEA application. Fiber diameter distribution is shown in Figure 4-1 and Figure 4-2 for RNL and OAN CNF, respectively. The average fiber diameter for RNL after carbonization was 333 ± 26 nm. Although the fiber diameter did not meet the target requirement of > 200 nm, the RNL diameter was deemed appropriate for the application because CB 81  agglomerates may span up to 5 µm depending on the mixture and processing of the catalyst. Conversely, OAN met the design target, achieving a fiber diameter of 248 ± 46 nm after carbonization. The parallel electrode set-up may have induced additional thinning because the fibers were suspended across a 7 cm gap [154]. Although negligible, OAN fibers may sag under its own weight compared to RNL fibers which were collected on an aluminum foil. More probable, the OAN fibers may experience more fiber shrinkage compared to RNL because of the more fragmented structure. Because RNL fibers were entangled and formed a cohesive and well-packed non-woven fabric, the random network may have prevented further shrinkage. In OAN, the fibers are separated and aligned in one direction, thus the applied tension would cause stretching in one direction resulting in further reduced diameter (Figure 4-3). Furthermore, OAN and RNL likely consist of different crystal lamellar structure because of the modified electrospinning method and solvent evaporation process. As a result, OAN and RNL would behave differently during heat treatment.       Figure 4-1: RNL fiber diameter distribution using optimized fabrication process (CNF) 82   Figure 4-2: OAN fiber diameter distribution using optimized fabrication process (CNF)  Figure 4-3: Schematic of OAN fiber organization and through pores Table 4-1: Porosity measurement by densometry of random and aligned CNF after Platinum and Ionomer deposition Random CNF (%) 79.82 Pt Random CNF 87.39 Ionomer/Pt Random CNF 74.29 Orthogonal 73.19 Pt Orthogonal 67.40 Ionomer/Pt Orthogonal 64.27 It was hypothesized that OAN would have higher porosity since the fiber alignment should produce through pores with large spacing (Figure 4-3). The porosity measurements were made using a densometer which utilizes Archimedes principle. The porosity of random nanofibers after Pt and ionomer deposition were roughly 74% porous, while the orthogonally aligned fibers were 83  64% porous. Though the porosity decreased, it still remains comparable to the porosity of carbon black based catalyst layers. The conventional agglomerated catalyst layer is between 30 – 60% porosity [155–157]. The porosity did not change significantly with the application of ionomer and Pt because the coating was very thin and did not block secondary pores (Table 4-1). The decrease in porosity for OAN may be explained by the overlapping and adhesion between unidirectional fibers shown in Figure 4-4. It is also possible that the submersion liquid did not thoroughly permeate the deeper pores during the densometry measurements because of lower wettability, resulting in lower porosity values. In comparison, the calculated % porosity using Equation (2.4), was 80% ± 5.05% for RNL, revealing reasonable agreement to the densometry porosity values.  Although previous studies have shown well organized orthogonal structures, the fiber diameter (1 µm) and pore spacing (10 µm) were much larger than the present work [158]. Moreover, the fiber mat geometrical area in prior work is typically very small (2 – 3 cm wide) because of the limitations of the parallel electrodes and fiber sagging across larger distances [112,113]. The main challenge for orthogonal nanofiber mats is maintaining the degree of perpendicularity due to the weakened electrical field with increasing gap distance and layer thickness [159,160].  OAN mats were electrospun until a suitable thickness was achieved, this involved stacking orthogonal layers manually ~40 times to reach a desired thickness. To the extent of the author’s knowledge, the mat thickness achieved in this study has not been realized in prior work. Fiber mesh thickness was measured by SEM to ensure consistency and uniformity. Carbonized layer thickness was measured by micrometer as a first approximation. To verify the values from physical method, MEA were embedded into resin for cross sectional imaging. The cross section image shows the CNF thickness, roughly 9.0 ± 0.9 µm and ~9.7 ± 1.2 µm for OAN and RNL samples respectively (Figure 4-5). It was important to maintain low and consistent thickness during 84  fabrication because it is related to the diffusivity which plays a key role in water management and fuel transport. A summary of the structural properties for OAN and RNL is presented in Table 4-2.    Figure 4-4: Orthogonally aligned carbon nanofiber mat fabricated by modified electrospinning method (SEM: 5 kV Secondary electron mode) Carbonized OAN and RNL samples were tested on the KES-G1 Kawabata microtensile testing system to evaluate the mechanical strength. Because OAN was carbonized at 900 °C, an additional set of RNL carbonized at the same temperature was tested to have direct comparison. The mechanical properties averaged over five samples is illustrated in Figure 4-6. The ultimate tensile strength (3.77 ± 1.71 MPa) was significantly lower compared to RNL (67.15 ± 5.16 MPa) samples. The same trend can be seen for the Young’s modulus where OAN and RNL achieved 1.16 ± 0.65 GPa and 16.18 ± 1.88 GPa, respectively. The OAN results reveal some scatter because of the inherent structure and fabrication method of the orthogonal fibers. The poor mechanical properties of the OAN were largely due to the orientation of the fibers. Since they were organized 85  in an orthogonal manner, approximately half of the fibers carried the load in the tensile axis. This drastically lowers the calculated tensile strength because the orthogonal fibers contribute to the cross sectional area as well. Another factor was the delamination during sample preparation. As explained in a prior section, because the orthogonal fiber layers were stacked consecutively instead of continuously drawn, the fibers lack cohesion through thickness. The residual static charge on the fibers caused the layers to peel off or wrinkle on the surface, causing variability in the measurements. Nevertheless, OAN showed reliable mechanical properties in compression, which is the main mechanical loading in a fuel cell stack. OAN did not crack or rip during the resistance under compression measurements (> 4 MPa).  Figure 4-5: Orthogonally aligned (A and C) and random nanofiber (B and D) cross sectional thickness (15 kV, secondary electron mode)  86  Table 4-2: Summary of OAN and RNL structural dimensions measured by SEM and densometry  Layer thickness (µm) Diameter (nm) Porosity (%) Porosity after I/Pt (%) Orthogonally aligned nanofiber 9.0 ± 0.9 248 ± 46 73 64 Random nanofiber 9.7 ± 1.2 333 ± 26 79 74   Figure 4-6: Ultimate tensile strength (left) and Young’s modulus (right) for OAN and RNL (average of 5 samples carbonized at 900 °C)  Through-plane and in-plane electrical properties  Table 4-3: Summary of electrical properties for OAN vs. RNL Sample  In-plane conductivity (S cm-1) Through-plane conductivity (S cm-1) Target > 5 > 0.05 RNL (1100 °C) 31.94 ± 2.95 0.453 ± 0.193 RNL (900 °C) 6.734 ± 0.320 0.266 ± 0.040 OAN (900 °C) 1.535 ± 0.140 0.13 ± 4.20 x 10-3 RNL (700 °C) 0.202 ± 0.021 4.41 x 10-3 ± 2.18 x 10-3 Electrical properties for OAN were measured by the same method as previously discussed. The in-plane and through-plane resistance were measured at increasing carbonization temperatures for both OAN and RNL samples (Table 4-3). The effect of graphitization and carbonization temperature can be seen in both in-plane and through-plane measurements. RNL carbonized at 1100 °C achieved an in-plane electrical conductivity of 32 S cm-1. Samples carbonized at 700 °C 67.153.770.00 20.00 40.00 60.00 80.00RNLOANUltimate Tensile Strength (MPa)16.181.160.00 5.00 10.00 15.00 20.00RNLOANYoung's Modulus (GPa)87  did not achieve the minimum target requirement of > 5 S cm-1. Similarly for through-plane resistance, the RNL sample carbonized at 700 °C did not meet the target requirement of > 0.05 S cm-1. OAN in-plane electrical conductivity was measured at 0° to the fiber axis. OAN showed weak electrical properties (1.5 S cm-1 and 0.13 S cm-1) compared to RNL. A reason for this could be because the contact resistance between each fiber layer. Since the RNL was fabricated continuously, the fibers were deposited overlapping each other and allowed to dry in a compressed state. OAN, on the other hand, were electrospun as separate orthogonal layers. The layers underneath had time to dry and was not easily adhering to the next layer being deposited. Therefore, there was a lot of resistance between each stacked layer, resulting in higher through-plane resistance (threefold increase between 900 °C RNL and OAN) (Table 4-3). In-plane conductivity was also significantly poorer than RNL because only the fibers parallel to the measurement direction was conducting electrical current. It should also be noted that the OAN fiber mat was delaminating in some areas, thus increasing contact resistance throughout the layers. Although OAN samples met the design requirement for through-plane resistance, the effect of orientation and spinning method is apparent in the present results.  Surface carbon content via x-ray photoelectron spectroscopy   The analysis of the surface chemistry was performed by XPS to determine the possible functional groups. Functional groups are often used as nucleation sites for Pt deposition. It is also a good indication of what types of carbon are on the surface; amorphous or graphitic. From the C 1s scan, the peak was calibrated at a binding energy of 284 eV. The convoluted peaks are resolved using XPS Peak software. The deconvoluted peaks revealed the presence of graphitic and disordered carbon (Table 4-4). Furthermore, the π-π satellite signal represents the graphitic carbon, demonstrating adequate amount of graphitic carbon on the fiber surface [161]. Other carbon 88  species include: C−OH, C−O−C, C−N, C=O, and COOH. These maybe residual species remaining after heat treatment or contaminants absorbed from the atmosphere.  Calculating the area under the peaks, the atomic % of each species is found. The results reveal CNF carbonized at 1100 °C possess ~70% graphitic carbon and 12% disordered carbon on the surface. This corresponds to a graphitic to disordered ratio of 5.8. XPS results suggest that abundant graphitic structures were present, indicating the chosen heat treatment procedure was sufficient. Considering the carbonization temperature used was not particularly high compared to other published results or industry standards, the graphitic yield is encouraging. Since the graphitic portion is responsible for high electrical conductivity and mechanical stability, it is desirable for the CNF support. CNF carbonized at 900 °C was also examined to compare the effect of carbonization temperature on graphitic content. As expected, ratio between graphitic to disordered carbon decrease to 3.2 for samples carbonized at 900 °C (Table 4-5). However, the results also reveal there is little oxygen containing functional groups which would assist Pt deposition. Nonetheless, it is possible for atomic size defects, such as graphene edges, to act as nucleation sites during Pt deposition [162]. Subsequently, surface chemistry will be analyzed after chemical functionalization and Pt deposition to observe chemical bonding and chemical state of Pt in Chapter 5.2.2. Because samples carbonized at 1100 °C showed higher graphitic content, it was considered more inert to chemical treatments or Pt deposition. Hence, the carbonization temperature was lowered to 900 °C for results presented in Chapter 5 and 6.     89  Table 4-4: 1100 °C carbonized 10 wt% PAN-co-MA deconvoluted XPS peaks for C 1s, pass energy 20 eV Peak RNL 1100 °C Position (eV) Area Component % 0 C1: C=C graphitic 284.45 11656.02 71.22 1 C2: C-C  aliphatic/dis-ordered 285 2026.402 12.38 2 C3: C-OH + C-O-C + C-N species 286.5 787.223 4.81 3 C4: C=O 287.7 563.494 3.44 4 C5: COOH 289.2 241.638 1.48 5 C6: π-π* shake up 290.7 1092.438 6.67  sum  16367.22 100 Table 4-5: 900 °C carbonized 10 wt% PAN-co-MA deconvoluted XPS peaks for C 1s, pass energy 20 eV Peak RNL 900 °C Position (eV) Area Component % 0 C1: C=C graphitic 284.5 23116.53 63.79 1 C2: C-C  aliphatic/dis-ordered 285.05 7306.078 20.16 2 C3: C-OH + C-O-C + C-N species 286.5 2261.594 6.24 3 C4: C=O 287.6 1374.142 3.79 4 C5: COOH 289.2 782.566 2.16 5 C6: π-π * shake up 290.7 1396.677 3.85  sum  36237.59 100 Since OAN was fabricated using a modified electrospinning set-up the chemical structure and composition may be altered. XPS was conducted to evaluate the chemical bonding and graphitization of OAN samples carbonized at 900 °C. Although the OAN was heat treated under the same conditions, XPS revealed differing results from RNL. Figure 4-7 shows the composition of the carbon species from the deconvoluted Carbon 1s peaks. Similar to the RNL peaks, there were 6 components detected. Again, the characteristic π –π* satellites indicate the presence of sp2 aromatic carbon. The individual carbon species on the OAN surface are labeled in the graphic, with the most notable results being the C=C and C-C components. Compared to the RNL samples, OAN comprises of notably less graphitic carbon (Figure 4-8). The graphitic to disordered carbon ratio is 1.33 and 3.16 for the OAN and RNL, respectively. It is plausible that the decrease in graphitic content is due to the electrospinning method rather than the carbonization process, since both samples are heat treated under the same conditions. Because OAN is electrospun using a desk 90  top set-up, the humidity and environment cannot be readily controlled [163]. Relative humidity levels has been shown to affect the nanofiber morphology, roughness, and porosity [163]. The difference between the two electrospinning methods may be the reason for dissimilar surface groups. The drying time and solvent evaporation is different in the OAN samples as a result of having to manually stack the fiber layers in an orthogonal pattern.   Figure 4-7: Example of deconvoluted C 1s XPS spectrum for OAN (10 wt% PAN-co-MA carbonized at 900 °C), calibrated at 284.6 eV, pass energy: 20 eV The RNL method can continuously deposit fiber and form a cohesive non-woven layer while the solvent is evaporating. RNL also experiences whipping motion during deposition whereas OAN experience little whipping and are aligned across two copper plates. The whipping motion from electrospinning results in thinner fibers and macromolecular chain alignment within 280285290295Intensity (a.u)Binding Energy (eV)C 1s deconvoluted (OAN) Raw Intensity Peak SumBackground C=C graphiticC-C  disordered C-OH + C-O-C + C-N speciesC=O COOHPi-Pi* shake up91  the fiber [164–166]. Additionally, the quality of carbon fibers reportedly increased by stretching (molecular alignment), which improved the formation of the ladder structure during stabilization [167]. Thus, the crystallinity and molecular alignment was possibly the reason for OAN and RNL disparity. Because of the lower graphitic content, the electrical properties in OAN would also be deficient. Although there was a large variance between the amount of disordered and graphitic carbon in the OAN and RNL samples, the remaining carbon-oxygen groups appear to be similar.  Figure 4-8: OAN and RNL (carbonized at 900 °C) surface chemistry by deconvoluted C 1s XPS spectrum, pass energy 20 eV  Summary  Novel orthogonally aligned carbon nanofibers were successfully fabricated using a modified electrospinning method. Following the same post processing method as RNL, the OAN were fully characterized to ensure similar properties were obtained. Fiber diameter, porosity, layer thickness, surface chemistry, through-plane electrical resistance, and in-plane electrical conductivity were compared to RNL. OAN CNF were carbonized at a lower temperature of 900 °C, following RNL fabrication, to ensure good compromise between mechanical and electrical properties. The measured fiber diameter was 248 ± 46 nm, compared to 333 ± 26 for RNL. Porosity for the OAN layer was 73% and 64% before and after I/Pt deposition, respectively. The porosity 010203040506070Graphitic Disordered C-OH + C-O-C + C-NC=O COOH Pi-Pi*Component %RNL 900 OAN 90092  for RNL was slightly higher than OAN, 79% and 74% porosity, respectively. Although OAN was expected to have through-thickness pores and organized structure, the electrospinning process caused some fibers to adhere together rather than spreading out. The layer thickness for OAN and RNL were largely similar, approximately 9 µm. This is comparable to conventional catalyst supports using carbon black [168]. Therefore, all structural and dimensional requirements for the catalyst design were met. Tensile test revealed that RNL (67 ± 5 MPa) possessed significantly higher tensile strength compared to OAN (3.77 ± 1.71 MPa). The lower strength in OAN was largely due to less fibers in the loading axis direction—unable to carry the load. The inherent stacking of the OAN resulted in higher electrical resistances through-plane compared to RNL samples. The same challenge was observed in the in-plane direction because fibers perpendicular to the probing axis carry little current or have large resistances because of the inefficient path. Although the in-plane conductivity and through-plane conductivity for OAN was 1.5 S cm-1 and 0.13 S cm-1, respectively, the electrical properties remain adequate for the current catalyst support design. Surface chemistry analysis by XPS also confirmed the decrease in graphitic content and increase in disordered carbon, leading to less conductive fibers. Although the graphitic to disordered ratio decreased, the OAN possessed similar amount of remaining carbon-oxygen species to those of RNL samples. The deconvoluted XPS results for RNL demonstrated higher graphitic content in 1100 °C carbonized sample compared to 900 °C carbonized sample. The higher graphitic content is beneficial for mechanical and electrical properties, however the graphitic structure may also inhibit Pt deposition. In Chapter 5.2.2, XPS reveals whether Pt preferentially bonds to functionalized CNF or original CNF. Ultimately, material characterization revealed that OAN met the key design criterion. Based on the material evaluations discussed, it appears RNL is a more suitable candidate for MEA application because of the superior electrical 93  properties and ease of controlling microstructure. The study concludes RNL and OAN processing parameters cannot be directly translated due to the difference in electrospinning method which yields dissimilar as-spun fiber properties (crystallinity, cohesion, porosity, etc.). The next chapter discusses the Pt plating and ionomer deposition process used for both OAN and RNL.   94   Platinum and Ionomer Deposition on CNF   Introduction Using the optimized parameters discussed in Chapter 4, the CNF were prepared for Pt deposition. Several deposition methods were explored to determine the simplest and effective process. Because CNF is not traditionally used as a catalyst support, not much information has been reported about Pt deposition methods. The main challenge was selecting a deposition process that yields a homogeneous Pt coating with small particle size and reduces the handling to prevent the thin CNF layers from tearing. Because Pt is the most expensive component of the MEA, it is important to utilize a method that would deposit a thin layer, adequate for achieving high catalytic surface area. Some of the deposition results discussed here include atomic layer deposition, electrodeposition methods using different reducing agents, and hydrogen reduction. Most of the methods were deemed unfeasible because samples were too delicate or the Pt was not uniformly distributed on the fiber surface. Electroless plating using formic acid as reducing agent was favorable because it was a facile method and produced nanowire-like clusters which exhibit high surface area. Pt loading was controlled and different particle size was also observed by changing concentration of the plating bath. Chemical analysis revealed the deposited Pt’s chemical state, which was consistent with Pt 4f XPS patterns reported in the literature. Next, ex situ electrochemical characterization via cyclic voltammetry (CV) was conducted to determine the effective platinum surface area (ECSA). The CV results reveal that the ECSA does not follow a linear relationship with increasing Pt concentration. Additionally, the ex situ testing revealed the presence of catalytically active facets which were consistent with the literature. Ionomer was also deposited to determine how the Pt would interact with ionomer. From the cross sectional STEM/EDX maps, it appears that the ionomer adhered well to CNF/Pt and formed a thin layer 95  around the entire fiber. The ionomer was shown to form a connected network between fibers, allowing the protons to travel efficiently through-plane. Optimization of the Pt loading and ionomer loading was necessary to obtain feasible catalyst supports for MEA fabrication and in situ FC testing.   Evaluation of Pt deposition process  Hydrogen Reduction  Different types of pre-treatment were investigated to form functional groups on the CNF surface as described in Chapter 2.4. After soaking in the Pt bath overnight, the bath solution completely evaporated and the dried sample was reduced in hydrogen furnace using the conditions explained in Chapter 2.4. Figure 5-1A shows small Pt particles were deposited on the CNF surface. The Pt formed a homogenous dense layer around the fiber and was distributed across the surface of the mat. Although the size of the Pt particle was desirable and the Pt loading was 115 µg cm-2, the Pt was not consistently distributed within the thickness.  SEM images of early samples revealed that the morphology was inconsistent and often does not produce the desired Pt loading. From Figure 5-1A, The CNF was pre-treated in citric acid and (3-Aminopropyl) triethoxysilane, soaked in a H2PtCl6 acetone bath, and reduced in H2 furnace. The method formed a thick coating with little or no exposed CNF, this may create a challenge for the formation of a TPB.   96  A B C H2 reduction; H2PtCl6 acetone evaporation; Citric acid and (3-Aminopropyl) triethoxysilane treated (115 µgPt cm-2) Atomic Layer Deposition on bare CNF Acetic acid method on stabilized fiber (45 µgPt cm-2)    D E F Acetic acid method on CNF (3.4 µgPt cm-2) Formic acid method; MeOH pre-treatment Formic acid method, pre-treatment w/ (3-Aminopropyl) triethoxysilane and MeOH    G H I Formic acid method; Triton X surfactant  (132 µgPt cm-2 loading) Formic acid method, (3-Aminopropyl) triethoxysilane pre-treatment for 2 hrs.; Triton X surfactant (407 µgPt cm-2) Formic acid method, pre-treatment for 2 hrs.; triton surfactant (51 µgPt cm-2)    Figure 5-1: Pt deposition by various methods (5 and 10 kV, secondary electron mode)  97  Acetic Acid Method Because the acetic method requires solution to be constantly stirring or boiling, the delicate CNF could easily crack, therefore stabilized fibers were also explored. The stabilized fiber achieved higher Pt loading than the CNF, 45 µg cm-2 and 3 µg cm-2 respectively (Figure 5-1C and D). However, the acetic acid method proved to be ineffective and cumbersome since stabilized fibers would still need to be carbonized at a later step to increase the electrical conductivity.  ALD Pt deposition  SEM images of CNF coated with ALD revealed poor deposition. The Pt was either too thin, depositing on an atomic level, or the CNF surface was overly inert for this method (Figure 5-1B). XPS of this sample also did not reveal Pt species on the surface. ALD may have deposited an atomic layer of Pt which is not enough for the fuel cell to function effectively. To achieve a thicker layer of Pt using ALD, the deposition cycles would need to be increased significantly, causing the process to be economically unfeasible. Formic Acid method  Pre-treating the CNF with methanol (MeOH) and depositing Pt using formic acid method, the Pt particle were evenly distributed on the fiber surface and dispersed evenly to create small islands Figure 5-1E. The morphology shown was promising and desirable, however, the MeOH treatment was not effective and produced inconsistent Pt loading. Similarly, Figure 5-1F reveals large nanowire clusters formed from the formic acid method. The Pt did not distribute homogenously on the fiber surface using MeOH and (3-Aminopropyl) triethoxysilane pre-treatment, large regions were bare. Lastly, Triton X and PVP surfactant was added to promote nanowire growth in one direction and to also prevent precipitation in solution [169]. In Figure 5-1G, a good dispersion of 98  small Pt particles were obtained using the formic acid method with addition of 0.1 mL Triton X. The method yielded 130 µg cm-2, which is near the targeted Pt loading. The addition of a pre-treatment step did not improve the Pt morphology or yielding, in fact, the 0.5 mL Triton X was extremely difficult to remove and can be seen as thick layers in the SEM images (Figure 5-1H).  Figure 5-1I shows a similar sample using 0.1 mL Triton X with inconsistent coating. Fibers beneath the surface had uneven or sparse coating of Pt. Because surfactants would be considered impurity in the MEA and can hinder the performance, a small amount of PVP was used for later optimized samples. Evaluating all the methods and challenges associated with each, the final procedure for Pt deposition includes: pre-treatment in 3M HNO3, followed by formic acid reduction method with addition of 2 mg of PVP in the plating bath. Results of the developed plating method can be seen in Figure 5-2. Nanowire were grown in small circular clusters with homogenous distribution and controllable Pt loading.     5.2.1 SEM/TEM evaluation of Pt nanowire coating  Low, medium, and high Pt loading samples were examined in SEM.  As can be seen from the SEM images, the Pt clusters were evenly and homogeneously distributed through the thickness of the CNF layer (Figure 5-2). The particles were densely dispersed along the fiber, yielding numerous catalyst reaction sites.  Increasing the Pt concentration yielded larger clusters and more densely packed Pt. High magnification images reveal that the particles appear to be a nanowire like structure arranged into a sphere (Figure 5-3 and Figure 5-4). This is advantageous since the nanowire structure increases the specific surface area, ultimately improving the catalyst reaction sites. The average cluster size measured by Image J was roughly 10, 25, and 70 nm, for low, medium, and high concentration, respectively.  99  low    med    high    Figure 5-2: SEM micrograph of electroless plated Pt on CNF (top to bottom: low, medium, and high concentration Pt precursor) (1 kV, secondary electron mode)  Figure 5-3: SEM image showing dense Pt nanowire clusters on CNF surface (120kx, 1 kV, secondary electron mode) TEM observation confirmed the presence of Pt nanowires with a preferential growth direction. Nanowire length and diameter is roughly 10 – 15 nm and 3 – 5 nm, as observed from 100  TEM micrographs (Figure 5-4). The nanowires protrude randomly from spherical clusters on the CNF surface. The crystallographic spacing shown in the TEM reveals a d-spacing of 0.23 nm which corresponds to a single crystal direction of <111> as seen in XRD peaks [141]. Comparable Pt nanowire morphology is obtained by Shangfeng et al. following the same Pt growth methodology [170]. Cross sectional EDX elemental mapping also revealed an even Pt coating around the fibers (Figure 5-4). The thickness of the Pt layer was uniform across all fiber cross sections. The 1D architecture of the Pt nanowire is expected to enhance mass transport by improving the oxygen diffusion to the Pt surface. Therefore, desired efficiency could be achieved with reduced Pt loading as a result of the higher Pt utilization. The applied electroless plating was found to be suitable for fabrication of CNF/Pt mats for fuel cell assembly and testing. Although challenges were met in exact Pt loading control, with careful handling, using as strict as possible deposition conditions, satisfactory control was achieved.  Figure 5-4: TEM of Pt Nanowire crystal structure (left) and cross sectional STEM EDX map of Pt around each fiber (right) (STEM/EDX, 200 kV) 5.2.2 Chemical composition by XRD and XPS  Surface chemistry analysis was conducted to compare the OAN and RNL before and after chemical functionalization and Pt deposition. Because OAN had much lower graphitic content 101  than RNL, the goal was to determine whether the Pt deposition would be more favorable for OAN layers. The effect of lower graphitic content was also investigated for RNL samples carbonized at 700 °C for comparison. The influence of chemical functionalization treatment was analyzed for OAN and RNL with regards to changes in nitrogen and carbon species. Lastly, XPS and XRD data validated the quality of the Pt coating by measuring Pt0/Ptn+ ratio and crystal structure.    5.2.2.1 Functionalization analysis  Chemical treatment using HNO3 was used to encourage even distribution of Pt during electroless plating. Acid treatment on CNT and CNF has been shown to introduce oxygen functional groups which assist in the deposition of Pt [61,171,172]. To verify the production of functional groups from acid treatment, the surface chemistry was analyzed by XPS. Figure 5-5 illustrates the nitrogen species before and after HNO3 treatment for RNL carbonized at various temperatures. Left over nitrates were seen in all three samples, with 900 °C and 1100 °C carbonized samples exhibiting higher amounts. From the non-treated samples, it appears that pyridinic nitrogen decreased with increasing carbonization temperature, while substitutional and pyridine nitrogen experienced the opposite trend. The nitrogen trends were consistent with literature regarding the heat treatment for PAN carbon fibers [167]. After HNO3 treatment, RNL carbonized at 1100 °C appear to exhibit unchanged ratio between nitrogen species. This suggests that acid treatment did not have large effect on this sample, possibly due to the higher content of graphitic carbon – which is typically inert and stable (Figure 5-6). Contrary, RNL 900 °C showed the largest change in nitrogen species, converting from predominately substitutional nitrogen before treatment to pyrrolic after treatment. Also, a large drop in pyridinic nitrogen was observed after acid treatment. These changes indicate the HNO3 treatment was very effective on RNL 900 °C samples. Interestingly, although RNL 700 °C had the least graphitic carbon, the acid treatment 102  did not appear to affect the chemical structure. Except for the increased nitrate amount, the remaining nitrogen species appear unchanged before and after treatment (Figure 5-5). It is not clear from observation of C and N species the reason for the large impact of treatment on RNL 900 °C samples. However, the overall atomic % composition of the sample reveals that the amount of nitrogen increased significantly after treatment (0.8% to 3.11%) (Table 5-1). Contrary, atomic % of nitrogen was relatively stable before and after treatment in RNL 700 °C (11.53% to 11.51%) and 1100 °C (1.39% to 1.37%) samples. This may be a result of left over nitrates by incomplete washing after treatment because RNL 900 °C retained the highest amount of nitrates.      Figure 5-5: Comparison of nitrogen species before (left) and after (right) HNO3 chemical treatment for RNL carbonized at 700 °C, 900 °C, and 1100 °C Following the same procedure, the effect of geometrical structure and spinning method on surface chemistry was investigated. In Figure 5-7, OAN and RNL carbonized at 900 °C was measured by XPS before and after HNO3 treatment. From the N 1s deconvoluted curves, the OAN appears to have dominant substitutional and pyridinic species. The percent composition of each species was significantly different from RNL samples, although they were carbonized at the same temperature. According to Laffont’s study, at a transition temperature of ~500 °C, pyridinic 010203040506070Component %RNL 700RNL 900RNL 1100010203040506070Component %RNL 700 + HNO3RNL 900 + HNO3RNL 1100 + HNO3103  structure partially changes to form substitutional nitrogen [167]. With increasing carbonization temperature, Watt proposed N atoms located at graphene edges link together and are eliminated as gaseous N2, leading to the formation of graphene structure [173]. As shown in Figure 5-7, RNL had a larger amount of substitutional nitrogen, suggesting the successful formation of substitutional nitrogen from pyridinic. Moreover, atomic % composition indicated only 0.8% nitrogen in RNL, while OAN 900 °C had 4% (Table 5-1). This observation may explain the discrepancy in chemical composition and surface properties of the OAN fibers. Because OAN was fabricated using modified electrospinning method, it is possible the alignment and geometrical structure of the assembly affected the carbonization process. As discussed previously, the whipping motion in electrospinning influences the molecular chain alignment, possibly inducing favorable gas diffusion and steric situation [167]. Moreover, the arrangement and elimination of atoms was further enhanced by applied tension during heat treatment [167]. Although moderate tension was applied, OAN experienced larger shrinkage than RNL. As a result, the chemical structure and the material properties differed because of the nature of OAN fibers.  Because of the inherent nature of the OAN fibers, it was expected that the composition of nitrogen before and after treatment would not follow the same trend as RNL. As shown in Figure 5-7, the OAN fibers showed a more balanced composition of nitrogen, with the addition of nitrates. Instead of experiencing a large increase in pyrrolic nitrogen, the OAN samples had a small decrease in pyridinic and substitutional. Based on the virtually identical nitrogen groups after treatment, it appears that the HNO3 treatment did not have a large influence on the surface chemistry. Again, the majority of nitrates detected are likely residual from incomplete sample washing. After HNO3 treatment, nitrogen content increased from 0.8% to 3.11% and 4% to 6.13% for RNL and OAN, respectively. Similarly, the HNO3 treatment did not seem to affect C 104  composition. With RNL samples, carbon content decreased after treatment, however in OAN samples the carbon content remained constant. Interestingly, oxygen atomic % is the highest for both treated and non-treated OAN sample (Table 5-1). It is possible that oxygen groups provided functionality for the Pt plating, thus increasing the Pt0 content [11].         Figure 5-6: Comparison of carbon species before (left) and after (right) HNO3 chemical treatment for RNL carbonized at 700 °C, 900 °C, and 1100 °C  Figure 5-7: Comparison of nitrogen species before (left) and after (right) HNO3 chemical treatment for OAN carbonized at 900 °C  01020304050607080RNL 700RNL 900RNL 110001020304050607080RNL 700 HNO3RNL 900 HNO3RNL 1100 HNO30102030405060RNL 900OAN 9000102030405060RNL 900 + HNO3OAN 900 + HNO3105  Table 5-1: Surface chemical composition denoted in atomic percentages Sample ID Atomic % C O N Na Pt RNL 900 96.53 2.64 0.83 0.00 0.00 RNL 900 HNO3 84.54 12.34 3.11 0.00 0.00 RNL 900 Pt 58.85 16.40 5.25 0.00 19.50 RNL 900 HNO3 Pt 59.55 16.26 8.53 0.00 15.66 RNL 700 86.26 2.21 11.53 0.00 0.00 RNL 700 HNO3 80.60 7.89 11.51 0.00 0.00 RNL 700 Pt 72.17 12.83 8.72 0.00 6.28 RNL 700 HNO3 Pt 71.74 10.43 10.84 0.00 6.99 RNL 1100 94.85 3.76 1.39 0.00 0.00 RNL 1100  HNO3 91.60 7.03 1.37 0.00 0.00 RNL 1100 Pt 55.47 17.54 6.70 0.00 20.29 RNL 1100 HNO3 Pt 85.33 8.45 3.68 0.00 2.54 OAN 900 78.49 15.87 4.00 1.64 0.00 OAN 900 HNO3 78.94 14.93 6.13 0.00 0.00 OAN 900 Pt 64.16 16.09 5.18 0.00 14.57 OAN 900 HNO3 Pt 70.72 19.26 2.46 0.00 7.56  5.2.2.2 Chemical composition of platinized CNF   Figure 5-8: Relationship between degree of graphitization and metallic Pt to oxidized Pt ratio 0.000.501.001.502.002.503.000.00 500.00 1000.00 1500.00Ratio of metallic Pt to oxidized PtCarbonization T  (°C)RNL 700 PtRNL 700 HNO3 PtRNL 900 PtRNL 900 HNO3 PtRNL 1100 PtRNL 1100 HNO3PtOAN 900 PtOAN 900 HNO3 Pt106  XPS analysis was conducted on samples after Pt deposition to observe the effect of HNO3 treatment. The results shown in Figure 5-8 reveal there is an increasing trend between carbonization temperature and the ratio Pt0/Ptn+. Interestingly, higher metallic content was achieved across all samples without HNO3 treatment. It is well known that higher degree of graphitization results in a more stable sp2 carbon structure, this generally means less number of nucleation sites for Pt reduction [174,175]. Although XPS results indicated lower metallic content for lower carbonization temperatures, it is not representative of the Pt distribution and coverage. Since XPS is a surface method and it scans a small area (20 µm2), it cannot give information about Pt yield or thickness of Pt coating throughout the CNF layer. From SEM observations, HNO3 treatment assisted with Pt deposition by yielding a denser Pt coating with homogenous distribution throughout the fibrous layer. The effect of the treatment was most clearly seen in 900 °C carbonized RNL samples. This observation may be explained by the Pt coverage and reduction process (Figure 5-9). It is assumed that Pt deposition is initiated by heterogeneous nucleation at an anchoring site (defect or functional group) and continues to deposit by crystal growth [176]. The slow reduction process allows for the formation of nanowires by anisotropic growth [141]. Thus, samples exposed to acid treatment possessed more anchoring sites for nucleation—leading to denser nanowire growth on the CNF surface. This led to homogenous distribution of nanowires with high surface area (Figure 5-10). Since the resulting Pt nanowire coating exhibited high surface area, it was more prone to developing a thin oxide layer when exposed to air [177]. This phenomena explains the reason samples treated with acid retains lower ratio of metallic to oxidized Pt content (Figure 5-11). The metallic ratio is higher for highly graphitized samples for the same reason—the Pt surface area is much lower with less nucleation sites. This thin oxide layer is present on all Pt coatings and would be removed during potential cycling before MEA testing. Overall, 107  the current method of acid treatment allows for dense and homogenous deposition of Pt nanowires with large surface area. It should be noted that CNF produced for the remainder of the thesis was carbonized at 900 °C instead of the original optimized 1100 °C because more uniform Pt was achieved. Since samples carbonized at 900 °C showed surface groups that would be beneficial for Pt deposition, higher Pt yield and stronger bonding may exist. It was not critical to carbonize at higher temperatures since carbonization at 900 °C already provides sufficient electrical and mechanical properties.  Figure 5-9: SEM images of CNF/Pt without HNO3 treatment (left) and with treatment (right); Pt surface area is higher after treatment    Figure 5-10: Nanowire growth on CNF surface with (left) limited anchoring sites and (right) many anchoring sites [176] Platinum electroless plated RNL was analyzed by XRD and shown in Figure 5-12. The diffraction shows strong characteristic peaks for crystalline platinum. The three major peaks detected here correspond to (111), (200), and (222) at 2θ = 39.77°, 45.99°, and 83.70°, which indicates that the deposited Pt is face centered cubic (FCC) structure (JCPDS Card 04-0802).  The presence of (111) plane normally indicates high catalytic activity, since the adsorption/desorption 108  occurs on this facet [20,178–181]. From the sharp and high intensity of the diffraction peak, it appears the catalyst layer contains an abundant amount of Pt(111) which would improve the catalytic activity of the FC and potentially decrease amount of Pt required.  Figure 5-11: Comparison of metallic to oxidized Pt ratio for RNL with and without HNO3 treatment  Figure 5-12: X-ray diffraction diagram for Pt electroless plated CNF layers    010203040506070Pt metallic Pt2+{Pt(OH)2}Pt2+ {PtO} Pt4+ {PtO2 +Pt(OH)4}RNL 900 PtRNL 900 HNO3 Pt109   Ex situ CV and effective platinum surface area   Figure 5-13: EPSA was calculated based on the area under the Had oxidation curve Although there are a number of studies using perchloric acid as the electrolyte solution to avoid adsorption of sulfate anions onto the Pt surface, prior work has shown that ECSA does not appear to be affected by the supporting electrolyte [182]. Moreover, H2SO4 electrolyte can mimic the anion−electrode interaction between Pt surface and surrounding nafion sulfonic groups [183–185]. To gain a preliminary understanding of how the catalyst layer will perform in fuel cell, ex situ CV was conducted to measure the electrochemical surface area. From the integrated charge in the hydrogen adsorption/desorption peak areas (Figure 5-13) in the CV curves and the Pt poly-crystallite hydrogen adsorptions constant 0.21 mC cm-2 Pt, the effective platinum surface area (EPSA, cm2Pt  cm-2) and the specific Pt surface area (APt, m2 gPt-1) were calculated according to the following equations; where QM is the charge associated with the formation of a monolayer of hydrogen [186]: EPSA (cm2Pt cm-2) = QM (mC cm-2)/0.21 mC cm-2 (5.1) APt (m2 gPt-1) = EPSA (cm2Pt cm-2)/Pt loading (mg cm-2)*0.1 (5.2) 110  𝑃𝑡 % 𝑈𝑡𝑖𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛 = 𝐸𝑃𝑆𝐴 (𝑐𝑚2𝑐𝑚2) ×1𝐴𝑃𝑡(𝑔𝑃𝑡𝑚2) ×1𝑃𝑡 𝑙𝑜𝑎𝑑𝑖𝑛𝑔(𝑚2𝑔𝑃𝑡) × 100% (5.3) EPSA is related to platinum utilization shown in Equation (5.3), therefore improving the EPSA would contribute to enhancing the fuel cell efficiency or decreasing the amount of Pt deposited on the catalyst layers. The voltammetric profiles of CNF/Pt membrane (Figure 5-14) displays the characteristic features with an H-sorption potential region (0.0 – 0.4 V), a double layer potential region (0.4 – 0.7 V) and an O-sorption potential region (0.7 – 1.0 V). From the cyclic voltammogram, the EPSA was calculated by hydrogen desorption region. The hydrogen desorption portion of the curve also revealed small humps that are characteristic of different Pt crystallographic planes that exists [187].  This is consistent with the XRD results presented earlier.    Figure 5-14: Cyclic voltammogram of CNF/Pt catalysts with 3 different Pt loadings at 25 °C in N2-saturated 0.09M H2SO4, scan rate: 20 mV s−1 (1.25, 1.5, and 2.0 g L-1 precursor), normalized to geometric surface area of electrode 5.3.1 Electrochemical surface area of RNL Low, medium, and high concentration (∼1.25, 1.5, and 2 g L-1) of Pt precursor was used to investigate the yielding of electroless plated Pt. The results suggest that the yielding is generally 111  presented by a linear relationship as seen in Figure 5-15, although not completely linear. However, further increasing the Pt loading would not be desirable since the Pt formed agglomerates and decreased the available active surface area. As seen from prior SEM images, the high concentration sample revealed large clusters that were densely packed. This dense morphology likely delays fuel diffusion and reduces the Pt utilization since the area beneath the surface is not easily accessible.   Thus, it appears the ECSA trend was slowly plateauing as the loading increased. This demonstrated a diminishing return between Pt loading and active surface area, which is undesirable since the intention is to reduce Pt cost by reducing the amount deposited. The current method appears to be a promising deposition method for applying a thin layer of Pt on carbon nanofibers. Previous studies using other methods of Pt deposition obtained electrochemical area (APt) of about 50 – 100 m2 gPt-1, while the current study achieved 66.5 m2 gPt-1. This method attained comparable electrochemical areas to industry and literature but has room for improvement [138,186,188–190].    Figure 5-15: Relationship between Pt precursor concentration, ECSA and Pt loading    112  5.3.2 Electrochemical surface area of OAN To evaluate the electrochemical performance of OAN sample, ECSA was evaluated by CV using the same method as mentioned in prior sections. As shown in Table 5-2, OAN follows the same trend as RNL sample with similar Pt loading. The results suggest that Pt deposition method was consistent between OAN and RNL samples. Additionally, the Pt nanowire morphology was preserved. Thus, OAN samples show comparable ECSA to RNL, suggesting orientation of the fiber does not have an effect for ex situ electrochemical performance. Because ex situ CV is performed in liquid electrolyte cell, the fibers would be completely wetted and in contact with reactants. The reaction in situ is slightly different since reactants are in gas phase, thus diffusion through the pores and efficient water management play a significant role. It is important to note that ex situ CV gives an indication of the available Pt surface area, however the in situ performance depends on multiple variables and operating conditions.  Table 5-2: ECSA for various RNL and OAN samples Sample description Pt loading (µg cm-2) ECSA ex situ (m2 gpt-1) RNL, med I/C 157 36.87 RNL, high I/C 141 26.02 OAN, med I/C 140 23.54 RNL, med I/C, 700°C 283 38.79 RNL, med I/C 279 45.60 RNL sputtered, med I/C 81 82.83  Table 5-3: H2 cross over current and shorting resistance for RNL and OAN samples tested in CAT cell at 105% and 50% RH for 68 °C and 85 °C, respectively (*V50 was tested in 68 °C for both RH) ID 68 °C 85 °C  ix (mA cm-2) Rs (Ω cm-2) ix (mA cm-2) Rs (Ω cm-2) RNL 150 µg cm-2, med I/C 2.923 1992.74 1.953 127.471 RNL 150 µg cm-2, low I/C 4.065 2856.125 0.379 123.981 RNL 150 µg cm-2, high I/C 4.26 6658.567 0.359 193.262 RNL 250 µg cm-2, med I/C 5.078 376.871 0.83 103.874 OAN 150 µg cm-2, med I/C 1.559 38.676 0.46 29.213 V50 150 µg cm-2, 1.1 I/C 6.702 141.4 5.084* 86.1* 113  H2 crossover current (ix) and shorting resistance (Rs) was estimated by linear sweep voltammetry [191]. Comparing the crossover current, it appears all the RNL samples have comparable crossover to the baseline V50 (Table 5-3). The OAN sample experienced an ix of 1.56 mA cm-2, the lowest of the MEAs. Comparing the OAN and RNL geometry, it appears that OAN has lower ix possibly due to the difference in fiber orientation and porosity. As shown in prior chapters, the OAN orientation revealed lower porosity compared to RNL because the fibers were densely adhered together. In relation to ECSA and Pt loading, the ix does not appear to show correlations between the samples. The results indicate that all RNL regardless of I/C and ECSA have similar ix. V50 baseline showed slightly higher ix likely due to the unoptimized fabrication of the catalyst layer, creating non-homogenous distribution of the carbon and ionomer. Compared to the conventional catalysts reported in the literature, CNF appears to have higher crossover current density due to the large open porous architecture [191–193].     Ionomer deposition and morphology  Preliminary ionomer deposition was achieved by using vacuum pump and funnel method as described in Chapter 2.5. The loading and distribution of the ionomer was controlled by diluting the ionomer solution to achieve lower loading. Using the funnel and pump efficiently distributed the ionomer and did not induce blocked pores. The ionomer was thinly coated on each fiber and throughout the thickness. Loading was measured by weighing the sample before and after deposition and dividing by geometric surface area. Ionomer to carbon weight ratio, I/C, was calculated from the measured ionomer loading and measured Pt loading (5.4).  𝐼𝐶=  𝑖𝑜𝑛𝑜𝑚𝑒𝑟 𝑙𝑜𝑎𝑑𝑖𝑛𝑔 (𝑚𝑔𝑐𝑚2)𝑃𝑡 𝑚𝑎𝑠𝑠 𝑝𝑒𝑟 𝑎𝑟𝑒𝑎 (𝑚𝑔𝑐𝑚2) − 𝑃𝑡 𝑙𝑜𝑎𝑑𝑖𝑛𝑔 (𝑚𝑔𝑐𝑚2) (5.4) 114  The SEM images suggest that no pores were blocked and a thin layer was distributed on each fiber (Figure 5-16). High magnification micrographs prominently show a thin layer of ionomer surrounding the Pt nanowire clusters. Cross sectional TEM images were obtained to observe the ionomer and Pt coating dispersion and thickness. The EDX chemical composition maps reveal a homogenous thin layer of Pt deposited around each fiber. Since ionomer is responsible for transporting H+ ions across the membrane, it is important to have a connected network in the catalyst support layer to improve the mass transport properties. From the EDX fluorine mapping images, there were thin strands connected across the fibers throughout the cross section. Because the fiber mat is porous, it is crucial that ionomer is dispersed through the thickness and connected between fibers to form a continuous network for proton transport. In the combined map, the ionomer uniformly surrounds the Pt layer creating a possible triple phase boundary region (Figure 5-17). Thickness measured from TEM images reveal that the ionomer coating was 29.46 ± 21.38, 63.08 ± 60.47, and 102.65 ± 84.90 nm thick, for low, medium, and high I/C, respectively. Depositing 5 wt% and 10 wt% ionomer solution resulted in I/C of 0.11 and 0.285 respectively (Table 5-4). Medium loading I/C was achieved by annealing at 140 °C after the first deposition and repeating with 5 wt% ionomer again. The standard deviation for the ionomer thickness was large because ionomer was stretched between fibers, forming bridge structures (Figure 5-17 and Figure 5-18). The ionomer adhering onto the CNF/Pt was fairly thin and consistent throughout the regions. From the distributions graphs, there is a clear trend showing ionomer thickness increases with increasing I/C (Table 5-4 and Figure 5-19). Thus, the results confirm that ionomer loading and thickness was tailored by using the current deposition method with satisfactory distribution. Further control of the ionomer distribution and coating thickness would require thorough future work.    115  Previous ionomer study has shown that increasing I/C increases oxygen transport resistance. The authors also revealed that lower pressure-independent oxygen transport resistance was attributed to homogenous distribution of ionomer [194]. Decreasing I/C improved the oxygen transport resistance while increasing protonic resistance. It was recently reported in literature that effective ionomer thickness is an important parameter. The authors found optimal ionomer distribution and protonic conductivity with I/C of 0.4 and effective ionomer thickness of 1.6 nm [194]. Although not directly comparable because of using CNF, the current results agree with the conclusions by Orfanidi et al. which suggests that oxygen mass transport was largely affected by ionomer distribution and thickness [194]. In situ results presented in Chapter 6.3.2 discusses the effect of I/C on fuel cell performance and oxygen transport resistance.      Figure 5-16: Ionomer distribution between the pores and all around CNF/Pt surface Table 5-4: Ionomer loading and I/C for CNF with similar Pt loading Pt loading (mg cm-2) I/C Ionomer loading (mg cm-2)  0.138 0.110 0.057 0.164 0.156 0.080 0.141 0.285 0.178  116   Figure 5-17: TEM/EDX mapping of ionomer by fluorine signal (left), combined map showing Pt, carbon and ionomer (right)   Figure 5-18: TEM/EDX maps of combined Platinum (red) and Fluorine in ionomer (green) distribution (A-D) and only ionomer distribution (E-F): Low I/C sample (A and E), medium I/C with a part of NRE membrane shown (B and F), high I/C (C and G), and high I/C OAN revealing ionomer pockets (D and H).  117   Figure 5-19: Ionomer thickness distribution for low, mid, high I/C on RNL (averaged over 200 readings) 5.4.1 Surface wettability of CNF catalyst layer   Surface wettability demonstrated a shift from hydrophilic to hydrophobic surface upon deposition of ionomer. Figure 5-20A reveals low contact angle for water wettability for CNF/Pt, meaning the platinized fibrous layer was moderately hydrophilic. Figure 5-20B, however, illustrates the contact angle after a thin layer of ionomer was coated on the fiber surface. The 125° contact angle suggests that the catalyst layer was hydrophobic. The results are promising since it is consistent with literature regarding ionomer chain alignment [30]. The hydrophilic CNF/Pt substrate encourages the hydrophilic side chains of ionomer to be preferentially oriented towards the CNF/Pt surface. The orientation allows for the formation of an intermediate water layer as shown in Figure 5-21 [30]. The hydrophobic backbone facing away from the CNF/Pt also improves water management by keeping water out of secondary pores [30]. This continuous water layer enhances the proton conductivity and electrocatalytic activity because the oriented hydrophilic ionomer group donates protons to this intermediate water film. In the practical scenario where ionomer is not in contact with all active Pt surface, a continuity of water channels can maintain the 118  proton conduction network [30]. Therefore, retaining the intermediate water layer in close contact with Pt is highly desirable for the proton conductivity and overall electrocatalytic activity.     Figure 5-20: Wettability of catalyst layer before and after ionomer deposition; contact angle of 50° and 125° (A and B)  Figure 5-21: Orientation of ionomer side chains towards hydrophilic Pt particles  Summary  To obtain the ideal Pt layer on the optimized CNF, several deposition methods were investigated. However, many challenges were encountered, proving some methods were unfeasible and inconsistent. The electroless plating method was the most suitable for CNF, producing well dispersed low Pt loading nanowires. The Pt nanowires were characterized in XPS, XRD, ex situ CV, and SEM to determine the Pt surface area and morphology. XPS revealed that more nucleation sites led to the formation of more oxidized Pt because of exposed surface area. XRD confirmed the presence of FCC Pt structure and the (111) crystal plane. This crystal plane is 119  desirable because literature suggests this is a catalytically active plane for ORR. Visual inspection by SEM indicated that increasing concentration of the Pt precursor increased the Pt loading and produced larger Pt clusters. Additionally, ex situ CV determined the optimal Pt precursor concentration to obtain highest active surface area. ECSA results were in agreement with SEM images, suggesting that the medium concentration (1.5 g L-1 Pt precursor) sample attained sufficient Pt surface area with a homogenous distribution. Trends from ECSA calculation suggests that lower Pt loading was more favorable because less Pt could be used while maintaining adequate active surface area. Lastly, the OAN ex situ properties were comparable to RNL, revealing similar ECSA at a given Pt loading. The consistency of the ECSA between RNL and OAN suggests Pt nanowires were of reasonable quality and repeatability, indicating the success of the Pt plating process and post processing. The funnel technique used to deposit the ionomer successfully formed a uniform and connected network without blocking pores throughout the catalyst layer thickness. Different concentrations of ionomer were deposited to determine the ideal I/C ratio for the CNF layers. SEM inspection revealed that ionomer was thoroughly coated on CNF and did not cover any pores. Ionomer thickness was measured by STEM/EDX mapping and revealed an increase in thickness with increasing ionomer concentration. Ionomer coating thickness was 29.46 ± 21.38, 63.08 ± 60.47, and 102.65 ± 84.90 nm thick, for low, medium, and high I/C, respectively. Additionally, ionomer bridges were formed between fibers for medium and high I/C samples. The bridges were expected to be beneficial since a homogenous network of ionomer would be formed throughout the CNF thickness, thereby facilitating a continuous TPB. Fuel cell performance is highly dependent on ionomer distribution and effective thickness, therefore it is crucial to compare the quality of the ionomer layer through visual techniques as well as in situ fuel cell measurements (Chapter 6.3).  120   In situ Evaluation of CNF Based MEA  Introduction  To understand and evaluate the feasibility of CNF catalyst support layers, the membrane electrode assembly (MEA) was fabricated for in situ fuel cell testing. A modified process for assembling the MEA was developed specifically for the CNF layers. Because the CNF was a standalone layer, conventional methods of assembly did not apply. Two types of MEA were fabricated, one for pre-screening and one for full scale fuel cell performance evaluation. The screening MEA had a small active area of 1.13 cm2 and was tested in HOT and NOC conditions at specific potentials. The obtained results provided insights into the optimal CNF designs. Next, the top five performing samples from the MAT cell were fabricated into MEAs for full in-depth fuel cell testing. Polarization curves and Nyquist plots were collected for each sample. Maximum power density, ohmic resistance, ionic resistance, and oxygen transport resistance were evaluated and compared to unoptimized V50 CB. Particularly, the performance results were interpreted in the context of Pt loading and morphology, ionomer loading and distribution, fiber orientation, and fiber electrical properties. The results of the pre-screening and full fuel cell test emphasized the effect of material variables on fuel cell polarization losses. Ultimately, ionomer distribution appeared to be correlated with the fiber orientation, Pt loading, and ionomer loading. Lastly, characterization of the MEA after fuel cell testing was performed to verify the integrity of the CNF catalyst.     In situ MAT cell performance (pre-screening of CNF) Following the testing procedure described in Chapter 2.7.9, in situ testing only under specific conditions and voltages were used to provide indications to the regions of polarization losses (Figure 6-1). Sample IDs and several variables of interest are listed in Table 6-2. Each 121  sample was run three to five times and the average of those runs are reported. For the RNL MEAs, the effects of ionomer loading, Pt loading, and carbonization temperature (CNF support conductivity) were explored. Three different I/C was analyzed to determine the ideal range for ionomer deposition on CNF structures. The I/C was lower compared to the baseline samples (1.1) because higher I/C would cause pore blockages and poor distribution in the CNF structure.  Two Pt loadings, 150 and 250 µg cm-2, for CNF were fabricated in order to directly compare with the baseline loadings. The effect of carbonization temperature was analyzed to investigate the effect of electrical conductivity and surface chemistry. Since XPS revealed varying chemical compositions, the resulting Pt growth and properties could have an effect on overall in situ performance. All MEAs were compared to the two baseline samples with Pt loading of 150 µg cm-2 and 250 µg cm-2 (described in Chapter 2.6). Vulcan catalyst was used as a baseline comparison because the Pt is distributed on the surface of the spherical carbon support, similarly to surface deposition of Pt on nanofibers. The difference is agglomerated structure in Vulcan catalyst layers, as opposed to aligned fibrous structure in CNF based MEAs, as well as diameter of the carbon support and porosity. A regional morphology comparison between CNF based catalyst layer with V50 CB is shown in Figure 6-2D and E. The V50 baseline used in this study was prepared using a Meyer bar method and has been previously measured and reported by Shukla et al. [195]. The current baseline was not optimized, however the porosity and characteristics reported in the literature provides an estimate of the properties. At an I/C of 1.1, the V50 is expected to have a porosity of roughly 40%, similar to the literature [195]. Table 6-3 compares testing results for all samples measured at 825 mV which is the activation region. The performance in this region is related to activation energy barrier between reacting species and is expected to be primarily affected by the electrode properties, such as ECSA, electrical properties of the electrode, and 122  ionomer contact with the active sites [44]. Also, overall pressure, temperature and reactant concentration play a role in this region [44]. The results demonstrate that all the CNF based MEAs outperformed the V50 baselines, in some cases even a 5x enhancement. Although V50 and CNF sample consist of the same Pt loading, CNF demonstrated superior performance owing to its highly porous structure and favorable Pt distribution. The very porous structure of CNF ensured uniform (high) concentration of the reactant gases throughout the layer, which was not the case with a dense unoptimized V50 catalyst layer. It is believed that TPB was more optimized since Pt was evenly distributed on CNF surface, as evident in the TEM images (Figure 5-18). Moreover, V50 showed poor performance in the activation region, indicating poor Pt utilization compared to CNF with Pt nanowire. The results suggest that the large surface area and high porosity of CNF contributed to the improved catalyst performance. Table 6-1: Operating conditions of single cell performance tests performed in MAT cell hardware  Normal condition (NOC) Hot condition (HOT) Temperature (°C) 68 85 Inlet relative humidity (%) 105 50 Balanced inlet pressure (bara) 2.5 Stoic (both H2 and Air) > 10 Anode 100% H2 Cathode Air FFP geometry Cathode: straight channel, Anode: radial, no gasket sealing     123  Table 6-2: Summary of in situ MAT cell samples and their properties Sample ID Pt loading (µg cm-2) I/C Alignment Carbonization T (oC) R150L 138 0.110 Random 900 R150M 157 0.158 Random 900 R150H 141 0.285 Random 900 R250M  279 0.156 Random 900 S80M 81 0.132 Random 900 V50-0.15 150 1.000 Baseline NA V50-0.25 250 1.000 Baseline NA O150H 140 0.238 Orthogonal 900 R250-T2 283 0.207 Random 700 R250-T1 246 0.177 Random 900    Figure 6-1: Regions of polarization losses in typical fuel cell. Voltages at which measurements were performed in this study are marked on the graph At 825 mV, the RNL samples R250M and R250-T1 had the highest current (Table 6-3) which indicates fast kinetics at the Pt surface. This is plausible since R250M and R250-T1 both have higher Pt loading (250 µg cm-2), improving the reaction kinetics (activation). It is worth to mention that both RNLs have I/C in the mid-range level. The next best performing sample was RNL R250-T2 with the same high Pt loading, medium I/C, but carbonized at 700oC. The CNF support in this sample had lower electrical conductivity due to the lower carbonization 124  temperature. This result verifies that electrical properties of the electrode support play an important role in the kinetic region as well.   Figure 6-2: High resolution STEM/EDX images of CNF/Pt and V50. (A) Electroless plating: nanowires showing (111) crystal planes (insert); (B) Cross section of CNF/Pt; (C) Pt sputter coated on fibers close to membrane; (D and E) STEM-EDX maps comparing carbon black V50 catalyst layer (0.25 mgPt cm-2), white rectangle selects a region visually comparable in morphology to the CNF region; (F) EDX map of Pt distribution (red) on fibers; Pt layer illustrates 0.25 mg cm-2 loading. Green color denotes Fluorine signal from ionomer, blue color denotes Carbon particles or fiber cross section, and red denotes Pt. Ionomer thickness or I/C ratio also played a critical role in this region. Comparing 900 °C carbonized RNLs with similar Pt loading, but different I/C, R150L (low I/C), R150M (mid I/C) and R150H (high I/C), the highest performance was obtained with R150M (mid I/C). Low I/C RNL, as shown in (Figure 5-18), had a very thin layer of ionomer surrounding CNF/Pt fibers, but no connecting bridges to provide path for proton transport. Medium I/C sample contained a number of bridges and medium thickness of the ionomer film, enabling more efficient transport of protons without inhibiting oxygen access to active sites too much. High I/C sample contained more 125  connecting ionomer bridges, but the thickness of ionomer on the fibers was too high, hence inhibiting oxygen transport to Pt. Comparing the geometry, orthogonally oriented (O150H) CNF achieved lower current density compared to randomly oriented (R150H) CNF in the activation region. This was possibly due to the lower electrical conductivity of the OAN sample, which affected the activation energy barrier for electron transport. Generally, the trend in the kinetic region reveals that CNF samples performed better in NOC conditions compared to HOT conditions in the activation region. This was expected since humid and colder NOC conditions enhance proton conductivity due to the presence of condensed water, while dry and warmer HOT conditions increased proton resistance [196]. The effects of parameters in the ohmic region were investigated under NOC operation at 650 mV and HOT operation at 600 mV. Ohmic losses in a fuel cell are typically caused by resistance to the electron transport through the electrodes and external circuit, by contact resistances between flow-field plates/gas diffusion layer/electrodes/membrane, as well as resistances to the ionic transport through the ionomer in electrodes and the membrane [197]. Assuming that all external electrical resistances, and membrane ionic resistance are similar for all samples, the effect of Pt loading and distribution, ionomer loading, CNF electrical conductivity, and fiber alignment on the performance was evaluated. Figure 6-3 (red bars) compares performance of the samples with varying Pt loading, while I/C and carbonization temperature are kept constant. As expected, Pt loading does not have a large impact on ohmic resistance in CNF samples under both HOT or NOC conditions. However, CNF sample R150M with Pt loading of 0.15 mg cm-2 showed more than two times higher performance compared to the same loading baseline V50 sample, indicating lower ohmic resistance (Figure 6-3 green bar). For this CNF sample, an in-plane conductivity of 6.73 S cm-1 was measured (Table 4-3), which is higher than 4 126  S cm-1 for CB in V50 [22]. In addition, a thin, almost continuous film of Pt on the surface of the fibers is providing additional conductivity to the R150M CNF layer, as through-plane conductivity increases from 0.266 ± 0.040 S cm-1 before Pt deposition to 0.388 ± 0.0289 S cm-1 after Pt deposition. When compared to the S80M sputtered sample (conductivity of 0.219 S cm-1 at 34 µgPt cm-2), which has Pt only on one side of CNF mat and a very low loading, the sputtered CNF sample performed much lower than the electroless plated sample and comparable to the V50 sample (Figure 6-3 green bar). Although sputtered sample had lower Pt loading, V50 and sputtered sample achieved comparable NOC performance. This comparison shows that Pt distribution matters and could provide additional conductivity for the catalyst layer. Although the electrical conductivities did not differ significantly, it is worthwhile to report the through-plane properties because the catalyst layer electrical properties are not well-understood or accurately measured in the current literature [198–200]. Understanding the through-plane properties is beneficial for managing the local thermal variations which is linked to degradation and stability issues [199,201]. Pt distributed on the surface of the carbon support was beneficial to lower the ohmic resistance. In addition to carbon support conductivity and Pt distribution effect, ionomer loading and distribution also played a role. Ohmic resistance of V50 catalyst layer could be high due to the fact that ionomer coated Pt/C agglomerates have less electrical contact points, hence increasing overall electrical resistance. The effect of ionomer in the ohmic region was also seen when comparing the performance in HOT and NOC conditions (Figure 6-3 green bar). The performance of CNF samples in HOT compared to the NOC condition was lower due to the increased resistance of the ionomer film under higher operating temperatures and dry conditions. Even when ionomer bridges are present, the distance between the ionomer coated fibers are large, hence making proton conductivity a major challenge for the CNF based samples. In the dry HOT conditions, this effect 127  was especially enhanced. V50 catalyst layer, however, performed better in HOT than in NOC conditions. Due to the unoptimized deposition of the catalyst ink and the nonhomogeneous film of ionomer (Figure 6-2E), it is possible that NOC conditions caused flooding of the catalyst layer even in the ohmic region, causing drop in performance. Thick and connected ionomer in V50 catalyst layer provided satisfactory protonic conductivity in HOT conditions, hence retained higher performance (Figure 6-3 green bar).  Figure 6-3 (blue bars) reveals the effect of ionomer loading (or I/C ratio) on performance of CNF samples in the ohmic region. The results follow the same trend as in the kinetic region, with the mid I/C being the optimal design. Low I/C had only a thin layer of ionomer film coating the Pt (Pt + ionomer film thickness ~ 30 nm), which does not contain ionomer bridges to provide protonic path. NOC performance was satisfactory but HOT performance suffered a significant drop due to drying of ionomer. High I/C sample had the lowest performance. This is possibly not so much the ohmic effect, but the fact that fibers are covered in a thick (103 nm) ionomer film, inhibiting oxygen transport to the active Pt sites, lowering the overall performance. Mid I/C sample, with moderately thick ionomer film and a number of ionomer bridges, provided the best performance.  Effect of electrical conductivity of the CNF support on the performance in the ohmic region is given in Figure 6-3 (purple bars). As expected, in agreement with ex situ electrical measurements, the CNF carbonized at lower temperature of 700°C showed lower performance due to lower CNF support conductivity. Figure 6-3 (yellow bars) compares performance of CNF samples with randomly aligned fibers and orthogonally aligned fibers. As the ex situ conductivity data revealed, RNL samples had higher electrical conductivity, hence improved performance. Although the expectation was that orthogonally aligned CNF samples would improve overall 128  performance, it seems that aligned fibers did not contain enough electrical contact points between the fibers, which was manifested in increased ohmic resistance.  Table 6-3: Summary of current densities at 825 and 100 mV for all samples measured in the MAT cell; 68 °C 105% RH and 85 °C 50% RH for NOC and HOT, respectively  Pt loading (µg cm-2) I/C Alignment Carbonization T (oC) Current Density at 825 mV (A cm-2) Limiting Density at 100 mV (A cm-2)     NOC HOT NOC HOT 138 0.110 Random 900 0.0361 0.0056 3.4410 3.0048 157 0.158 Random 900 0.0852 0.0509 3.8579 3.2762 141 0.285 Random 900 0.0572 0.0128 3.1850 2.0863 279 0.156 Random 900 0.1426 0.0676 3.6235 2.6889 81 0.132 Random 900 0 0 2.2786 1.9766 150 1.000 Baseline NA 0.0306 0.0435 0.8142 1.7973 250 1.000 Baseline NA 0.0120 0.0509 0.7653 1.5950 140 0.238 Orthogonal 900 0.0454 0.0250 3.0011 2.7621 246 0.177 Random  900 0.1228 0.0550 3.3001 2.6693 283 0.207 Random 700 0.0954 0.0963 3.3938 3.0928 Finally, Table 6-3 compares the performance of all samples in the limiting current (or mass transport loss) region measured at 100 mV. In this region the rate of fuel consumption is equal to its maximum supply speed [44]. Thus, the current density cannot rise because the fuel cannot be supplied at a greater rate. Generally, higher current density results in lower surface concentration of reactant species. In this region, the surface concentration of reactants reaches zero, thus the current is limited [197]. When compared to CNF samples, the baseline V50 samples performed very poorly. The lower porosity carbon black structure flooded in the NOC conditions, thus inhibiting reactant oxygen transport to the reaction sites. Nevertheless, the HOT performance in V50 samples was almost two times higher compared to NOC, as condensed water was no longer present and the dense network of ionomer still provides satisfactory transport of oxygen and protons for the reaction. Most CNF samples performed well in this region, especially in NOC conditions. The open structure of CNF cathode provided an easy path for oxygen diffusion and 129  although water was produced, it did not block the pores. The best performance was achieved for sample R150M, ~160 µg cm-2 Pt loading, followed by R250M, ~280 µg cm-2 Pt loading, both with mid I/C level. Comparing Pt loading effect (including R250-T1, 250 µg cm-2), the results confirm already reported trend that catalyst layers with the same thickness, but lower Pt loading (lower Pt/C ratio), provide lower mass transport resistance [202]. Even the sputtered sample S80M with only 81 µg cm-2 Pt showed decent performance in this region, confirming the trend, although its overall performance was lower compared to other CNF samples due to high activation and ohmic overpotential.  Effect of I/C ratio, represented by samples R150H (high I/C), R150M (mid I/C) and R150L (low I/C), reveals that the sample with high I/C had the lowest performance in the mass transport region. That is because the thick ionomer film increases resistance to the oxygen transport. The effect was even more pronounced in HOT conditions. Sample with the low I/C performed moderately, as thin ionomer films provided lower oxygen transport resistance, but its overall performance was still low due to proton transport limitations. Mid I/C ratio again proved to be the optimal design. When comparing CNF samples with different carbonization temperatures (R250-T1 and R250-T2), it is evident that electrical conductivity of the support does not play a significant role in the mass transport region, as expected. Alignment of the fibers was expected to improve the mass transport due to the aligned pores. However, our results show that the performance of the aligned CNF sample (O150H) was lower than that of samples with randomly oriented fibers (R150H). This is in agreement with the densometry results, where OAN porosity was lower compared to RNL. TEM-EDX maps (Figure 6-4) reveal that the aligned CNF cathode contained pockets of ionomer in the contact points of the fibers. These pockets reduced the total porosity of the cathode affecting oxygen diffusion and increased oxygen transport resistance through the 130  ionomer film in these areas, reducing the performance in the mass transport region. Optimization of ionomer deposition is required to improve the aligned design.  Overall, the results indicate that CNF catalyst support demonstrated promising performance in the current fuel cell operating conditions. The effect of the large porosity and interconnected fiber network was evident in the in situ results in comparison to traditional CB catalysts. For comparison, a commercial CB catalyst can achieve a current density of roughly 1.5 A cm-2 at 0.6 V, which is notably higher than that of the CNF MEAs [203,204]. Overall, the effect of ionomer distribution and concentration is paramount and can be clearly seen in activation, ohmic and mass transport overpotential regions. The mid-level ionomer concentration performed best in all three regions because it displayed a good balance between ionomer thickness and distribution. The ionomer bridges provided sufficient proton transport paths while remaining openly porous to support oxygen transport. The effect on oxygen transport resistance can be seen when ionomer loading was too high, producing a thick layer surrounding the fibers (> 100 nm), as shown in R150H RNL in the mass transport region. As evident from the various performance results, the performance improvement using CNF as catalyst support is encouraging.  131   Figure 6-3: Comparison between CNF samples in the ohmic loss region (85 °C, 50% RH, 600 mV and 68 °C, 105% RH, 650 mV conditions): high and low Pt loading; Pt deposition methods; low, medium, and high I/C; carbonization temperature; and orthogonal and random fiber mat. 132   Figure 6-4: STEM/EDX mapping of OAN catalyst layer with Pt loading of 0.15 mg cm-2 (red: platinum, green: ionomer)  In situ CAT cell performance (full evaluation of CNF) From the screening in situ test, the highest performing samples were evaluated in a larger sized (5 cm2 CAT cell) fuel cell. The tested samples are listed in Table 6-4. The subsequent analysis involved recording EIS, CV, and polarization curves. The results were analyzed and compared to ex situ observations to determine structural property and performance correlations.  Table 6-4: Summary of CNF sample properties tested in CAT cell for polarization analysis Sample Name Orientation Pt loading (µg cm-2) I/C RNL 150, med I/C RNL 170 0.149 RNL 150, low I/C RNL 166 0.109 RNL 150, high I/C RNL 149 0.271 RNL 250, med I/C RNL 250 0.203 OAN 150, med I/C OAN 150 0.179  6.3.1 Current and power density analysis  Samples were tested under the same HOT and NOC conditions and polarization curves were plotted against V50 baseline sample (Table 6-5). The collected data were iR-corrected and appeared to have the same general trend as pre-screening results. The exception is the high I/C 133  RNL (150 µg cm-2) sample which performed better than low I/C. The OAN and low I/C RNL appears to achieve very similar current densities. Still, all RNL samples attained higher current densities than the V50 baseline. As shown in Figure 6-6, power density plot reveals a clearer trend and maximum power density achievable for CNF catalyst. As expected, NOC samples achieved higher maximum power densities because of the higher RH and low operating temperatures allowing sufficient proton conduction. The high I/C RNL (150 µg cm-2) achieved the highest maximum power density of roughly 1.0 W cm-2. Comparing the effect of geometry under NOC conditions, OAN power density was ~30% less than that of RNL with the same I/C and Pt loading. The negative effect of high Pt loading can be clearly seen in Figure 6-6. Maximum power density dropped nearly 50% when Pt loading was increased from 150 µg cm-2 to 250 µg cm-2 (Figure 6-6). The poor performance was largely due to the decreased electrochemical surface area because of the much larger particle size. As a result, Pt utilization was low and the catalyst remains costly with poor performance. In the interest of preserving parameters used in the CNF system, when preparing the V50 catalyst layer, the reported V50 may not be of optimal structure and performance. Thus, the current power density results for V50 are subpar compared to industrial performance standards. Nonetheless, the reported trends are helpful in determining MEA manufacturing variables to control and optimize. Furthermore, the present results are encouraging and comparable to published results using nanofibers as catalyst supports. At the same voltage (0.6 V), Park et al. reported a max current density of 900 mA cm-2 for Pt/e-CNF slurry coated on carbon paper by spray method [35]. The pioneer work by Pintauro et al., using Nafion catalyst ink and Polyacrylic acid (PAA) as the electrospinning carrier polymer, revealed a current density of 873 mA cm-2 at 0.6 V with Pt loading of 0.1 mg cm-2 [94,95]. Electrospinning using graphene doped PAN/PVDF (Polyvinylidene fluoride) with 0.1 mg cm-2 Pt loading also revealed similar results, 134  achieving roughly 1.4 A cm-2 at 0.6 V [98]. More recently, a porous CNF Pt support (0.4 mg cm-2) achieved a current density of 200 mA cm-2 at 0.6 V [39]. Additionally, the present study revealed a significant performance improvement compared to a recently reported freestanding electrospun CNF catalyst with similar fiber diameter and layer thickness [205]. Kayarkatte et al. reported a current density of roughly 400 mA cm-2 at 0.6 V for Pt loading of 0.3 mgPt cm−2, while the present study achieved ~1.0 A cm-2 with half the Pt loading [205]. Therefore, the benefits of the large porosity and interconnected fiber network is evident from the in situ results in comparison to traditional CB catalysts and published work. Table 6-5: Operating conditions of single cell performance tests performed in CAT cell hardware   Normal condition (NOC) Hot condition (HOT) Temperature (°C) 68 85 Inlet relative humidity (%) 105 50 Balanced inlet pressure (bara) 2.5 Stoic (both H2 and Air) 10.2 Anode 100% H2 Cathode Air FFP geometry Straight channels  135    Figure 6-5: Polarization characteristics of the MEAs fabricated with RNL and OAN cathode catalyst (0.15 and 0.25  mgPt cm−2) for H2/Air at 68 ◦C and 85 ◦C, Nafion 211 membrane, 2.5bara balanced anode and cathode inlet pressure. Dotted lines indicate iR-corrected data for comparison 00.20.40.60.810 1 2 3 4Potential (V)Current Density A cm-2105% RH, 68 °CRNL, 150ug/cm2, I/C Med iR-corrected RNL 150 MedRNL, 150ug/cm2, I/C Low iR-corrected RNL 150 LowRNL, 150ug/cm2, I/C High iR-corrected RNL 150 HighRNL, 250ug/cm2, I/C Med iR-corrected RNL 250 MedOAN, 150ug/cm2, I/C Med iR-corrected OAN 150 MedV50 150ug/cm2, baseline00.20.40.60.810 0.5 1 1.5 2 2.5 3Potential (V)Current Density A cm-250% RH, 85 °CRNL, 150ug/cm2, I/C Med iR-corrected RNL 150 MedRNL, 150ug/cm2, I/C Low iR corrected RNL 150 LowRNL, 150ug/cm2, I/C High iR corrected RNL 150 HighRNL, 250ug/cm2, I/C Med iR corrected RNL 250 MedOAN, 150ug/cm2, I/C Med iR corrected OAN 150 MedV50 150ug/cm2, baseline136    Figure 6-6: Power density characteristics of the MEAs fabricated with RNL and OAN cathode catalyst (0.15 and 0.25  mgPt cm−2) for H2/Air at 68 ◦C and 85 ◦C, Nafion 211 membrane, 2.5 bara balanced anode and cathode inlet pressure 00.20.40.60.811.20 1 2 3 4Power Density (W cm-2)Current Density (A cm-2)105% RH, 68 °C RNL, 150ug/cm2, I/C Med RNL, 150ug/cm2, I/C LowRNL, 150ug/cm2, I/C High RNL, 250ug/cm2, I/C MedOAN, 150ug/cm2, I/C Med V50 150ug/cm2, baseline00.10.20.30.40.50.60 0.5 1 1.5 2 2.5 3Power density (W cm-2)Current Density (A cm-2)50% RH, 85°CRNL, 150ug/cm2, I/C Med RNL, 150ug/cm2, I/C LowRNL, 150ug/cm2, I/C High RNL, 250ug/cm2, I/C MedOAN, 150ug/cm2, I/C Med V50 150ug/cm2, baseline137  Table 6-6: The shorting resistance and the H2 cross-over currents measured with H2/N2 at 2.5 bara balanced inlet, 105% RH 68 °C and 50% RH 85 °C operating conditions; *measured at 68 °C 50% RH ID 68 °C 85 °C  ix (mA cm-2) Rs (Ω cm-2) ix (mA cm-2) Rs (Ω cm-2) RNL 150 µg cm-2, med I/C 2.923 1992.74 1.953 127.471 RNL 150 µg cm-2, low I/C 4.065 2856.125 0.379 123.981 RNL 150 µg cm-2, high I/C 4.26 6658.567 0.359 193.262 RNL 250 µg cm-2, med I/C 5.078 376.871 0.83 103.874 OAN 150 µg cm-2, med I/C 1.559 38.676 0.46 29.213 V50 150 µg cm-2, 1.1 I/C 6.702 141.4 5.084* 86.1* 6.3.2 Impedance and transport analysis  To understand the performance and effect of the material properties, EIS was used to determine the ohmic and ionic resistances of the catalyst layer. Oxygen transport resistance was measured to determine the pressure dependant and pressure independent resistance. Equivalent circuit using a modified Randles and Warburg element circuit was used to fit the impedance data. Oxygen transport resistance was determined following the technique described in Baker’s publication [206]. The pressure dependent (RPD) portion accounts for the Fickian intermolecular diffusion through large pores (> 100 nm). Pressure independent resistance (RPI) consists of Knudsen diffusion through small pores (< 100 nm) and the ionomer layer surrounding Pt particles. Because the CNF layer is a very open structure with large pores, the majority of the calculated pressure independent resistance was caused by the diffusion through the thin ionomer layer. From the measured RPI shown in Table 6-7, the highest oxygen transport resistance was observed in sample with medium I/C and 250 µg cm-2. The trend clearly shows that increasing the I/C ratio increases RPI from 38.1 to 64.5 s m-1. The experimental results are plausible since TEM ionomer measurements reveal a thickness of 29.46 ± 21.38 and 102.65 ± 84.90 nm for samples with low and high I/C, respectively. Certainly, diffusion through thicker ionomer films would be slower; even so, 250 µg cm-2 sample revealed significantly slower diffusion compared to all 150 µg cm-2 138  samples. This could be attributed to poorer distribution of ionomer and Pt since it was observed that larger clusters formed at high Pt loadings. The larger clusters may cause the ionomer to coat non-uniformly throughout the CNF layer creating a more tortuous path for oxygen transport. Similar results were obtained for OAN sample. RPI was higher in OAN sample, likely because the ionomer was poorly distributed and concentrated in the junctions, as seen in TEM images (Figure 5-18). The ionomer pockets in the OAN samples would drastically increase the oxygen resistance and create a more tortuous path for oxygen transport Table 6-7. Total mass transport resistance (RMT) also reveal consistent trends with small fluctuations between samples. Since the total mass transport resistance includes transport through other components of the fuel cell, it can be assumed that the resistance in other components is consistent across all samples. Thus, only the catalyst layer transport resistance is considered. Subtracting the RPI from RMT gives the estimated RPD, assuming all samples have comparable RPI through respective components. Since CNF were fabricated by the same method, they should have similar porosity and hence similar transport resistance through intermolecular pores. The RMT results appear to be consistent across all samples, although OAN and low I/C RNL appear to have slightly higher RPD. RPD appears to be higher in OAN sample possibly because of the lower porosity caused by CNF adhering together and forming bundles. The structure of the OAN and ionomer pockets in and around the pores would cause RPD and RMT to increase (Table 6-7). Additionally, water trapped in the pores, causes flooding which prevents oxygen gas transport to the active sites. OAN may have challenges with water management since it displayed overall higher RMT compared to other samples. Same observations appeared in RNL 250 µg cm-2 medium I/C sample. The poor distribution of ionomer may induce flooding, hence lowering oxygen transport. Overall, the oxygen transport resistance proved that thin ionomer films improve the mass transport and ionomer distribution is an important parameter.  139  Table 6-7: Total mass transport resistance and pressure independent transport resistance in CAT samples  RNL, 150µg/cm2, I/C Med RNL, 150µg/cm2, I/C Low RNL, 150µg/cm2, I/C High RNL, 250µg/cm2, I/C Med OAN, 150µg/cm2, I/C Med RPI s/m 46.3 38.1 64.5 122.9 105.9 RMT (2.6bara)s/m 109.8 136.5 116.9 174.8 177.5  Table 6-8: Ohmic and ionic resistance measured by EIS for CAT samples under NOC and HOT operating conditions  Operating condition RNL, 150µg/cm2, I/C Med RNL, 150µg/cm2, I/C Low RNL, 150µg/cm2, I/C High RNL, 250µg/cm2, I/C Med OAN, 150µg/cm2, I/C Med Rohm, Ωcm2 68 °C, 30% RH 0.0653 0.0706 0.0718 0.0813 0.0683 68 °C, 100% RH 0.0125 0.0132 0.0126 0.0133 0.0120 √𝑅𝑖𝑜𝑛𝑖𝑐𝐶𝑑, Ωcm2s-1/2 68 °C, 30% RH 2.48820 3.05540 0.9656 4.3135 4.6508 68 °C, 100% RH 0.8637 0.7044 0.245 0.8718 2.6544 Rohm, Ωcm2 85 °C, 50% RH 0.0285 0.0268 0.0289 0.0288 0.0264 √𝑅𝑖𝑜𝑛𝑖𝑐𝐶𝑑, Ωcm2s-1/2 85 °C, 50% RH 1.2641 0.9710 0.3990 1.7856 3.2437   Figure 6-7: Nyquist plot of CAT samples in NOC conditions under H2/N2 conditions (right: zoom in of x-intercepts) -2.0E-01-1.5E-01-1.0E-01-5.0E-020.0E+000 0.025 0.05 0.075 0.1Z"Z' RNL, 150ug/cm2, I/C MedRNL, 150ug/cm2, I/C lowRNL, 150ug/cm2, I/C HighRNL, 250ug/cm2, I/C MediumOAN, 150ug/cm2, I/C Med-2.0E-02-1.5E-02-1.0E-02-5.0E-030.0E+000 0.01 0.02 0.03 0.04Z"Z'140   Figure 6-8: Equivalent Circuit used to fit Nyquist plot for CAT samples   Figure 6-9: |Z| as a function of ω−1/2 for the cell in NOC conditions, slope represents a relationship for Rionic Total ohmic and ionic resistances were measured by EIS for all CAT samples. EIS using H2/N2 method was recorded under HOT and NOC conditions, same as the polarization curves. Nyquist plots are shown in Figure 6-7 in the high frequency region with a close-up of the intercept. The curves were fit using a modified Randles circuit shown in Figure 6-8. The x-axis intercept of the Nyquist plot was determined as the ohmic resistance, which is largely the membrane resistance, of the MEA [207]. The ionic resistance (Rionic) was determined by following the method described in Hou’s work [208]. According to Hou’s method, the results from the Bode plots could be used to determine the Rionic by plotting the magnitude (|Z|) against the frequency (ω). From the |Z| vs ω-1/2 plot in Figure 6-9, the Rionic was determined by fitting a linear trend line and calculating the slope, which equates to√𝑅𝑖𝑜𝑛𝑖𝑐𝐶𝑑 . The presented capacitance in the equivalent circuit is a combination of the interfacial capacitances which were expected to be similar across all samples. R1 C1R3 Ws1Element Freedom Value Error Error %R1 Free(+) 0.013236 N/A N/AC1 Free(+) 0.0018649 N/A N/AR3 Free(+) 0.012601 N/A N/AWs1-R Fixed(X) 0.675 N/A N/AWs1-T Free(+) 0.17332 N/A N/AWs1-P Fixed(X) 0.5 N/A N/AData File:Circuit Model File: C:\Users\Sophia\Documents\ImpedanceModel\FittedWalburg.mdlMode: Run Fitting / Freq. Range (50 - 1000000)Maximum Iterations: 100Optimization Iterations: 0Type of Fitting: ComplexType of Weighting: Data-Modulusy = 0.7044x + 0.0116R² = 0.999300.040.080.120.160.20.240.280 0.2 0.4|Z| Ωcm2ω-1/2141  Since double layer capacitance (Cd) is not readily deconvoluted from the chosen equivalent circuit, it is assumed to be a constant value. Thus, comparing the slopes reveals the Rionic of the MEA system, mainly contribution from catalyst layer. Total Rohm represents the (a) resistance to ion migration within the electrolyte, (b) resistance to electron transport within the cell components (electrodes, gas diffusion layer, and flow field/current collectors), and (c) contact resistances. Although typically, Rionic dominates the resistance contribution since ionic transport is more difficult than electronic charge transport at low frequency. Ohmic (Rohm) and Rionic was measured at 30% and 100% RH at 68 °C. Comparisons between all samples reveal that Rohm was lower at 100% RH and was virtually identical across all samples (Table 6-8). At 30% RH, Rohm across all samples was roughly 0.07 Ωcm2, which is an order of magnitude less than Rionic. The difference between the OAN and RNL Rohm does not appear to be significant, 0.0120 Ωcm2 and 0.0125 Ωcm2 respectively (Table 6-8). The low Rohm was consistent with the CNF ex situ through-plane resistance presented in Chapter 4.2. The trend was consistent with polarization results where RNL150 µg cm-2 with Med I/C obtained the lowest Rohm at both 30% and 100% RH, hence higher performance. For the same sample in HOT conditions (85 °C and 50% RH), Rohm was slightly higher, 0.0285 Ωcm2 and 0.0125 Ωcm2 for HOT and NOC conditions, respectively. Comparison of 30%, 50%, and 100% RH reveals that Rohm decreases with increasing RH, since water is a good conductor and also hydrates the proton membrane. The results suggest Rohm was generally consistent owing to a highly conductive stable CNF structure. Similarly, under NOC conditions Rionic was lower at 100% RH compared to 30% RH. At 100% RH, RNL 150 µg cm-2 with high I/C achieved the lowest Rionic (Table 6-8). Since high I/C results in thick ionomer coating on the fibers, it may have improved bonding with the electrolyte membrane. As a result, the mobility of the protons was increased due to a well-connected ionomer network for efficient transport. OAN 142  sample possessed the highest Rionic at both RHs, indicating the ionomer was not well distributed and was not well-bonded with the membrane. Because the OAN had pockets of ionomer trapped at the fiber junctions, the TPB was not continuous, resulting in a slower path for proton transport. At 68 °C under 30% and 100% RH, Rionic increased as follows: 150 µg cm-2 high I/C < 150 µg cm-2 medium/low I/C and 250 µg cm-2 medium I/C < OAN 150 µg cm-2 medium I/C. The same trend was observed for HOT conditions (85 °C under 50% RH). Again, lower Rionic was seen in the 150 µg cm-2 high I/C sample since the thicker ionomer coating allows for proton transport even in dryer environments. OAN possessed the highest Rionic because of the ionomer pockets and lack of ionomer bridges connecting the fibers through-plane (Table 6-8). In summary, the 150 µg cm-2 high I/C sample showed the highest performance as a result of low ohmic and ionic resistances in both NOC and HOT conditions. However, 150 µg cm-2 medium I/C showed higher oxygen mass transport because of the thinner ionomer layer and better water management at high current densities, allowing efficient diffusion through the ionomer. As a result, both 150 µg cm-2 medium and high I/C have sufficient TPB, translating to improved peak power density compared to traditional V50 CB.        Degradation post fuel cell test   Stability and durability is a major concern for PEMFCs which is outlined by the DoE as an industry goal for 2020. So far, CNF has been favorable as a catalyst layer, therefore it was important to verify the integrity and microstructure post-fuel cell testing. Visual methods via SEM were used to evaluate the condition of the catalyst post fuel cell testing.    6.4.1 Fiber and Pt distribution evaluation  A common issue with conventional carbon black catalyst is the agglomeration of Pt and carbon corrosion during start-up and shut down procedures [209]. It is detrimental to the fuel cell 143  performance because degradation will lead to decreased Pt surface area and lower the reaction kinetics. Therefore, a support material with high electrical conductivity and chemical stability is desirable for maintaining performance. Herein, the fiber and Pt morphology were examined in the SEM before and after fuel cell testing to determine signs of degradation (Figure 6-10, Figure 6-11, and Figure 6-12). From the images shown, CNF has maintained its integrity after fuel cell testing, the fibers did not shatter or change in morphology. As shown, the post-tested CNF layer consists of an interconnected network providing optimal electrical conductivity. The images suggest that CNF were mechanically robust and may be feasibly utilized as a support material without suffering degradation or defects. On the other hand, Pt shows variable amount of corrosion or instability. In Figure 6-10  and Figure 6-12 there were notable changes in the Pt coating. There were regions where Pt was stripped off the CNF support revealing bare CNF underneath. The Pt also became sparse and formed channels where Pt was stripped off or coalesced. The mechanism of Pt degradation and loss of surface area is widely detailed in the literature. Generally, researchers have found that loss of Pt surface area is caused by increased potentials and voltage cycling [210]. For instance, Pt dissolution can occur at voltages higher than 0.8 V vs. RHE by a process similar to Ostwald ripening. Ostwald ripening happens when the Pt dissolution occurs on small nanoparticles and is transported to larger nanoparticles where reduction (redepositon) occurs [210]. The reduction of small Pt nanoparticles is driven by the reduction of surface energy, so nanoparticles with high surface energy are dissolved and redeposited onto larger Pt particles to lower the surface energy [210]. As a result, coarsening of Pt nanoparticles reduces the available active surface area. Moreover, research has found that Pt in solution may diffuse into the proton membrane at higher cathode potentials and exit the cell via waste water on the cathode side [211]. The rate at which Pt coarsening occurs increases with voltage cycling between 0.87 V and 1.20 V. Since the MEA were 144  cycled prior to recording stable polarization curves, it is possible that the CNF/Pt experienced some coarsening and loss of active surface area. It was shown that Pt nanoparticles grew from ~2-3 nm to ~6 nm on CNF after MEA cycling. Because of the nanowire cluster structure, it was challenging to precisely measure the size of individual wires and its’ growth. At most, cluster size was measured but is not a reliable indication of whether Ostwald ripening occurred or not. The average cluster size measured in the long axis is ~22.5 ± 5.47 nm (60 points) and a distribution of the Pt cluster size for RNL sample with 167 µgPt cm-2 is shown in Figure 6-13. Chapter 5.2.1 reported an average cluster size of ~25 nm for similar Pt loading. The manual measurement of cluster size did not show significant increase, however any increment would likely be within the standard deviation.    Figure 6-10: Before and after fuel cell testing; RNL with Pt loading of 167 µg cm-2 (left before, right after) (2 kV, secondary electron mode) 145   Figure 6-11: Before and after fuel cell testing; RNL with Pt loading of 148 µg cm-2 (left before, right after) (2 kV, secondary electron mode)  Figure 6-12: Before and after fuel cell testing; R150M with Pt loading of 157 µg cm-2 (left before, right after) (2 kV, secondary electron mode)  Figure 6-13: Distribution of Pt particle cluster size measured in the long axis post fuel cell testing  0%20%40%60%80%100%120%024681012141618FrequencyDiameter (nm)Frequency Cumulative %146  6.4.2 Cross sectional evaluation   Figure 6-14: Cross sectional evaluation post fuel cell testing: Pt shown on fibers (15 kV, secondary electron mode)  Figure 6-15: Cross sectional evaluation post fuel cell testing: R150H cross section, Pt shown on fibers (15 kV, secondary electron mode)  Figure 6-16: O150H OAN cross section post MAT FC testing (15 kV, secondary electron mode) 147   Figure 6-17: Post fuel cell testing with sufficient Pt distribution and adhesion (left OAN O150H, right RNL R150M) (STEM EDX, 200 kV)  Figure 6-18: Cross section of R150H (left) and R250-T1 (right) showing regions of delamination (15 kV, secondary electron mode)  Figure 6-19: Cross sectional evaluation post fuel cell testing: O150H OAN (left) R150H RNL (right) (15 kV, secondary electron mode) MAT (small fuel cell) tested samples were embedded into resin and imaged in SEM and TEM for cross sectional analysis. The cross section revealed that Pt was securely adhered onto the 148  fiber surface and remained homogenously distributed throughout the thickness (Figure 6-14 and Figure 6-15). OAN cross section also revealed the orientation of fibers remained intact (Figure 6-16). The Pt appears to be homogenously adhered to OAN and did not show signs of degradation and defects (Figure 6-17). Although the Pt and CNF morphology did not change post fuel cell testing, some regions experienced delamination. Selected SEM shown in Figure 6-18 reveal CNF delaminating from the membrane. Although this was only a small section from the overall active surface area, the regions of delamination from the membrane possibly led to increased ohmic overpotentials in R150H and R250-T1. It is possible that R150H (high I/C) performed worse in the MAT testing compared to CAT testing because of delamination and poor contact, resulting in high ohmic resistances. Although delamination was found in the cross sections, it is possible the delamination occurred during sample preparation and embedding into the resin. Delamination likely occurred because the proton membrane shrinks and deforms when the water fully evaporates after testing. During testing, the CNF catalyst should be sufficiently attached to membrane because the cell operates under compression. Certainly, ohmic resistance may be mitigated by improving interfacial properties as discussed by Leeuwner [51]. Generally, the majority of CNF samples did not have delamination issues, with OAN adhering remarkably well (Figure 6-19).   Summary  MEA were fabricated using a modified technique to sandwich the standalone CNF into the assembly. The hot pressed MEA were then prepared for pre-screening in situ fuel cell testing. The pre-screening experiment involved running small (1.13 cm2) MEA in key overpotential regions such as: kinetic, ohmic, and mass transport loss overpotential regions. The effect of material properties were obtained by analyzing the current density achieved in the key potential ranges. Material and design parameters that were tested include: Pt loading, I/C ratio, substrate electrical 149  conductivity, fiber geometry, and Pt deposition method. Polarization curves were recorded for CAT samples under NOC and HOT conditions and evaluated for peak current density and power density. The performance was evaluated by determining the oxygen transport, ionic, and ohmic resistance in the cell through EIS measurements. From the power density plots, the two best performing (~1 W cm-2) samples were 150 µg cm-2 with high and medium I/C. Although, 150 µg cm-2 with medium I/C showed slightly higher Rionic and Rohmic, it had lower RMT and RPI which allowed for efficient oxygen transport at high current densities. Vice versa, 150 µg cm-2 with high I/C possessed low Rionic and Rohmic, signifying good electrical contact and efficient charge transfer, but high RMT and RPI because of the thick ionomer film reducing oxygen diffusion. Through the EIS and polarization data, ionomer distribution and thickness appeared to be the greatest factors in optimizing the fuel cell performance. RNL performed significantly better than the unoptimized V50 because of the open pore structure and large surface area for catalyst reaction sites. Additionally, the large open pore structure in RNL improved water management, allowing sufficient oxygen transport at high current density (mass transport loss overpotential region). The RNL structure established a continuous TPB by addressing the key issues with traditional catalyst layers: decreased Pt surface area, tortuous path, non-homogenous microstructure and distribution, and water management. The performance results indicate RNL performed better than OAN largely because of having a non-optimized OAN fabrication method. Porosity and ionomer distribution was challenging for OAN because of the tendency for fibers to adhere forming large ionomer pockets between layers at the junctions. Therefore, TPB was likely lacking, resulting in high resistances in the cell. Overall, RNL with Pt loading of 150 µg cm-2 showed the highest performance, meaning high Pt utilization and potential to decrease catalyst costs. The current results demonstrated the importance of having a balance between RMT and Rionic and with further 150  understanding and enhanced ionomer deposition methods, the ionomer thickness could be optimized to yield low RMT and Rionic in the MEA. 151   Conclusions and Recommendations  Although fuel cells have been identified as a viable solution for sustainable energy and reduction of greenhouse emissions, the technology has yet to become successfully commercialized due to cost and durability challenges. The carbon black catalyst layer, one of the most expensive components of the fuel cell, has many shortcomings and issues due to the unpredictable microstructure. The random structure leads to inhomogeneous TPB (catalyst, carbon, and ionomer junction), which decreases the efficiency because of having a limited number of reactions sites. Several studies have developed new support materials to improve the catalyst surface area, stability, and TPB [21,22,24,31,64,212]. Although materials such as nanotubes, nanowires, and novel nanoporous structures have shown notable performance enhancements, there remains a lack of understanding between the material and structural properties and the influence it has on the fuel cell’s performance.  Therefore, the objective of this thesis was to design a novel catalyst support with predictable and controllable architecture and material properties to understand the in situ performance differences in the context of certain material properties. In order to understand the effect of organized and random (analogous to CB) microstructure, electrospinning was selected as a method to produce tailorable fiber orientation with high surface area and porosity. In addition, the CL fabrication process involved controlled deposition of Pt and ionomer, yielding predictable morphology and loading. Thus, through this controlled fabrication method, the CL was fully characterized at every stage of the process. By coupling the characterization results with in situ fuel cell testing at NOC and HOT conditions, performance differences were interpreted with greater insight and understood in the context of fiber orientation, fiber electrical conductivity, Pt loading, Pt deposition method, and ionomer loading. Understanding the relationship between fuel cell 152  performance and material properties and architecture could vastly improve the technology by reducing instability, Pt usage, and costs.   Development of nanofiber based catalyst layer  In order to investigate the in situ performance of this novel catalyst support, a bottom-up processing approach to optimize the CNF was employed. First, fabrication of the randomly deposited and orthogonally aligned nanofiber mats were developed through systematic experimental optimization and modelling via response surface method (Chapter 3.3.5). Refinement and validation of this process led to material and structural properties that satisfied industry design targets (Chapter 4). Secondly, the optimized CNF support was successively coated with Pt and ionomer to complete the CL in preparation for MEA assembly. The ex situ characterization of the Pt and ionomer layer revealed controllable Pt distribution, Pt loading, ionomer thickness, and ionomer loading (Chapter 5). Finally, the freestanding CNF catalyst layer was embedded into the MEA for in situ testing (Chapter 6). The effect of fiber geometry, Pt loading, ionomer concentration, Pt deposition method, and support electrical conductivity on fuel cell polarization losses were analyzed from SEM/TEM, CV, EIS, and polarization data. In Chapter 6.3, relationships between microstructure and fuel cell performance were established, fulfilling the research objectives in this study. This fabrication approach led to a systematic method to compare the effect of distinct material and structural characteristics on fuel cell performance metrics. Researching the feasibility and material properties of CNF catalyst supports gave insight into how future catalyst layers may be redesigned to target specific structural or material characteristics for increased efficiency. With regards to process and material design, the outcomes from this study could be applicable to other porous membrane systems including electrolysers, batteries, supercapacitors, and filtration devices. 153   Research outcomes and contributions   The goal of the research was to determine the relationship between the catalyst’s material and geometrical properties and the influence it has on the fuel cell’s polarization losses. Properties of interest include: fiber orientation, support electrical conductivity, Pt loading, Pt deposition method, and ionomer distribution. The conclusions and contributions are presented according to the main objectives of this study (Chapter 1.6), which are summarized in the following sections.      7.2.1 Effect of fiber orientation   To understand the role of geometry and structure of the support, randomly oriented and orthogonally aligned CNF were fabricated and tested in situ. EIS and polarization curves were analyzed to determine the performance associated with fiber orientation. It was found that both geometries had similar total Rohm, suggesting that the OAN was relatively conductive under operating conditions despite subpar ex situ conductivity. Compared to RNL, the ~30% drop in maximum power density was likely the result of the threefold increase in Rionic. Rionic was caused by the poor distribution of the ionomer coating resulting in slow proton transport. Moreover, limited current analysis between OAN and RNL showed OAN achieved slightly lower current density, 3.0 compared to 3.19 A cm-2, suggesting the challenges of flooding and lack of porosity for expelling water. The lower porosity, permeability, and poor ionomer distribution caused the RMT and RPI to increase significantly. It was found that the OAN structure did not enable homogenous distribution nor ionomer bridges between fibers leading to accumulation of ionomer in the junctions and pores. This again confirms the importance of ionomer distribution and coating thickness as it affects both oxygen transport and protonic resistances. Compared to CB baseline, the OAN design achieved an order of magnitude increase in maximum current density as a result of the advantageous CNF architecture and large pores. The study concludes that the geometry and 154  structure greatly influenced the ionomer distribution and hence fuel cell performance in the ohmic and mass transport overpotential regions. Specifically, the lower OAN performance was caused by the increased protonic and oxygen transport resistance as a result of unoptimized mesh porosity and permeability leading to inhomogeneous ionomer distribution – inconsistent TPB. Finally, contrary to our initial hypothesis, the results suggest that the organized OAN structure did not show improved fuel cell performance. However, it is believed that OAN performance may be improved by controlling the pore spacing and optimizing the ionomer deposition method to improve ionomer distribution.  7.2.2 Effect of fiber electrical conductivity  To understand the role and effect of the carbon support, CNF graphitic content was varied by carbonizing at 700 °C and 900 °C. Characterization involved observation by SEM and electrical conductivity measurements. SEM micrographs revealed visually identical and homogenous Pt distribution on both sets of CNF. Ex situ measurements revealed electrical conductivity of RNL is far greater than V50 structure due to the interconnected network of the thin CNF layer. From ex situ results, in-plane and through-plane electrical conductivity measurements confirmed that higher carbonization temperature increased electrical conductivity by nearly two orders of magnitude. The results are also in agreement with XPS, where carbonization at 700 °C and 900 °C achieved a graphitic to amorphous carbon ratio of 1.6 and 3.2, respectively. As expected, samples carbonized at 900 °C achieved higher maximum current density in situ because high electrical conductivity is crucial for charge transfer and decreasing the ohmic overpotential. The pre-screening in situ tests demonstrated that graphitic content and electrical conductivity did not have a significant impact on activation overpotential and mass transport overpotential. However, in the ohmic overpotential region, 700 °C and 900 °C carbonized MEA achieved current density of 0.7 155  and 1.0 A cm-2, respectively. Nonetheless, the CNF (700 °C) MEA outperformed the unoptimized V50 with the same Pt loading by nearly 3.5 times even with lower electrical conductivity. This observation suggests that the CNF, as a catalyst support material, consists of an efficient microstructure compared to CB. Finally the study demonstrated that samples carbonized at 700 °C and 900 °C showed notable performance difference in the ohmic overpotential region. The performance may be further improved by increasing electrical conductivity (carbonization temperature) and reducing interfacial resistances between the individual components of the fuel cell.             7.2.3 Effect of Pt deposition and loading  To understand the role of Pt on the fuel cell performance, two different deposition methods were evaluated with varying Pt loadings. Results indicated that electroless plated RNL achieved five times higher maximum current density than V50 with the same amount of Pt. Moreover, sputter coated RNL with half the Pt loading achieved almost identical maximum current density to V50, revealing the superiority of CNF porous structure even when Pt and ionomer were poorly distributed. Although MAT test evaluation of the Pt loading revealed that 150 and 250 µg cm-2 loadings achieved the same current density (~1 A cm-2 at 650 mV), the CAT cell tests revealed a different trend. In the CAT cell, the maximum power density dropped by 50% when Pt loading was increased to 250 µg cm-2. This observation was unexpected because higher ECSA and activation current density was achieved with higher Pt loadings. This observation indicated that the polarization losses were mostly caused by the ohmic and mass transport losses. The Rionic was particularly high under dryer (30% RH) conditions, suggesting the lack of ionomer bridges and a connected ionomer network. Moreover, RNL sample with 250 µg cm-2 Pt loading exhibited the lowest limiting current out of all the RNL samples, suggesting poor oxygen transport or water 156  management. Since the higher Pt loading sample exhibits more surface area and denser coatings, the deposited ionomer loading may not have been sufficient to form the TPB. Therefore, the results suggest that the ratio between I/Pt might be more critical than I/C ratio. This imbalanced ratio between ionomer and Pt resulted in higher Rionic, RMT, and RPI decreasing the overall fuel cell performance. Therefore, this study furthermore shows that properties such as ECSA or Pt loading cannot be considered in isolation when investigating the performance. In this study, RNL with 150 µg cm-2 Pt loading demonstrated exceptional power density performance which means less Pt could be used to achieve satisfactory performances. Nonetheless, the complex combination of TPB and the material’s porous structure are of utmost importance and should be further optimized to reduce Pt usage and costs.  7.2.4 Effect of ionomer distribution and loading  To understand the effect of ionomer loading and distribution on fuel cell performance, three different I/C loadings were tested in situ. Both sets of in situ testing confirmed that as concentration was further increased, power density showed a diminishing return. Ionomer concentration affects ohmic overpotential and mass transport overpotential regions because of increased oxygen transport resistance and Rohm. Low ionomer concentration showed a significant disadvantage in power density (30% drop) when compared to medium concentration (1 W cm-2). When ionomer was insufficient, the Rionic increased due to poor proton transport. However, very thick ionomer coatings increased the oxygen transport resistance. Hence, the EIS and polarization results were in agreement and confirmed that fuel cell performance was highly influenced by ionomer distribution and thickness. The study concludes, medium and high I/C achieved the highest performance, up to 5x higher maximum power density than conventional V50 CB. The main challenge was utilizing the appropriate amount of ionomer to encourage TPB while maintaining free and direct pathways 157  for gas transport. Although the current results provided insight into the importance of ionomer distribution and thickness, rigorous optimization of the deposition process would be necessary to achieve an efficient balance between oxygen transport resistance and protonic resistance. Finally, the study concludes ionomer distribution has significant influence on fuel cell performance and was a factor of carbon support geometry, Pt loading, and Pt/ionomer ratio. The optimization of ionomer in the CNF based MEA would balance RMT, Rohmic, and Rionic, ultimately reducing potential losses in the ohmic and mass transport overpotential regions – improving fuel cell efficiency and reducing material usage and costs.  Ultimately, based on the aforementioned conclusions, ionomer distribution and loading had the greatest influence on fuel cell performance. Regarding optimization of CNF layers, the connected ionomer network should be optimized to achieve a balance between Rohm, Rionic, and RMT to reduce polarization losses in the ohmic and mass transport regions. Compared to the unoptimized baseline CB, the results demonstrated that CNF catalyst support increased maximum power density even without an optimized TPB due to having a well-connected network and porous structure. With regards to material and structural properties, the established relationships provide a guideline for enhancing future catalyst designs to improve durability and lower costs.                   Health and safety concerns During the progression of this research, safety and health concerns were raised by AFCC regarding the handling of nanomaterials. Industry laboratories often have serious procedures and protocols to ensure a safe and standardized working environment. For this reason, new materials introduced to the AFCC laboratories must pass health and safety requirements. Unlike CBs, which 158  have well-documented hazard prevention procedures, CNFs have not been previously used in the AFCC laboratories, therefore new procedures were developed to decrease risks and exposure. It is well known that nanostructures raise occupational health concerns because it is not well documented how the nanomaterials would interact with human bodies. Particularly, carbon is considered a carcinogen since it has been shown to cause lung cancer in animals [213]. Because of the health risks involved, manufacturers implement strict safety procedures to reduce exposure to airborne particles. The National Institute for Occupational Safety and Health (NIOSH) outlined a recommended airborne exposure limit (REL) of 3.5 mg m-3 averaged over 10 hours for carbon black [214]. Electrospun CNF, on the other hand, has not been rated by NIOSH. The most comparable material would be CNT and CVD CNF (5-30 µm length). For CNF, NIOSH recommends an exposure of below 1 μg m-3 for an 8 hour Time Weighted Average (TWA),  respirable fraction (elemental carbon) during a 40 hour work week [215]. Because this is for short CNFs, it is not representative of electrospun fibers. Electrospinning forms fibers that are several centimetres long. In the case of electrospun CNF, the fibers are in a cohesive non-woven fabric structure with virtually zero fallout. The perceived risk of exposure during the processing of MEA is when the CNF layer is cut with scissors. To reduce exposure in the current study, CNF samples were cut and handled under a fume hood or well ventilated area. It was estimated that the current experiment and procedure produces 0.53 μg m-3 per 8 hour of airborne substance. Compared to carbon black, which is an easily airborne powder, electrospun CNF is safer and has a much lower risk of airborne particle exposure. Therefore, following the developed working procedures, it was found that CNFs were relatively safe and met AFCC safety standards. For these reasons, it is worthwhile to consider replacing carbon black catalysts with electrospun CNF to lower occupational health risks in the future.         159   Recommendations for future work   From this study, CNF was demonstrated as a feasible catalyst support with controllable structural and material properties. To build upon this work, there are several opportunities for further research to improve performance and applicability in this area.   Geometrical effect  Although OAN was expected to improve mass transport and aid with water management, the current results are inconclusive. Because the OAN structure was fabricated using a different electrospinning apparatus which displays an altered electrospinning electric field, the process parameters should be optimized separately from RNL electrospinning. Additional properties of interest include fiber spacing and pore size. It is plausible that the OAN did not achieve high porosity and mass transport because the fibers did not have optimal pore spacing. As a result, future work should focus on developing a fabrication method which deposits orthogonal fibers simultaneously with controllable fiber diameter, pore size, and pore spacing. Prior work has primarily demonstrated feasibility of controlling alignment direction, however the results are limited to small areas (2 cm2) and minimal layer thickness [90,112,216,217]. The orthogonal layer should be cohesively deposited to reduce delamination and contact resistance between layers. In addition, the study could determine the effect of the degree of alignment and pore size/spacing on diffusivity and mass transport. The optimized orthogonal structure may show improved ionomer distribution leading to higher power densities. Optimized RNL showed significant improvement in maximum power density compared to V50. It is apparent that geometrical effect is a factor. Whether OAN or RNL achieves higher power density requires further development of the orthogonal deposition technique and systematic optimization of the process parameters. The effect 160  of oriented fibers as it relates to the performance of the catalyst samples in hydrogen fuel cells warrants further scrutiny in a future study.  Protonic conductivity  In addition to ex situ electrical conductivity, an apparatus may be designed to measure protonic conductivity. Since ionomer plays a critical role in the catalyst design, it is necessary to create a method to characterize the protonic conductivity ex situ. The standardised ex situ measurement would assess the quality of the ionomer coating without destroying the sample. Current methods involve measuring the entire MEA and separating the ionic response from the EIS data [218,219]. Measuring the entire MEA as a screening method to understand ionomer deposition is a costly and destructive technique. Another method measures the in-plane protonic conductivity of the polymer membrane using a four point probe humidity controlled apparatus [220]. This may be an applicable method, however, the method has been primarily used for polymer membranes rather than catalyst layers. Additionally, electronic and protonic conductivities should be separated in the raw data. As a first approximation in-plane measurements could be useful however, through-plane is a more accurate representation of the fuel cell operating conditions. Measuring the protonic conductivity would allow for quick evaluation of the ionomer layer when investigating advanced deposition methods.        Durability and stability testing  Fuel cell operating conditions should be optimized for CNF MEAs because the microstructure is vastly different from traditional catalyst layers. Applying CNF MEAs to fuel cell stacks optimized for CB may not give an accurate representation of the CNFs potential. In addition to evaluating the CNF catalyst in situ, the experiment should include stability and durability testing. Durability and stability is a major problem in the current catalyst supports because of carbon 161  agglomeration and Pt dissolution. To confirm the feasibility of CNF as catalyst supports, in situ fuel cell testing should be scanned over 10,000 cycles to evaluate the durability and stability [221,222]. From the current results, post-fuel cell testing Pt and fiber morphology did not reveal visible degradation or changes. However, the SEM images indicated some delamination from the membrane which may have been caused by sample preparation or the harsh fuel cell environment. Therefore, to have a deeper understanding of the CNF and Pt material properties in situ, stability testing should be performed. Additional tests could be conducted to evaluate the effect of surface treatment or post processing on durability.    Imaging technique  Although the obtained STEM micrographs displayed presence of ionomer film around the fibers, it was a challenging method for imaging due to electron beam sensitivity. As a result, some ionomer samples were melted and destroyed during high resolution imaging. Refining operation parameters and imaging regions quickly may reduce melting; however other methods should be included for accurate representation. Cryogenic SEM may reduce ionomer melting as it is often used for samples with beam sensitivity or instability issues. Additional imaging techniques could include atomic force microscope infrared-spectroscopy (AFM-IR), which can give spatial resolution down to < 10 nm. A standardized procedure should be implemented for cross sectioning to ensure preservation of the imaging surface. However, it may be difficult to observe a large cross sectional area with these techniques. The ideal technique would allow for in situ or hydrated environment imaging because ionomer swells in humid conditions. Understanding the ionomer distribution and thickness is necessary to develop improved catalyst structures in the future. 162  Bibliography [1] BP, BP Statistical Review of World Energy 2017, London, 2017. https://www.bp.com/content/dam/bp/en/corporate/pdf/energy-economics/statistical-review-2017/bp-statistical-review-of-world-energy-2017-full-report.pdf (accessed May 22, 2018). [2] U.S.E.P.A. EPA, Sources of Greenhouse Gas Emissions, (2016). https://www.epa.gov/ghgemissions/sources-greenhouse-gas-emissions (accessed May 8, 2018). [3] U.S. Department of energy, Fuel Cell Technologies Office Multi-Year Research, Development, and Demonstration Plan, 2016. [4] Charging electric vehicles (EVs), Minist. Transp. (2018). http://www.betterplace.com/solution/charging/ (accessed May 23, 2018). [5] J. Wang, H. Wang, Y. Fan, Techno-Economic Challenges of Fuel Cell Commercialization, Engineering. 4 (2018) 352–360. doi:10.1016/J.ENG.2018.05.007. [6] J. Wang, Barriers of scaling-up fuel cells: Cost, durability and reliability, Energy. 80 (2015) 509–521. doi:10.1016/j.energy.2014.12.007. [7] K. Stolzenburg, V. Tsatsami, H. Grubel, Lessons learned from infrastructure operation in the CUTE project, Int. J. Hydrogen Energy. 34 (2009) 7114–7124. doi:10.1016/j.ijhydene.2008.06.035. [8] S.D. Stephens-Romero, T.M. Brown, J.E. Kang, W.W. Recker, G.S. Samuelsen, Systematic planning to optimize investments in hydrogen infrastructure deployment, Int. J. Hydrogen Energy. 35 (2010) 4652–4667. doi:10.1016/j.ijhydene.2010.02.024. [9] D. Haeseldonckx, W. D’haeseleer, Concrete transition issues towards a fully-fledged use of hydrogen as an energy carrier: Methodology and modelling, (2011). doi:10.1016/j.ijhydene.2011.01.113. [10] A. Serov, C. Kwak, Review of Non-platinum Anode Catalysts for DMFC and PEMFC Application, Appl. Catal. B Environ. 90 (2009) 313–320. doi:10.1016/j.apcatb.2009.03.030. [11] R. Singh, R. Awasthi, C. Sharma, Review : An Overview of Recent Development of Platinum- Based Cathode Materials for Direct Methanol Fuel Cells, Int. J. Electrochem. Sci. 9 (2014) 5607–5639. http://electrochemsci.org/papers/vol9/91005607.pdf. [12] H. Lv, D. Li, D. Strmcnik, A.P. Paulikas, N.M. Markovic, V.R. Stamenkovic, Recent advances in the design of tailored nanomaterials for efficient oxygen reduction reaction, Nano Energy. 29 (2016) 149–165. doi:10.1016/J.NANOEN.2016.04.008. [13] R. Othman, A.L. Dicks, Z. Zhu, Non precious metal catalysts for the PEM fuel cell cathode, Int. J. Hydrogen Energy. 37 (2012) 357–372. doi:10.1016/j.ijhydene.2011.08.095. [14] A. Bonnefont, P. Ruvinskiy, M. Rouhet, A. Orfanidi, S. Neophytides, E. Savinova, Advanced catalytic layer architectures for polymer electrolyte membrane fuel cells, Wiley Interdiscip. Rev. Energy Environ. 3 (2014) 505–521. doi:10.1002/wene.110. [15] E. Antolini, Palladium in fuel cell catalysis, Energy Environ. Sci. 2 (2009) 915. doi:10.1039/b820837a. [16] S. Sharma, B.G. Pollet, Support materials for PEMFC and DMFC electrocatalysts—A review, J. Power Sources. 208 (2012) 96–119. doi:10.1016/j.jpowsour.2012.02.011. 163  [17] A. Wilson, J. Marcinkoski, D. Papageorgopoulos, R. Ahluwalia, B. James, C. Houchins, J. Huya-Kouadio, Fuel Cell System Cost -2016, Washington, 2016. https://www.hydrogen.energy.gov/pdfs/16020_fuel_cell_system_cost_2016.pdf (accessed May 23, 2018). [18] G. Hoogers, Fuel Cell Components and Their Impact on Performance, in: G. Hoogers (Ed.), Fuel Cell Technol. Handb., 2003. doi:10.1201/9781420041552.ch4. [19] H. Lv, D. Li, D. Strmcnik, A.P. Paulikas, N.M. Markovic, V.R. Stamenkovic, Recent advances in the design of tailored nanomaterials for efficient oxygen reduction reaction, Nano Energy. 29 (2016) 149–165. doi:10.1016/j.nanoen.2016.04.008. [20] I.E.L. Stephens, A.S. Bondarenko, U. Gr?nbjerg, J. Rossmeisl, I. Chorkendorff, M. Odelius, H. Ogasawara, L.G.M. Pettersson, A. Nilsson, T.-H. Lim, R. Adzic, G. Karapetrov, D. Strmcnik, N.M. Markovic, V.R. Stamenkovic, J.E. McGrath, M. Inaba, K. Miyatake, M. Hori, K. Ota, Z. Ogumi, S. Miyata, A. Nishikata, Z. Siroma, Y. Uchimoto, K. Yasuda, K.I. Kimijima, N. Iwashita, Understanding the electrocatalysis of oxygen reduction on platinum and its alloys, Energy Environ. Sci. 5 (2012) 6744. doi:10.1039/c2ee03590a. [21] E. Antolini, Composite materials: An emerging class of fuel cell catalyst supports, Appl. Catal. B Environ. 100 (2010) 413–426. doi:10.1016/j.apcatb.2010.08.025. [22] E. Antolini, Carbon supports for low-temperature fuel cell catalysts, Appl. Catal. B Environ. 88 (2009) 1–24. doi:10.1016/j.apcatb.2008.09.030. [23] Y. Wang, D.P. Wilkinson, J. Zhang, Noncarbon Support Materials for Polymer Electrolyte Membrane Fuel Cell Electrocatalysts, Chem. Rev. 111 (2011) 7625–7651. [24] O.-H. Kim, Y.-H. Cho, S.H. Kang, H.-Y. Park, M. Kim, J.W. Lim, D.Y. Chung, M.J. Lee, H. Choe, Y.-E. Sung, Ordered macroporous platinum electrode and enhanced mass transfer in fuel cells using inverse opal structure, Nat. Commun. 4 (2013) 1–9. doi:10.1038/ncomms3473. [25] D. Thompsett, Catalysts for Proton Exchange Membrane Fuel Cell, in: G. Hoogers (Ed.), Fuel Cell Technol. Handb., 2002. doi:10.1201/9781420041552.ch6. [26] L.M. Roen, C.H. Paik, T.D. Jarvi, Electrocatalytic Corrosion of Carbon Support in PEMFC Cathodes, Electrochem. Solid-State Lett. 7 (2004) A19. doi:10.1149/1.1630412. [27] R. Borup, J. Meyers, B. Pivovar, Y.S. Kim, R. Mukundan, N. Garland, D. Myers, M. Wilson, F. Garzon, D. Wood, P. Zelenay, K. More, K. Stroh, T. Zawodzinski, J. Boncella, J.E. McGrath, M. Inaba, K. Miyatake, M. Hori, K. Ota, Z. Ogumi, S. Miyata, A. Nishikata, Z. Siroma, Y. Uchimoto, K. Yasuda, K.-I. Kimijima, N. Iwashita, Scientific aspects of polymer electrolyte fuel cell durability and degradation., Chem. Rev. 107 (2007) 3904–51. doi:10.1021/cr050182l. [28] M. Cai, M.S. Ruthkosky, B. Merzougui, S. Swathirajan, M.P. Balogh, S.H. Oh, Investigation of thermal and electrochemical degradation of fuel cell catalysts, J. Power Sources. 160 (2006) 977–986. doi:10.1016/j.jpowsour.2006.03.033. [29] R. O’Hayre, D.M. Barnett, F.B. Prinz, The Triple Phase Boundary, J. Electrochem. Soc. 152 (2005) A439. doi:10.1149/1.1851054. [30] M.E. Eikerling, A. Kulikovsky, Catalyst Layer Structure and Operation, in: Polym. Electrolyte Fuel Cells Phys. Princ. Mater. Oper., 19th ed., CRC Press 2014, 2015: pp. 155–262. [31] S. Tang, G. Sun, J. Qi, S. Sun, J. Guo, Q. Xin, G.M. Haarberg, Review of New Carbon 164  Materials as Catalyst Supports in Direct Alcohol Fuel Cells, Chinese J. Catal. 31 (2010) 12–17. doi:10.1016/S1872-2067(09)60034-6. [32] L. Calvillo, M. Gangeri, S. Perathoner, G. Centi, R. Moliner, M.J. Lázaro, Synthesis and performance of platinum supported on ordered mesoporous carbons as catalyst for PEM fuel cells: Effect of the surface chemistry of the support, Int. J. Hydrogen Energy. 36 (2011) 9805–9814. doi:10.1016/j.ijhydene.2011.03.023. [33] S. Song, Y. Liang, Z. Li, Y. Wang, R. Fu, D. Wu, P. Tsiakaras, Effect of pore morphology of mesoporous carbons on the electrocatalytic activity of Pt nanoparticles for fuel cell reactions, Appl. Catal. B Environ. 98 (2010) 132–137. doi:10.1016/j.apcatb.2010.05.021. [34] A.L. Dicks, The role of carbon in fuel cells, J. Power Sources. 156 (2006) 128–141. doi:10.1016/j.jpowsour.2006.02.054. [35] J.-H.H. Park, Y.-W.W. Ju, S.-H.H. Park, H.-R.R. Jung, K.-S.S. Yang, W.-J.J. Lee, Effects of electrospun polyacrylonitrile-based carbon nanofibers as catalyst support in PEMFC, J. Appl. Electrochem. 39 (2009) 1229–1236. doi:10.1007/s10800-009-9787-4. [36] Z. Lin, L. Ji, O. Toprakci, W. Krause, X. Zhang, Electrospun carbon nanofiber-supported Pt-Pd alloy composites for oxygen reduction, J. Mater. Res. 25 (2010) 1329–1335. doi:10.1557/jmr.2010.0163. [37] A. Smirnova, X. Dong, H. Hara, N.M. Sammes, New generation of catalyst layers for PEMFCs based on carbon aerogel supported Pt catalyst (CASPC), Fuel Cell Technol. (2006) 237–251. doi:10.1007/1-84628-207-1_7. [38] X. Sun, M.S. Saha, Nanotubes, nanofibers and nanowires as supports for catalysts, in: PEM Fuel Cell Electrocatal. Catal. Layers Fundam. Appl., Springe, London, 2008: pp. 655–714. doi:10.1007/978-1-84800-936-3_14. [39] Y. Wang, J. Jin, S. Yang, G. Li, J. Qiao, Highly active and stable platinum catalyst supported on porous carbon nanofibers for improved performance of PEMFC, Electrochim. Acta. 177 (2015) 181–189. doi:10.1016/j.electacta.2015.01.134. [40] G.S. Chai, S.B. Yoon, J.-S. Yu, Highly efficient anode electrode materials for direct methanol fuel cell prepared with ordered and disordered arrays of carbon nanofibers, Carbon N. Y. 43 (2005) 3028–3031. doi:10.1016/j.carbon.2005.05.041. [41] P. Ramesh, M.E. Itkis, J.M. Tang, R.C. Haddon, SWNT - MWNT Hybrid Architecture for Proton Exchange Membrane Fuel Cell Cathodes, (2008) 9089–9094. [42] K. Okano, Development Histories: Fuel Cell Technologies, in: Springer, Tokyo, 2016: pp. 93–115. doi:10.1007/978-4-431-56042-5_5. [43] N.H. Behling, Chapter 7 – History of proton exchange membrane fuel cells and direct methanol fuel cells, in: Fuel Cells Curr. Technol. Challenges Futur. Res. Needs, 1st ed., Newnes, 2013: pp. 423–600. doi:10.1016/B978-0-444-56325-5.00004-1. [44] J. Larminie, A. Dicks, Fuel Cell Systems Explained, 2001. doi:10.1016/S0378-7753(00)00571-1. [45] Our History - Power to Change the World | Ballard Power, (n.d.). http://ballard.com/about-ballard/our-history (accessed August 26, 2018). [46] S. Curtin, J. Gangi, Fuel Cell Technologies Market Report 2016, Washington, 2017. https://www.energy.gov/sites/prod/files/2017/10/f37/fcto_2016_market_report.pdf (accessed May 8, 2018). [47] F. Barbir, PEM Fuel Cells: Theory and Practice, 2005. doi:10.1016/b978-012078142-3/50002-1. 165  [48] W. Gu, D.R. Baker, Y. Liu, H.A. Gasteiger, Proton exchange membrane fuel cell (PEMFC) down-the-channel performance model, in: W. Vielstich, A. Lamm, H.A. Gasteiger, H. Yokokawa (Eds.), Handb. Fuel Cells, John Wiley & Sons, Ltd, Chichester, UK, 2010: pp. 1–27. [49] F. Barbir, PEM Fuel Cells, in: N.M. Sammes (Ed.), Fuel Cell Technol. Reach. Towar. Commer., 1st ed., Springer-Verlag London, London, 2006: pp. 27–51. doi:10.1007/1-84628-207-1. [50] M. Andisheh-Tadbir, M. El Hannach, E. Kjeang, M. Bahrami, An analytical relationship for calculating the effective diffusivity of micro-porous layers, Int. J. Hydrogen Energy. 40 (2015) 10242–10250. doi:10.1016/j.ijhydene.2015.06.067. [51] M.J. Leeuwner, D.P. Wilkinson, E.L. Gyenge, Novel Graphene Foam Microporous Layers for PEM Fuel Cells: Interfacial Characteristics and Comparative Performance, Fuel Cells. 15 (2015) 790–801. doi:10.1002/fuce.201500031. [52] M.J.L. Janse van Vuuren, Novel microporous layers with improved interfacial characteristics for PEM fuel cells, University of British Columbia, 2017. doi:10.14288/1.0360782. [53] C. a. Reiser, L. Bregoli, T.W. Patterson, J.S. Yi, J.D. Yang, M.L. Perry, T.D. Jarvi, A Reverse-Current Decay Mechanism for Fuel Cells, Electrochem. Solid-State Lett. 8 (2005) A273. doi:10.1149/1.1896466. [54] E. Yli-Rantala,  a. Pasanen, P. Kauranen, V. Ruiz, M. Borghei, E. Kauppinen,  a. Oyarce, G. Lindbergh, C. Lagergren, M. Darab, S. Sunde, M. Thomassen, S. Ma-Andersen, E. Skou., Graphitised Carbon Nanofibres as Catalyst Support for PEMFC, Fuel Cells. 11 (2011) 715–725. doi:10.1002/fuce.201000180. [55] S.H. Joo, H.I. Lee, D.J. You, K. Kwon, J.M.H. Kim, Y.S. Choi, M. Kang, J.M.H. Kim, C. Pak, H. Chang, D. Seung, Ordered mesoporous carbons with controlled particle sizes as catalyst supports for direct methanol fuel cell cathodes, Carbon N. Y. 46 (2008) 2034–2045. doi:10.1016/j.carbon.2008.08.015. [56] Y. Yuan, J. a. Smith, G. Goenaga, D.-J. Liu, Z. Luo, J. Liu, Platinum decorated aligned carbon nanotubes: Electrocatalyst for improved performance of proton exchange membrane fuel cells, J. Power Sources. 196 (2011) 6160–6167. doi:10.1016/j.jpowsour.2011.03.026. [57] R. Wang, D.C. Higgins, D.U. Lee, S. Prabhudev, F.M. Hassan, V. Chabot, G. Lui, G. Jiang, J.Y. Choi, L. Rasenthiram, J. Fu, G. Botton, Z. Chen, Biomimetic design of monolithic fuel cell electrodes with hierarchical structures, Nano Energy. 20 (2016) 57–67. doi:10.1016/j.nanoen.2015.11.033. [58] Z. Xia, S. Wang, L. Jiang, H. Sun, S. Liu, X. Fu, B. Zhang, D. Sheng Su, J. Wang, G. Sun, Bio-inspired Construction of Advanced Fuel Cell Cathode with Pt Anchored in Ordered Hybrid Polymer Matrix, Sci. Rep. 5 (2015) 16100. doi:10.1038/srep16100. [59] K.-D. Kreuer, A. Rabenau, W. Weppner, Vehicle Mechanism, A New Model for the Interpretation of the Conductivity of Fast Proton Conductors, Angew. Chemie Int. Ed. English. 21 (1982) 208–209. doi:10.1002/anie.198202082. [60] C.J.T. de Grotthuss, Sur la décomposition de l’eau et des corps qu’elle tient en dissolution à l’aide de l’électricité galvanique, Ann. Chim. 58 (1806) 54–73. [61] Z. Liu, X. Lin, J.Y. Lee, W. Zhang, M. Han, L.M. Gan, Preparation and characterization of platinum-based electrocatalysts on multiwalled carbon nanotubes for proton exchange 166  membrane fuel cells, Langmuir. 18 (2002) 4054–4060. doi:10.1021/la0116903. [62] C. Du, M. Chen, X. Cao, G. Yin, P. Shi, A novel CNT@SnO2 core–sheath nanocomposite as a stabilizing support for catalysts of proton exchange membrane fuel cells, Electrochem. Commun. 11 (2009) 496–498. doi:10.1016/j.elecom.2008.12.034. [63] W. Li, M. Waje, Z. Chen, P. Larsen, Y. Yan, Platinum nanopaticles supported on stacked-cup carbon nanofibers as electrocatalysts for proton exchange membrane fuel cell, Carbon N. Y. 48 (2010) 995–1003. doi:10.1016/j.carbon.2009.11.017. [64] W. Xiong, F. Du, Y. Liu, A. Perez, M. Supp, 3-D Carbon Nanotube Structures Used as High Performance Catalyst for, (2010) 15839–15841. [65] B.O. Boskovic, V. Stolojan, R.U.A. Khan, S. Haq, S.R.P. Silva, Large-area synthesis of carbon nanofibres at room temperature, Nat. Mater. 1 (2002) 165–168. doi:10.1038/nmat755. [66] S.L. Knupp, W. Li, O. Paschos, T.M. Murray, J. Snyder, P. Haldar, The effect of experimental parameters on the synthesis of carbon nanotube/nanofiber supported platinum by polyol processing techniques, Carbon N. Y. 46 (2008) 1276–1284. doi:10.1016/j.carbon.2008.05.007. [67] N.M. Rodriguez, M.S. Kim, R.T.K. Baker, Carbon Nanofibers - a Unique Catalyst Support Medium, J. Phys. Chem. 98 (1994) 13108–13111. doi:Doi 10.1021/J100101a003. [68] Z. Lin, L. Ji, W.E. Krause, X. Zhang, Synthesis and electrocatalysis of 1-aminopyrene-functionalized carbon nanofiber-supported platinum-ruthenium nanoparticles, J. Power Sources. 195 (2010) 5520–5526. doi:10.1016/j.jpowsour.2010.03.059. [69] S. Kang, S. Lim, D.H. Peck, S.K. Kim, D.H. Jung, S.H. Hong, H.G. Jung, Y. Shul, Stability and durability of PtRu catalysts supported on carbon nanofibers for direct methanol fuel cells, Int. J. Hydrogen Energy. 37 (2012) 4685–4693. doi:Doi 10.1016/J.Ijhydene.2011.04.119. [70] D. Sebastián, M.J. Lázaro, I. Suelves, R. Moliner, V. Baglio, A. Stassi, A.S. Aric, The influence of carbon nanofiber support properties on the oxygen reduction behavior in proton conducting electrolyte-based direct methanol fuel cells, in: Int. J. Hydrogen Energy, 2012: pp. 6253–6260. doi:10.1016/j.ijhydene.2011.07.004. [71] T. Maiyalagan, Pt-Ru nanoparticles supported PAMAM dendrimer functionalized carbon nanofiber composite catalysts and their application to methanol oxidation, in: J. Solid State Electrochem., 2009: pp. 1561–1566. doi:10.1007/s10008-008-0730-0. [72] C.A. Bessel, K. Laubernds, N.M. Rodriguez, R.T.K. Baker, Graphite nanofibers as an electrode for fuel cell applications, J. Phys. Chem. B. 105 (2001) 1115–1118. doi:10.1021/jp003280d. [73] J.-S. Zheng, X.-S. Zhang, P. Li, J. Zhu, X.-G. Zhou, W.-K. Yuan, Effect of carbon nanofiber microstructure on oxygen reduction activity of supported palladium electrocatalyst, Electrochem. Commun. 9 (2007) 895–900. doi:10.1016/j.elecom.2006.12.006. [74] J.S. Zheng, X.S. Zhang, P. Li, X.G. Zhou, W.K. Yuan, Microstructure effect of carbon nanofiber on electrocatalytic oxygen reduction reaction, Catal. Today. 131 (2008) 270–277. doi:Doi 10.1016/J.Cattod.2007.10.104. [75] J.S. Zheng, M.X. Wang, X.S. Zhang, Y.X. Wu, P. Li, X.G. Zhou, W.K. Yuan, Platinum/carbon nanofiber nanocomposite synthesized by electrophoretic deposition as electrocatalyst for oxygen reduction, J. Power Sources. 175 (2008) 211–216. 167  doi:10.1016/j.jpowsour.2007.09.058. [76] G. Álvarez, F. Alcaide, P.L. Cabot, M.J. Lázaro, E. Pastor, J. Solla-Gullón, Electrochemical performance of low temperature PEMFC with surface tailored carbon nanofibers as catalyst support, Int. J. Hydrogen Energy. 37 (2012) 393–404. doi:10.1016/j.ijhydene.2011.09.055. [77] Y. Ding, M. Chen, J. Erlebacher, Metallic mesoporous nanocomposites for electrocatalysis, J. Am. Chem. Soc. 126 (2004) 6876–6877. doi:10.1021/ja0320119. [78] L. Calvillo, M.J. Lázaro, E. García-Bordejé, R. Moliner, P.L. Cabot, I. Esparbé, E. Pastor, J.J. Quintana, Platinum supported on functionalized ordered mesoporous carbon as electrocatalyst for direct methanol fuel cells, J. Power Sources. 169 (2007) 59–64. doi:10.1016/j.jpowsour.2007.01.042. [79] L. Zhang, Carbon Nanofibers from Electrospinning – A Promising Electrode Material for Energy Storage and Conversion, Int. J. Nano Stud. Technol. III (2014) 1–2. [80] Z. Lin, Y. Yao, X. Zhang, Electrospun Nanofibers for Design and Fabrication of Electrocatalysts and Electrolyte Membranes for Fuel cells, in: B. Ding, J. Yu (Eds.), Electrospun Nanofibers Energy Environ. Appl., 1st ed., Springer-Verlag Berlin Heidelberg, 2014: p. 525. doi:10.1007/978-3-642-54160-5. [81] C. Ma, Y. Li, J. Shi, Y. Song, L. Liu, High-performance supercapacitor electrodes based on porous flexible carbon nanofiber paper treated by surface chemical etching, Chem. Eng. J. 249 (2014) 216–225. doi:10.1016/j.cej.2014.03.083. [82] M. Inagaki, Y. Yang, F. Kang, Carbon nanofibers prepared via electrospinning., Adv. Mater. 24 (2012) 2547–66. doi:10.1002/adma.201104940. [83] M. Li, G. Han, B. Yang, Fabrication of the catalytic electrodes for methanol oxidation on electrospinning-derived carbon fibrous mats, Electrochem. Commun. 10 (2008) 880–883. doi:10.1016/j.elecom.2008.04.002. [84] S. Cavaliere, S. Subianto, I. Savych, D.J. Jones, J. Rozière, Electrospinning: designed architectures for energy conversion and storage devices, Energy Environ. Sci. 4 (2011) 4761. doi:10.1039/c1ee02201f. [85] S. Chawla, M. Naraghi, A. Davoudi, Effect of twist and porosity on the electrical conductivity of carbon nanofiber yarns., Nanotechnology. 24 (2013) 255708. doi:10.1088/0957-4484/24/25/255708. [86] S. Sukigara, M. Gandhi, J. Ayutsede, M. Micklus, F. Ko, Regeneration of Bombyx mori silk by electrospinning. Part 2. Process optimization and empirical modeling using response surface methodology, Polymer (Guildf). 45 (2004) 3701–3708. doi:10.1016/j.polymer.2004.03.059. [87] F.K. Ko, Nanomaterials Handbook, illustrate, CRC Press, 2006. doi:1420004018. [88] F.K. Ko, Y. Wan, Introduction to Nanofiber Materials, Cambridge University Press, 2014. [89] L. Lin, Y. Li, F.K. Ko, Fabrication and Properties of Lignin Based Carbon Nanofiber, J. Fiber Bioeng. Informatics. 6 (2013) 335–347. doi:10.3993/jfbi12201301. [90] D. Li, Y. Xia, Electrospinning of nanofibers: Reinventing the wheel?, Adv. Mater. 16 (2004) 1151–1170. doi:10.1002/adma.200400719. [91] K. Nasouri, A.M. Shoushtari, M.R.M. Mojtahedi, Evaluation of effective electrospinning parameters controlling polyvinylpyrrolidone nanofibers surface morphology via response surface methodology, Fibers Polym. 16 (2015) 1941–1954. doi:10.1007/s12221-015-5263-4. 168  [92] C. RAO, D. Trivedi, Chemical and electrochemical depositions of platinum group metals and their applications, Coord. Chem. Rev. 249 (2005) 613–631. doi:10.1016/j.ccr.2004.08.015. [93] M. Brodt, R. Wycisk, P. Pintauro, T. Han, N. Dale, K. Adjemian, Nanofiber Fuel Cell Electrodes I. Fabrication and Performance with Commercial Pt/C Catalysts, ECS Trans. 58 (2013) 381–390. [94] W. Zhang, P.N. Pintauro, High-performance nanofiber fuel cell electrodes., ChemSusChem. 4 (2011) 1753–7. doi:10.1002/cssc.201100245. [95] M. Brodt, T. Han, N. Dale, E. Niangar, R. Wycisk, P. Pintauro, Fabrication, In-Situ Performance, and Durability of Nanofiber Fuel Cell Electrodes, J. Electrochem. Soc. 162 (2014) F84–F91. doi:10.1149/2.0651501jes. [96] R. Padmavathi, D. Sangeetha, Synthesis and characterization of electrospun carbon nanofiber supported Pt catalyst for fuel cells, Electrochim. Acta. 112 (2013) 1–13. doi:10.1016/j.electacta.2013.08.078. [97] M.T. Sung, M.H. Chang, M.H. Ho, Investigation of cathode electrocatalysts composed of electrospun Pt nanowires and Pt/C for proton exchange membrane fuel cells, J. Power Sources. 249 (2014) 320–326. doi:10.1016/j.jpowsour.2013.10.119. [98] M. Wei, M. Jiang, X. Liu, M. Wang, S. Mu, Graphene-doped electrospun nanofiber membrane electrodes and proton exchange membrane fuel cell performance, J. Power Sources. 327 (2016) 384–393. doi:10.1016/j.jpowsour.2016.07.083. [99] Q. Duan, B. Wang, J. Wang, H. Wang, Y. Lu, Fabrication of a carbon nanofiber sheet as a micro-porous layer for proton exchange membrane fuel cells, J. Power Sources. 195 (2010) 8189–8193. doi:10.1016/j.jpowsour.2010.07.032. [100] S. Uhm, B. Jeong, J. Lee, A facile route for preparation of non-noble CNF cathode catalysts in alkaline ethanol fuel cells, Electrochim. Acta. 56 (2011) 9186–9190. doi:10.1016/j.electacta.2011.07.108. [101] W.E. Teo, R. Inai, S. Ramakrishna, Technological advances in electrospinning of nanofibers, Sci. Technol. Adv. Mater. 12 (2011) 013002. doi:10.1088/1468-6996/12/1/013002. [102] J.B. Ballengee, P.N. Pintauro, Composite Fuel Cell Membranes from Dual-Nanofiber Electrospun Mats, 44 (2011) 7307–7314. doi:10.1021/ma201684j. [103] R. Sood, S. Cavaliere, D.J. Jones, J. Rozière, Electrospun nanofibre composite polymer electrolyte fuel cell and electrolysis membranes, Nano Energy. 26 (2016) 729–745. doi:10.1016/j.nanoen.2016.06.027. [104] M.M. Mannarino, D.S. Liu, P.T. Hammond, G.C. Rutledge, Mechanical and Transport Properties of Layer-by-Layer Electrospun Composite Proton Exchange Membranes for Fuel Cell Applications, ACS Appl. Mater. Interfaces. 5 (2013) 8155–8164. doi:10.1021/am402204v. [105] S.W. Choi, Y.Z. Fu, Y.R. Ahn, S.M. Jo, A. Manthiram, Nafion-impregnated electrospun polyvinylidene fluoride composite membranes for direct methanol fuel cells, J. Power Sources. 180 (2008) 167–171. doi:10.1016/j.jpowsour.2008.02.042. [106] C. Feng, K.C. Khulbe, T. Matsuura, Recent progress in the preparation, characterization, and applications of nanofibers and nanofiber membranes via electrospinning/interfacial polymerization, J. Appl. Polym. Sci. 115 (2010) 756–776. doi:10.1002/app.31059. [107] R. Gosalawit, S. Chirachanchai, S. Shishatskiy, S.P. Nunes, Sulfonated 169  montmorillonite/sulfonated poly(ether ether ketone) (SMMT/SPEEK) nanocomposite membrane for direct methanol fuel cells (DMFCs), J. Memb. Sci. 323 (2008) 337–346. doi:DOI 10.1016/j.memsci.2008.06.038. [108] X. Xiong, Q. Li, X.-C. Zhang, L. Wang, Z.-X. Guo, J. Yu, Poly(vinylidene fluoride)/silica nanocomposite membranes by electrospinning, J. Appl. Polym. Sci. 129 (2013) 1089–1095. doi:10.1002/app.38787. [109] H. Junoh, J. Jaafar, M.N.A. Mohd Norddin, A.F. Ismail, M.H.D. Othman, M. a. Rahman, N. Yusof, W.N. Wan Salleh, H. Ilbeygi, M. Noorul, A. Mohd, A.F. Ismail, M. Hafiz, D. Othman, M. a. Rahman, N. Yusof, W. Norhayati, W. Salleh, H. Ilbeygi, A Review on the Fabrication of Electrospun Polymer Electrolyte Membrane for Direct Methanol Fuel Cell, J. Nanomater. 2015 (2014) 1–16. doi:10.1088/0960-1317/16/4/R01. [110] A. Baji, Y.-W. Mai, S.-C. Wong, M. Abtahi, P. Chen, Electrospinning of polymer nanofibers: Effects on oriented morphology, structures and tensile properties, Compos. Sci. Technol. 70 (2010) 703–718. doi:10.1016/j.compscitech.2010.01.010. [111] A. Saraf, G. Lozier, A. Haesslein, F.K. Kasper, R.M. Raphael, L.S. Baggett, A.G. Mikos, Fabrication of Nonwoven Coaxial Fiber Meshes by Electrospinning, Tissue Eng. Part C-Methods. 15 (2009) 333–344. doi:DOI 10.1089/ten.tec.2008.0422. [112] H.-Y.Y. Mi, M.R. Salick, X. Jing, W.C. Crone, X.-F.F. Peng, L.-S.S. Turng, Electrospinning of unidirectionally and orthogonally aligned thermoplastic polyurethane nanofibers: Fiber orientation and cell migration., J. Biomed. Mater. Res. A. 105504 (2014) 1–11. doi:10.1002/jbm.a.35208. [113] W.E. Teo, S. Ramakrishna, A review on electrospinning design and nanofibre assemblies., Nanotechnology. 17 (2006) R89–R106. doi:10.1088/0957-4484/17/14/R01. [114] D.H. Reneker, A.L. Yarin, H. Fong, S. Koombhongse, Bending instability of electrically charged liquid jets of polymer solutions in electrospinning, J. Appl. Phys. 87 (2000) 4531–4547. doi:10.1063/1.373532. [115] S. Rafiei, S. Maghsoodloo, B. Noroozi, V. Mottaghitalab,  a K. Haghi, Mathematical modeling in electrospinning process of nanofibers: a detailed review, Cellul. Chem. Technol. 47 (2013) 323–338. [116] C.J. Thompson, G.G. Chase, A.L. Yarin, D.H. Reneker, Effects of parameters on nanofiber diameter determined from electrospinning model, Polymer (Guildf). 48 (2007) 6913–6922. doi:10.1016/j.polymer.2007.09.017. [117] B.L. Hager, G.C. Berry, Moderately concentrated solutions of polystyrene. I. Viscosity as a function of concentration, temperature, and molecular weight, J. Polym. Sci. Polym. Phys. Ed. 20 (1982) 911–928. doi:10.1002/pol.1982.180200513. [118] S. V. Fridrikh, J.H. Yu, M.P. Brenner, G.C. Rutledge, Controlling the Fiber Diameter during Electrospinning, Phys. Rev. Lett. 90 (2003) 144502. doi:10.1103/PhysRevLett.90.144502. [119] M.S. a. Rahaman,  a. F. Ismail,  a. Mustafa, A review of heat treatment on polyacrylonitrile fiber, Polym. Degrad. Stab. 92 (2007) 1421–1432. doi:10.1016/j.polymdegradstab.2007.03.023. [120] R.C. Houtz, “Orlon” acrylic fiber: chemistry and properties, Text. Res. J. v 20 (n.d.) 786–801. [121] L.I.. David, A.. Ismail, Influence of the thermastabilization process and soak time during pyrolysis process on the polyacrylonitrile carbon membranes for O2/N2 separation, J. 170  Memb. Sci. 213 (2003) 285–291. doi:10.1016/S0376-7388(02)00513-6. [122] D. Zhu, C. Xu, N. Nakura, M. Matsuo, Study of carbon films from PAN/VGCF composites by gelation/crystallization from solution, Carbon N. Y. 40 (2002) 363–373. doi:10.1016/S0008-6223(01)00116-6. [123] Z. Bashir, A critical review of the stabilisation of polyacrylonitrile, Carbon N. Y. 29 (1991) 1081–1090. doi:10.1016/0008-6223(91)90024-D. [124] S. Dalton, F. Heatley, P.M. Budd, Thermal stabilization of polyacrylonitrile fibres, Polymer (Guildf). 40 (1999) 5531–5543. doi:10.1016/S0032-3861(98)00778-2. [125] P.J. Donnet JB, Wang TK, Carbon Fiber, 3rd editio, Marcel Dekker Inc., New York, 1998. [126] N. Grassie, R. McGuchan, Pyrolysis of polyacrylonitrile and related polymers-III. Thermal analysis of preheated polymers, Eur. Polym. J. 7 (1971) 1357–1371. doi:10.1016/0014-3057(71)90032-2. [127] J. Donnet, R. Bansal, Carbon Fibers, 1990. doi:10.1361/asmhba0003354. [128] T.-H. Ko, The influence of pyrolysis on physical properties and microstructure of modified PAN fibers during carbonization, J. Appl. Polym. Sci. 43 (1991) 589–600. doi:10.1002/app.1991.070430321. [129] D.D. Edie, The effect of processing on the structure and properties of carbon fibers, Carbon N. Y. 36 (1998) 345–362. doi:10.1016/S0008-6223(97)00185-1. [130] J.-S. Tsai, C.-H. Lin, The effect of the side chain of acrylate comonomers on the orientation, pore-size distribution, and properties of polyacrylonitrile precursor and resulting carbon fiber, J. Appl. Polym. Sci. 42 (1991) 3039–3044. doi:10.1002/app.1991.070421123. [131] J. Mittal, O.P. Bahl, R.B. Mathur, Single step carbonization and graphitization of highly stabilized PAN fibers, Carbon N. Y. 35 (1997) 1196–1197. doi:10.1016/S0008-6223(97)84653-2. [132] Y. Yamashita, N. Aoki, Carbonization conditions for electrospun nanofibre of polyacylonitrile copolymer, 33 (2008) 345–353. [133] W.J. Ritchie, Electrospun carbon model system for a nanofiber transparent conductor, (2013). http://circle.ubc.ca/handle/2429/43865 (accessed February 26, 2015). [134] W.J. Burlant, J.L. Parsons, Pyrolysis of polyacrylonitrile, J. Polym. Sci. 22 (1956) 249–256. doi:10.1002/pol.1956.1202210107. [135] Z. Zhou, C. Lai, L. Zhang, Y. Qian, H. Hou, D.H. Reneker, H. Fong, Development of carbon nanofibers from aligned electrospun polyacrylonitrile nanofiber bundles and characterization of their microstructural, electrical, and mechanical properties, Polymer (Guildf). 50 (2009) 2999–3006. doi:10.1016/j.polymer.2009.04.058. [136] S.J. Park, G.Y. Heo, Precursors and manufacturing of carbon fibers, Springer Ser. Mater. Sci. (2015). doi:10.1007/978-94-017-9478-7_2. [137] G.J. Kim, P. Jao, Y.-K. Yoon, Fabrication of Nanoporous Membrane, US 2014/0124131 A1, 2014. [138] V. Kamavaram, V. Veedu,  a. M. Kannan, Synthesis and characterization of platinum nanoparticles on in situ grown carbon nanotubes based carbon paper for proton exchange membrane fuel cell cathode, J. Power Sources. 188 (2009) 51–56. doi:10.1016/j.jpowsour.2008.11.084. [139] S. Björklund, V. Kocherbitov, Alcohols react with MCM-41 at room temperature and chemically modify mesoporous silica, Sci. Rep. 7 (2017) 9960. doi:10.1038/s41598-017-171  10090-x. [140] N.G. Sahoo, S. Rana, J.W. Cho, L. Li, S.H. Chan, Polymer nanocomposites based on functionalized carbon nanotubes, Prog. Polym. Sci. 35 (2010) 837–867. doi:10.1016/j.progpolymsci.2010.03.002. [141] S. Sun, F. Jaouen, J.P. Dodelet, Controlled growth of Pt nanowires on carbon nanospheres and their enhanced performance as electrocatalysts in PEM fuel cells, Adv. Mater. 20 (2008) 3900–3904. doi:10.1002/adma.200800491. [142] S.P. Rwei, T.F. Way, W.Y. Chiang, S.Y. Pan, Effect of tacticity on the cyclization of polyacrylonitrile copolymers, Colloid Polym. Sci. 295 (2017) 803–815. doi:10.1007/s00396-017-4062-4. [143] E. Zengeni, P.C. Hartmann, R.D. Sanderson, P.E. Mallon, Poly(acrylonitrile-co-methyl acrylate) copolymers: Correlation between copolymer composition, morphology and positron annihilation lifetime parameters, J. Appl. Polym. Sci. (2011). doi:10.1002/app.32550. [144] S. Lee, J. Kim, B. Ku, J. Kim, H. Joh, Structural Evolution of Polyacrylonitrile Fibers in Stabilization and Carbonization, 2012 (2012) 275–282. [145] A. Gupta, I.R. Harriso, New aspects in the oxidative stabilization of pan-based carbon fibers, Carbon N. Y. 34 (1996) 1427–1445. doi:10.1016/S0008-6223(96)00094-2. [146] Z. Song, X. Hou, L. Zhang, S. Wu, Enhancing Crystallinity and Orientation by Hot-Stretching to Improve the Mechanical Properties of Electrospun Partially Aligned Polyacrylonitrile (PAN) Nanocomposites, Materials (Basel). 4 (2011) 621–632. doi:10.3390/ma4040621. [147] M. Panapoy, A. Dankeaw, B. Ksapabutr, Electrical Conductivity of PAN-based Carbon Nanofibers Prepared by Electrospinning Method, 13 (2008) 11–17. [148] L. Vaisman, H.D. Wagner, G. Marom, The role of surfactants in dispersion of carbon nanotubes., Adv. Colloid Interface Sci. 128–130 (2006) 37–46. doi:10.1016/j.cis.2006.11.007. [149] W. Weibull, The Phenomenon of Rupture in Solids, Proc. R. Swedish Inst. Eng. Res. 153 (1939) 1–55. [150] E. Ridge, D. Kudenko, E. Ridge, D. Kudenko, Sequential Experiment Designs for Screening and Tuning Parameters of Stochastic Heuristics, Work. Empir. METHODS Anal. ALGORITHMS, REYKJAVIK, Icel. (2006) 27–34. http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.64.3468 (accessed April 27, 2018). [151] M. Ziabari, V. Mottaghitalab, A.K. Haghi, Evaluation of electrospun nanofiber pore structure parameters, Korean J. Chem. Eng. 25 (2008) 923–932. https://link-springer-com.ezproxy.library.ubc.ca/content/pdf/10.1007%2Fs11814-008-0151-x.pdf (accessed June 13, 2018). [152] R. H.Myers, D. C.Montgomery, C. Anderson-Cook, M, Response Surface Methodology: Process and Product Optimization Using Designed Experiments, 4th ed., John Wiley & Sons, Hoboken, 2016. doi:10.1017/CBO9781107415324.004. [153] L. Huang, N.-N. Bui, S.S. Manickam, J.R. Mccutcheon, Controlling Electrospun Nanofiber Morphology and Mechanical Properties Using Humidity, J Polym Sci Part B Polym Phys. 49 (2011) 1734–1744. doi:10.1002/polb.22371. [154] J. Rafique, J. Yu, X. Zha, K. Rafique, Fabrication of ultra thin and aligned carbon 172  nanofibres from electrospun polyacrylonitrile nanofibres, Bull. Mater. Sci. 33 (2010) 553–559. https://www.ias.ac.in/article/fulltext/boms/033/05/0553-0559 (accessed June 1, 2018). [155] T. Suzuki, H. Tanaka, M. Hayase, S. Tsushima, S. Hirai, Investigation of porous structure formation of catalyst layers for proton exchange membrane fuel cells and their effect on cell performance, Int. J. Hydrogen Energy. 41 (2016) 20326–20335. doi:10.1016/j.ijhydene.2016.09.078. [156] Z. Zhang, L. Jia, Parametric study of the porous cathode in the PEM fuel cell, Int. J. Energy Res. 33 (2009) 52–61. doi:10.1002/er.1471. [157] M. El Hannach, M. Prat, J. Pauchet, Pore network model of the cathode catalyst layer of proton exchange membrane fuel cells: Analysis of water management and electrical performance Pore network model of the cathode catalyst layer of proton exchange membrane fuel cells: Analysis of water management and electrical per- formance, Int. J. Hydrogen Energy. 37 (2012) 18996–19006. doi:10.1016/j.ijhydene. [158] L.S. Carnell, E.J. Siochi, N.M. Holloway, R.M. Stephens, C. Rhim, L.E. Niklason, R.L. Clark, Aligned Mats from Electrospun Single Fibers, (2008) 5345–5349. [159] G.R. Rakesh, G.S. Ranjit, K.K. Karthikeyan, P. Radhakrishnan, P. Biji, A facile route for controlled alignment of carbon nanotube-reinforced, electrospun nanofibers using slotted collector plates, Express Polym. Lett. 9 (2015) 105–118. doi:10.3144/expresspolymlett.2015.12. [160] D. Li, Y. Wang, Y. Xia, V. Uni, Electrospinning of Polymeric and Ceramic Nanofibers as Uniaxially Aligned Arrays, (2003). [161] R. Blume, D. Rosenthal, J.-P. Tessonnier, H. Li, A. Knop-Gericke, R. Schlögl, Characterizing Graphitic Carbon with X-ray Photoelectron Spectroscopy: A Step-by-Step Approach, ChemCatChem. 7 (2015) 2871–2881. doi:10.1002/cctc.201500344. [162] K. Kim, H.-B.-R. Lee, R.W. Johnson, J.T. Tanskanen, N. Liu, M.-G. Kim, C. Pang, C. Ahn, S.F. Bent, Z. Bao, Selective metal deposition at graphene line defects by atomic layer deposition, Nat. Commun. 5 (2014) 4781. doi:10.1038/ncomms5781. [163] B. Barua, M.C. Saha, Influence of Humidity , Temperature , and Annealing on Microstructure and Tensile Properties of Electrospun Polyacrylonitrile Nanofibers, (2017) 1–12. doi:10.1002/pen. [164] D. Kołbuk, P. Sajkiewicz, T.A. Kowalewski, Optical birefringence and molecular orientation of electrospun polycaprolactone fibers by polarizing-interference microscopy, Eur. Polym. J. 48 (2012) 275–283. doi:10.1016/j.eurpolymj.2011.11.012. [165] Y. Liu, L. Cui, F. Guan, Y. Gao, N.E. Hedin, L. Zhu, H. Fong, Crystalline Morphology and Polymorphic Phase Transitions in Electrospun Nylon 6 Nanofibers., Macromolecules. 40 (2007) 6283–6290. doi:10.1021/ma070039p. [166] M. Richard-lacroix, C. Pellerin, Molecular Orientation in Electrospun Fibers: From Mats to Single Fibers, (2013). [167] L. Laffont, M. Monthioux, V. Serin, R.B. Mathur, C. Guimon, M.F. Guimon, An EELS study of the structural and chemical transformation of PAN polymer to solid carbon, Carbon N. Y. 42 (2004) 2485–2494. doi:10.1016/j.carbon.2004.04.043. [168] M.H. Eikerling, K. Malek, Q. Wang, Catalyst layer modeling: Structure, properties and performance, in: J. Zhang (Ed.), PEM Fuel Cell Electrocatal. Catal. Layers Fundam. Appl., 1st ed., Springer-Verlag London, 2008: pp. 381–446. doi:10.1007/978-1-84800-173  936-3_8. [169] J.A. Michel, W.H. Morris III, C.M. Lukehart, W.H.M. Iii, C.M. Lukehart, Synthesis of shaped Pt nanoparticles using common anions or small molecules as shape-directing agents: observation of a strong halide or pseudo-halide effect, J. Mater. Chem. A. 3 (2015) 2012–2018. doi:10.1039/C4TA05952B. [170] S. Du, K. Lin, S.K. Malladi, Y. Lu, S. Sun, Q. Xu, R. Steinberger-Wilckens, H. Dong, Plasma nitriding induced growth of Pt-nanowire arrays as high performance electrocatalysts for fuel cells, Sci. Rep. 4 (2014) 6439. doi:10.1038/srep06439. [171] J.R.C. Salgado, R.G. Duarte, L.M. Ilharco, A.M. Botelho do Rego, A.M. Ferraria, M.G.S. Ferreira, Effect of functionalized carbon as Pt electrocatalyst support on the methanol oxidation reaction, Appl. Catal. B Environ. 102 (2011) 496–504. doi:10.1016/j.apcatb.2010.12.031. [172] G. Zhang, S. Sun, D. Yang, J.P. Dodelet, E. Sacher, The surface analytical characterization of carbon fibers functionalized by H2SO4/HNO3 treatment, Carbon N. Y. 46 (2008) 196–205. doi:10.1016/j.carbon.2007.11.002. [173] W. Watt, Carbon work at the royal aircraft establishment, Carbon N. Y. 10 (1972). doi:10.1016/0008-6223(72)90036-X. [174]  a. Ferrari, J. Robertson, Interpretation of Raman spectra of disordered and amorphous carbon, Phys. Rev. B. 61 (2000) 14095–14107. doi:10.1103/PhysRevB.61.14095. [175] C. Zhang, J. Hu, X. Zhang, X. Wang, Y. Meng, Certain nitrogen functionalities on carbon nanofiber support for improving platinum performance, Catal. Today. 256 (2015) 193–202. doi:10.1016/j.cattod.2015.01.026. [176] Y. Lu, S. Du, R. Steinberger-Wilckens, Temperature-controlled growth of single-crystal Pt nanowire arrays for high performance catalyst electrodes in polymer electrolyte fuel cells, Appl. Catal. B Environ. 164 (2015) 389–395. doi:10.1016/J.APCATB.2014.09.040. [177] J.C. Chaston, Reactions of Oxygen with the Platinum Metals; I-The oxidation of platinum, Platin. Met. Rev. 8 (1964) 50–51. http://www.ingentaconnect.com/content/matthey/pmr/1965/00000009/00000002/art00007 (accessed April 11, 2018). [178] A.M. Gómez-Marín, R. Rizo, J.M. Feliu, Some reflections on the understanding of the oxygen reduction reaction at Pt(111), Beilstein J. Nanotechnol. 4 (2013) 956–967. doi:10.3762/bjnano.4.108. [179] P.P. Lopes, D. Strmcnik, D. Tripkovic, J.G. Connell, V. Stamenkovic, N.M. Markovic, Relationships between Atomic Level Surface Structure and Stability/ Activity of Platinum Surface Atoms in Aqueous Environments, (2016). doi:10.1021/acscatal.5b02920. [180] F. Calle-Vallejo, J. Tymoczko, V. Colic, Q.H. Vu, M.D. Pohl, K. Morgenstern, D. Loffreda, P. Sautet, W. Schuhmann, A.S. Bandarenka, Finding optimal surface sites on heterogeneous catalysts by counting nearest neighbors, Science (80-. ). 350 (2015) 185–189. doi:10.1126/science.aab3501. [181] F. Calle-Vallejo, M.D. Pohl, D. Reinisch, D. Loffreda, P. Sautet, A.S. Bandarenka, Why conclusions from platinum model surfaces do not necessarily lead to enhanced nanoparticle catalysts for the oxygen reduction reaction, Chem. Sci. 8 (2017) 2283–2289. doi:10.1039/C6SC04788B. [182] M. Nesselberger, S. Ashton, J.C. Meier, I. Katsounaros, K.J.J. Mayrhofer, M. Arenz, The Particle Size Effect on the Oxygen Reduction Reaction Activity of Pt Catalysts: Influence 174  of Electrolyte and Relation to Single Crystal Models - ESI, J. Am. Chem. Soc. 133 (2011) 17428–17433. doi:Doi 10.1021/Ja207016u. [183] S.J.C. Cleghorn, X. Ren, T.E. Springer, M.S. Wilson, C. Zawodzinski, T.A. Zawodzinski, S. Gottesfeld, Pem fuel cells for transportation and stationary power generation applications, Int. J. Hydrogen Energy. 22 (1997) 1137–1144. doi:10.1016/S0360-3199(97)00016-5. [184] E. Herrero, K. Franaszczuk, A. Wieckowski, Electrochemistry of methanol at low index crystal planes of platinum. An integrated voltammetric and chronoamperometric study, J. Phys. Chem. 98 (1994) 5074–5083. doi:10.1021/j100070a022. [185] G.Q. Lu, A. Wieckowski, Heterogeneous electrocatalysis: A core field of interfacial science, Curr. Opin. Colloid Interface Sci. 5 (2000) 95–100. doi:10.1016/S1359-0294(00)00041-8. [186] T.J. Schmidt, H.A. Gasteiger, G.D. Stäb, P.M. Urban, D.M. Kolb, R.J. Behm, Characterization of High-Surface-Area Electrocatalysts Using a Rotating Disk Electrode Configuration, J. Electrochem. Soc. 145 (1998) 2354–2358. doi:10.1149/1.1838642. [187] F.C. Nart, W. Vielstich, Normalization of porous active surfaces, (2010) 1–14. [188] M. Shao, J.H. Odell, S.-I. Choi, Y. Xia, Electrochemical surface area measurements of platinum- and palladium-based nanoparticles, Electrochem. Commun. 31 (2013) 46–48. doi:10.1016/j.elecom.2013.03.011. [189] K.J.J. Mayrhofer, D. Strmcnik, B.B. Blizanac, V. Stamenkovic, M. Arenz, N.M. Markovic, Measurement of oxygen reduction activities via the rotating disc electrode method: From Pt model surfaces to carbon-supported high surface area catalysts, Electrochim. Acta. 53 (2008) 3181–3188. doi:10.1016/j.electacta.2007.11.057. [190] S.-Y. Fu, X.-Q. Feng, B. Lauke, Y.-W. Mai, Effects of particle size, particle/matrix interface adhesion and particle loading on mechanical properties of particulate–polymer composites, Compos. Part B Eng. 39 (2008) 933–961. doi:10.1016/j.compositesb.2008.01.002. [191] M. Schoemaker, U. Misz, P. Beckhaus, A. Heinzel, Evaluation of Hydrogen Crossover through Fuel Cell Membranes ~, 2013 (2014) 412–415. doi:10.1002/fuce.201300215. [192] M. Inaba, T. Kinumoto, M. Kiriake, R. Umebayashi, A. Tasaka, Z. Ogumi, Gas crossover and membrane degradation in polymer electrolyte fuel cells, Electrochim. Acta. 51 (2006) 5746–5753. doi:10.1016/j.electacta.2006.03.008. [193] A. Jung, J. Oh, K. Han, M.S. Kim, An experimental study on the hydrogen crossover in polymer electrolyte membrane fuel cells for various current densities, Appl. Energy. 175 (2016) 212–217. doi:10.1016/j.apenergy.2016.05.016. [194] A. Orfanidi, P. Madkikar, H.A. El-Sayed, G.S. Harzer, T. Kratky, H.A. Gasteiger, The Key to High Performance Low Pt Loaded Electrodes, J. Electrochem. Soc. 164 (2017) F418–F426. doi:10.1149/2.1621704jes. [195] S. Shukla, D. Stanier, M.S. Saha, B. Zahiri, M. Tam, J. Stumper, M. Secanell, Characterization of inkjet printed electrodes with improved porosity, in: ECS Trans., 2017. doi:10.1149/07711.1453ecst. [196] F. Yang, R. Pitchumani, Transport and Electrochemical Phenomena, in: N.M. Sammes (Ed.), Fuel Cell Technol. Reach. Towar. Commer., 2006. doi:10.1007/1-84628-207-1_4. [197] F. Barbir, Fuel cell basic chemistry, electrochemistry and thermodynamics frano barbir, in: L.. kakac, Sadik; Pramuanjaroenkij, A.; Vasiliev (Ed.), Mini-Mirco Fuel Cells, 1st ed., 175  Springer Netherlands, 2008: pp. 13–26. doi:10.1007/978-1-4020-8295-5. [198] M. Ahadi, M. Tam, A. Putz, J. Stumper, C. Mccague, M. Bahrami, Electrical conductivity of PEM fuel cell catalyst layers : Through- plane vs . in-plane ( 124 ), 1 (2018) 206–207. [199] M. Ahadi, M. Tam, A. Putz, J. Stumper, C. Mccague, M. Bahrami, Challenges facing thermal and electrical conductivity measurements in catalyst layers of PEM fuel cells ( 125 ), (2018) 208–209. [200] D.R.P. Morris, S.P. Liu, D. Villegas Gonzalez, J.T. Gostick, Effect of Water Sorption on the Electronic Conductivity of Porous Polymer Electrolyte Membrane Fuel Cell Catalyst Layers, ACS Appl. Mater. Interfaces. 6 (2014) 18609–18618. doi:10.1021/am503509j. [201] S.G. Kandlikar, Z. Lu, Thermal management issues in a PEMFC stack - A brief review of current status, Appl. Therm. Eng. 29 (2009) 1276–1280. doi:10.1016/j.applthermaleng.2008.05.009. [202] N. Nonoyama, S. Okazaki, A.Z. Weber, Y. Ikogi, T. Yoshida, Analysis of Oxygen-Transport Diffusion Resistance in Proton-Exchange-Membrane Fuel Cells, J. Electrochem. Soc. 158 (2011) B416. doi:10.1149/1.3546038. [203] E. You, M. Min, S.-A. Jin, T. Kim, C. Pak, Highly Durable, Cost-Effective, and Multifunctional Carbon-Supported IrRu-Based Catalyst for Automotive Polymer Electrolyte Fuel Cell Anodes, J. Electrochem. Soc. 165 (2018) F3094–F3099. doi:10.1149/2.0121806jes. [204] D. Banham, S. Ye, Current Status and Future Development of Catalyst Materials and Catalyst Layers for Proton Exchange Membrane Fuel Cells: An Industrial Perspective, (2017). doi:10.1021/acsenergylett.6b00644. [205] M.K. Kayarkatte, Ö. Delikaya, C. Roth, Freestanding Catalyst Layers: A Novel Electrode Fabrication Technique for PEM Fuel Cells via Electrospinning, ChemElectroChem. 4 (2017) 404–411. doi:10.1002/celc.201600530. [206] D.R. Baker, D.A. Caulk, K.C. Neyerlin, M.W. Murphy, Measurement of Oxygen Transport Resistance in PEM Fuel Cells by Limiting Current Methods, J. Electrochem. Soc. 156 (2009) B991. doi:10.1149/1.3152226. [207] X.-Z. Yuan, C. Song, H. Wang, J. Zhang, Electrochemical Impedance Spectroscopy in PEM Fuel Cells, Electrochem. Impedance Spectrosc. PEM Fuel Cells. (2010) 39–93. doi:10.1007/978-1-84882-846-9. [208] J. Hou, W. Song, H. Yu, Y. Fu, L. Hao, Z. Shao, B. Yi, Ionic resistance of the catalyst layer after the PEM fuel cell suffered freeze, J. Power Sources. 176 (2008) 118–121. doi:10.1016/j.jpowsour.2007.10.035. [209] Y. Yu, H. Li, H. Wang, X.-Z. Yuan, G. Wang, M. Pan, A review on performance degradation of proton exchange membrane fuel cells during startup and shutdown processes: Causes, consequences, and mitigation strategies, J. Power Sources. 205 (2012) 10–23. doi:10.1016/j.jpowsour.2012.01.059. [210] Y. Shao-Horn, W.C. Sheng, S. Chen, P.J. Ferreira, E.F. Holby, D. Morgan, Instability of supported platinum nanoparticles in low-temperature fuel cells, Top. Catal. 46 (2007) 285–305. doi:10.1007/s11244-007-9000-0. [211] J. Xie, D.L. Wood, K.L. More, P. Atanassov, R.L. Borup, Microstructural Changes of Membrane Electrode Assemblies during PEFC Durability Testing at High Humidity Conditions, J. Electrochem. Soc. 152 (2005) A1011. doi:10.1149/1.1873492. [212] Z. Tang, C.K. Poh, K.K. Lee, Z. Tian, D.H.C. Chua, J. Lin, Enhanced catalytic properties 176  from platinum nanodots covered carbon nanotubes for proton-exchange membrane fuel cells, J. Power Sources. 195 (2010) 155–159. doi:10.1016/j.jpowsour.2009.06.105. [213] R.W. Field, B.L. Withers, Occupational and environmental causes of lung cancer., Clin. Chest Med. 33 (2012) 681–703. doi:10.1016/j.ccm.2012.07.001. [214] D.N. Publication, No, NIOSH POCKET GUIDE TO CHEMICAL HAZARDS, (2007) 1–888. https://www.cdc.gov/niosh/docs/2005-149/pdfs/2005-149.pdf (accessed April 2, 2018). [215] R.D. Zumwalde, NIOSH Nanotechnology Research Center, Occupational exposure to carbon nanotubes and nanofibers, Curr. Intell. Bull. (2013) xxiv, 156 pages. https://www.cdc.gov/niosh/docs/2013-145/pdfs/2013-145.pdf (accessed April 2, 2018). [216] D. Yang, B. Lu, Y. Zhao, X. Jiang, Fabrication of aligned fibrous arrays by magnetic electrospinning, Adv. Mater. 19 (2007) 3702–3706. doi:10.1002/adma.200700171. [217] L. Carnell, E. Siochi, R. Wincheski, N. Holloway, R. Clark, Electric field effects on fiber alignment using an auxiliary electrode during electrospinning, Scr. Mater. 60 (2009) 359–361. doi:10.1016/j.scriptamat.2008.09.035. [218] T. Suzuki, H. Murata, T. Hatanaka, Y. Morimoto, Analysis of the Catalyst Layer of Polymer Electrolyte Fuel Cells, R&D Rev. Toyota CRDL. 39 (n.d.). https://www.tytlabs.com/japanese/review/rev393pdf/393_033suzuki.pdf (accessed March 29, 2018). [219] J. Miyake, R. Taki, T. Mochizuki, R. Shimizu, R. Akiyama, M. Uchida, K. Miyatake, Design of flexible polyphenylene proton-conducting membrane for next-generation fuel cells, Sci. Adv. 3 (2017) eaao0476. doi:10.1126/sciadv.aao0476. [220] B. Muriithi, D.A. Loy, Proton Conductivity of Nafion/Ex-Situ Sulfonic Acid-Modified Stöber Silica Nanocomposite Membranes As a Function of Temperature, Silica Particles Size and Surface Modification., Membranes (Basel). 6 (2016). doi:10.3390/membranes6010012. [221] S. Zhang, X. Yuan, H. Wang, W. Mérida, H. Zhu, J. Shen, S. Wu, J. Zhang, A review of accelerated stress tests of MEA durability in PEM fuel cells, Int. J. Hydrogen Energy. 34 (2009) 388–404. doi:10.1016/j.ijhydene.2008.10.012. [222] S. Zhang, X.-Z. Yuan, J.N.C. Hin, H. Wang, K.A. Friedrich, M. Schulze, A review of platinum-based catalyst layer degradation in proton exchange membrane fuel cells, J. Power Sources. 194 (2009) 588–600. doi:10.1016/j.jpowsour.2009.06.073. [223] D. He, C. Zeng, C. Xu, N. Cheng, H. Li, S. Mu, M. Pan, Polyaniline-functionalized carbon nanotube supported platinum catalysts., Langmuir. 27 (2011) 5582–8. doi:10.1021/la2003589. [224] W. Zhang, Electrospinning Pt/C Catalysts into a Nanofiber Fuel Cell Cathode, Electrochem. Soc. (2010) 51. [225] E. Barsoukov, J.R. Macdonald, Impedance spectroscopy: theory, experiment, and applications, 2nd ed., Wiley-Interscience, 2005. https://www.wiley.com/en-ca/Impedance+Spectroscopy%3A+Theory%2C+Experiment%2C+and+Applications%2C+2nd+Edition-p-9780471647492 (accessed June 7, 2018). [226] Gamry Instruments, Basics of EIS: Electrochemical Research-Impedance, (n.d.). https://www.gamry.com/application-notes/EIS/basics-of-electrochemical-impedance-spectroscopy/ (accessed June 10, 2018). [227] J.F. Rubinson, Y.P. Kayinamura, Charge transport in conducting polymers: insights from 177  impedance spectroscopy, Chem. Soc. Rev. 38 (2009) 3339. doi:10.1039/b904083h.    178  A Appendix  Appendix A provides a supplementary literature review summary, some which were not mentioned in the discussion.  Table A-1: Summary of prior research related to CNF/CNT as catalyst supports in FC Material Application Catalyst Deposition Parameters (catalyst loading, porosity, thickness, etc.) Performance Advantages Challenges Ref. Twisted CNFs Study effect of twist on electrical conductivity N/A Diameter of single fiber: 280 nm Yarn diameter: 25 mm  Electrical Conductivity increase with increasing porosity (~1 S/m at 90% porosity) Higher porosity with lower twist (90% porosity at 300 tpm) Increase in conductivity due to permanent junctions between fibers, forming alternate paths Fiber breakage during twisting, decreases conductivity [85] Stacked-cup CNFs Catalyst layer, performance compared to CB Modified ethylene Glycol synthesis method 2-3 nm Pt particles SC-CNF diameter: 75-100 nm  Higher performance than CB, different Pt loadings were tested Can maintain continuous electrical conduction path, not easily cut down by Nafion micelles Agglomeration observed for 30wt% Pt/SC-CNFs  [63] CNT on carbon paper In situ CNT on carbon paper, combined GDL and catalyst layer (Cathode) Sputter coat 2-3 nm Pt Nanodots Loading: 0.04mg/cm (cathode) Pore size: <100 nm MWCNT OD: 30nm ID: 10nm Max power Density: 595mW/cm2 Significantly higher power density compared to Pt/Vulcan XC72R with equal Pt loading Requires acid treatment, limiting the effectiveness of CNT as catalyst support Optimization needed to control particles [212] 179  size and dispersion Electrospun Random CNF Catalyst layer improve morphology Pt slurry spray method Diameter: 250 nm Pore size: 2.36 nm Loading: 0.5 mg/cm2 (anode and cathode) Electrical conductivity: 9.9 S/cm Current density: 900mA/cm2 compared to 450mA/cm2 for Pt/XC-72R Pt/XC-72R: 4.5 S/cm Pt utilization 69% compared to 35% on Pt/XC-72R Difficult for Pt particles to penetrate the well-developed shallow pores [35] CNF mats using PAN Catalyst support layer, study Pt loading and performance Pt clusters electrodeposited Pt size: (50 – 200 nm) Fiber diameter: 150 nm Mat thickness: ~50 µm Loading: 0.21mg/cm2 Specific surface area: ~7 m2/g Peak current density: 117.5 mA/cm2 electrical conductivity of mats: ~50 S/cm  90mA/cm2 compared to 39mA/cm2 for Pt/CFM and commercial Pt/C electrodes Use less Pt loading Pt/C electrode electrical conductivity only ~42 S/cm Performance of Pt/CFM superior to Pt/C in both stability and activity, although dispersivity and size of Pt particles on CFM are inferior to Pt/C [83] CNT@SnO2 catalyst support, invested core-sheath nanocomposite structure Precursor H2PtCl6 and EG as solvent SnO2 sheath pore size: 2-5 nm SnO2 particle size: 4.7 nm MWCNT D: 40 – 60 nm MWCNT L: 5 – 15 µm   Current densities for CNT@SnO2 higher than that for Pt/CNT Long term stability of Pt/(CNT@SnO2) significantly higher than that of Pt/CNT SnO2 is sufficiently steady at high potentials, high corrosion resistance Mesopores formed by SnO2 mitigate the Pt migration or aggregation N/A [62] 180  Platelet, tubular, and fishbone CNF  Conducted voltammetric analysis, effect of microstructure on ORR activity Catalyst powder suspension, dispersed by ultrasonic in Nafion solution Pd NP size: 3 – 8 nm f-CNF: 60 nm Diameter, 0.3cm3/g pore volume, 86.6m2/g specific surface area p-CNF: 100 nm Diameter, 0.35cm3/g pore volume, 204.7 m2/g specific surface area No ORR peak found on Pd/AC,ORR of Pd/f-CNF is diffusion controlled and ORR of Pd/p-CNF controlled jointly by surface reaction and diffusion High active surface area of Pd/p-CNF (0.9m2/g), much higher than Pd/AC and Pd/f-CNF (0.37m2/g; 0.21m2/g) high active surface area reduce over-potential of electrochemical reaction N/A [73] Pt-PANI/CNT PANI introduced to bridge Pt NP and CNTs, catalyst support layer Pt colloidal NP, narrow size distribution Pt loading: 20% Pt NP size: 2 -4 nm ECA: Pt/PANI/CNT 64.5 m2/g   Pt-PANI/CNT has 3 times longer durability than Pt/C and 1.5 times longer than Pt/CNT, 40% as benchmark under same ADT conditions ECA increase 28% compared to Pt/C Pt/C durability is much lower than Pt/CNT PANI-Pt/CNT experiences 4500 cycles before ECA drops to 40% Pt NP tightly anchored onto surface of CNT by covalent bonding with N atoms in PANI N/A [223] Pt/CNT Catalyst support layer, evaluate effect of different oxidation treatments on CNT Electroless plating of Pt Synthesized CNT Functionalized and sensitized C/Pt: 1.34E-4 mA/cm2 Pt/CNT electrocatalyst contained 67% Pt(0) and 32.7% Pt(IV) Showed high electrocatalytic N/A [61] 181  activity in PEM stack Hybrid MWCNT/SWCNT Catalyst support layer, optimized SWNT-MWNT hybrid structure and high mass activity at high power output Reduction of Pt salts with EG SWCNT thickness: 0.3 – 1.4 µm Pt Loading: 20 – 100 µg/cm2 optimum blend: 20-30 µgPt/cm2 on SWNT and 5 – 10 µgPt/cm2 on MWNT Peak power density: ~700 mW/cm2   Porosity of MWNT and proton conductivity of Pt-SWNT achieve very active catalyst layer N/A [41] Vertically aligned nitrogen doped carbon nano tubes Investigate ORR rate/activity, for catalyst support layer Metal-free catalyst system, use only CNT NCNT length: 7.5 µm NCNT width: ~50 nm ECSA: 0.29 cm2   Current density: 2.62E-3 A/cm2  ORR rate constant: 5.38E-3 cm/s ORR rate compares favorably with those obtained on Pt-based 62% higher performance than benchmark accepted for ORR catalysts N/A [64] Electrospun Pt/C Fabricate electrospun Pt/C onto carbon paper Pt in spinning dope Nanofiber D: ~800 nm Pt Loading: either 0.4 mg/cm2 or 0.2 mg/cm2 Max power density: 705 mW/cm2 Current density: 1080 mA/cm2 28% improvement in FC performance compared to MEA with decal cathode/anode N/A [224]  182  B Appendix  Appendix B provides further discussion and theory about irreversible losses in the fuel cell.  Polarization losses  Polarization loss refers to the decrease in electrode potential from ideal equilibrium values due to the irreversible losses in the electrochemical system. The losses are represented in performance (polarization) curves which display the electrical potential as a function of current density. As mentioned in Chapter 1.2.2, the four irreversible losses which are briefly described here are: OCV loss kinetic loss, ohmic loss, and concentration loss.   OCV loss  The OCV represents the cell voltage when zero current is discharged. In practice, the OCV is always lower than the theoretical equilibrium cell voltage because of inherent fuel crossover. Fuel crossover describes the H2 that crosses through the proton conducting membrane to the cathode side to react with O2 directly. Other factors related to OCV loss include short circuits and impurities.  Kinetic loss  Kinetic loss refers to the energy incurred to overcome the activation barrier to drive the reaction forward. This loss occurs in the low current density region of the polarization curve and is estimated by the Butler-Volmer formula Equation (B.1); where i is the current density, i0 is the exchange current density, α is the kinetic transfer coefficient and ηkin is the activation overpotential. The two factors that determine the sluggishness of the reaction are the transfer coefficient and exchange current density. The transfer coefficient refers to the experimental coefficients for the forward and backward reactions at the electrode (reduction and oxidation). It is the proportion of electrical energy applied that is harnessed in changing the rate of an 183  electrochemical reaction. Exchange current density refers to the back and forth flow of electrons from and to the electrolyte. If the i0 is high, the more active the surface of the electrode is and more likely the current is to flow in one direction.   (B.1) ∆𝑉𝑜ℎ𝑚 = 𝑖𝑅𝑖 (B.2)  (B.3) ∆𝑉𝑡𝑟𝑎𝑛𝑠 = −𝑅𝑇𝑛𝐹ln (1 −𝑖𝑖𝑙𝑖𝑚) (B.4) Ohmic loss  The linear portion of the polarization curve is dominated by ohmic loss. Ohmic loss refers to the cell resistance and includes bulk resistances of the components, interfacial resistances, protonic and electronic resistances. These losses can be expressed by Ohm’s law (B.2); where Ri is the total cell resistance. Total Ri is shown in Equation (B.3); where Ri,i is the ionic resistance, Ri,e is the electronic resistance, and Ri,c is the contact resistance.   Mass transport loss   Mass transport (concentration) polarization refers to the concentration gradient caused by rapid consumption of the fuel at the electrode. The reactant concentration decreases with increasing current density. Thus, the surface concentration reaches zero when the rate of consumption exceeds the diffusion rate. The current at this stage is called the (ilim) limiting current density (B.4). A fuel cell cannot produce more than the limiting current because there are no reactants at the catalyst surface. Mass transport loss is more prominent in the cathode side because if water is not removed efficiently, excess water accumulates in the pores causing flooding which hinders the diffusion of oxygen to the reaction sites.    184  C Appendix  Appendix C provides raw data for the factorial design, regression modeling statistics, and theory of response surface method.  C.1 Raw data for factorial design experiments  Table C-1: Summary of 4 factor 2 level factorial design   Factor 1 Factor 2 Factor 3 Factor 4 Std Run A:Concentration B:Stabilization Temp C:CNT loading D:Carbonization Temperature   wt% °C wt% °C 1 5 7 230 0 850 2 2 10 230 0 850 3 4 7 260 0 850 4 8 10 260 0 850 5 6 7 230 1 850 6 11 10 230 1 850 7 1 7 260 1 850 8 12 10 260 1 850 9 9 7 230 0 1100 10 14 10 230 0 1100 11 15 7 260 0 1100 12 13 10 260 0 1100 13 7 7 230 1 1100 14 3 10 230 1 1100 15 10 7 260 1 1100 16 16 10 260 1 1100  185   Figure C-1: Histogram of average fiber diameter of Sample #10 (nm)   Figure C-2: Fiber mat factorial design average fiber diameter (nm)   Figure C-3: Fiber mat factorial design in-plane electrical conductivity (S cm-1)   Figure C-4: Fiber mat factorial design % porosity   Figure C-5: Fiber mat factorial design results for elastic modulus (GPa)   Figure C-6: Fiber mat factorial design through-plane electrical resistance (mΩ∙cm2)  186  C.2 Regression modelling for process optimization     Table C-2: ANOVA analysis for diameter, porosity, electrical conductivity, and young’s modulus regression models Source Sum of Squares df Mean Square F Value p-value Prob > F Model (Diameter) 41724.50 9 4636.06 8.50 0.0085   A-Polymer Concentration 5184.00 1 5184.00 9.50 0.0216   B-Stab Temperature 144.00 1 144.00 0.26 0.6258   C-Carb Temperature 4489.00 1 4489.00 8.23 0.0285   D-CNT Concentration 1640.25 1 1640.25 3.01 0.1336   AB 144.00 1 144.00 0.26 0.6258   AC 169.00 1 169.00 0.31 0.5980   AD 12210.25 1 12210.25 22.38 0.0032   BC 10000.00 1 10000.00 18.33 0.0052   ABC 7744.00 1 7744.00 14.19 0.0093 Residual 3273.50 6 545.58   Corrected Total 44998.00 15          Model (Porosity) 359.25 4 89.81 8.39 0.0023   B-Stab Temperature 1.56 1 1.56 0.15 0.7096   C-Carb Temperature 0.56 1 0.56 0.05 0.8228   D-CNT Concentration 280.56 1 280.56 26.22 0.0003   BC 76.56 1 76.56 7.16 0.0216 Residual 117.69 11 10.70   Corrected Total 476.94 15          Model (Conductivity)  16.97 2 8.48 81.28 < 0.0001   C-Carb Temperature 14.77 1 14.77 141.50 < 0.0001   D-CNT Concentration 2.20 1 2.20 21.07 0.0005 Residual 1.36 13 0.10   Corrected Total 18.33 15          Model (Modulus) 205.10 9 22.79 5.27 0.0280   A-Polymer Concentration 35.62 1 35.62 8.24 0.0284   B-Stab Temperature 18.59 1 18.59 4.30 0.0834   C-Carb Temperature 1.08 1 1.08 0.25 0.6357   D-CNT Concentration 14.49 1 14.49 3.35 0.1168   AB 50.89 1 50.89 11.77 0.0140   BC 20.17 1 20.17 4.67 0.0741   BD 6.33 1 6.33 1.46 0.2717   CD 2.63 1 2.63 0.61 0.4652   BCD 55.30 1 55.30 12.79 0.0117 Residual 25.94 6 4.32   Corrected Total 231.04 15    187  Table C-3: Response surface function predicted optimized solutions ranked by desirability  Number Polymer Concentration Stab Temp Carb Temp CNT concentration Diameter Conductivity Porosity Modulus Desirability 1 10.00 230.00 1100.00 0.00 149.25 37.96 78.81 6.79 0.88 2 9.64 230.00 1099.99 0.00 158.26 37.9545 78.8122 7.53704 0.882754 3 9.90 230.01 1099.16 0.00 152.127 37.7131 78.7952 6.95798 0.882541 4 9.67 230.42 1100.00 0.00 158.967 37.9557 78.7598 7.43480 0.882095 5 9.82 230.00 1089.96 0.00 158.998 35.1341 78.6223 6.90035 0.880323 6 10.00 232.06 1099.93 0.00 156.566 37.9345 78.5538 6.61754 0.879461 7 10.00 233.15 1100.00 0.00 160.391 37.9560 78.4184 6.53089 0.874506 8 10.00 230.00 1100.00 0.08 155.272 35.7879 79.4765 6.47390 0.866599 9 9.30 230.00 1099.85 0.00 166.905 37.9042 78.8124 8.32109 0.864310 10 10.00 230.00 1070.31 0.00 165.169 30.2104 78.2489 6.08202 0.861251 11 10.00 234.59 1099.96 0.00 165.482 37.9445 78.2383 6.41608 0.860404 12 10.00 232.32 1067.17 0.00 173.136 29.4910 77.9876 5.93799 0.839244 13 10.00 231.49 1058.50 0.00 175.198 27.5906 77.9103 5.79191 0.833368 14 8.50 230.34 1100.00 0.00 187.817 37.9560 78.7703 10.45610 0.813782 15 8.41 230.00 1099.99 0.00 189.378 37.9524 78.8123 10.82220 0.810429   188  C.3 Response surface methodology principles  Factorial experiments can be designed in various ways depending on the number of factors and levels. The specific case used for the optimization of CNF properties was a 2k factorial design, i.e. two levels (high and low level) for each factor. Typically, 2k designs are useful at the start of a response surface study where screening experiments are performed to identify the important system variables. The 2k design is also used to fit a first-order response surface model to generate factor effect estimates. Using Design-expert software, the design was analyzed to determine interactions and main effects for each response. For example, the software uses visual methods to assist in choosing the significant main effects. In a half-normal plot, significant effects lie far away from the “error” linear line, any effect lying on the line near zero is considered insignificant (Figure C-7). Additionally, the Pareto graph illustrates the magnitude of the effect’s t-value. Effects below the “t-value limit” (black line) are considered insignificant and should be deselected for the model (Figure C-8).      189   Figure C-7: Example of half-normal plot in Design-Expert, here C and D are main effects as they fall far from the linear line 190   Figure C-8: Pareto graph displaying the magnitude of the chosen effects, effects below the lower limit are insignificant  191   Figure C-9: Residuals vs. run number; diagnostic method to validate the model In DOE, the “main effect is defined as the change in response produced by a change in the level on that factor averaged over the levels of the other factor” [152]. By convention, A refers to the effect of factor A, B refers to the effect of factor B, and AB refers to the AB interaction, etc. For instance, mathematically, the main effect of factor A is represented in (C.1). Here, yA+ represents the average response at the points where A is at the high level and yA- is the average response at low level; a, b, ab, and (1) represents the response totals of all replicates (n) taken at the points in the design. Similarly, for main effect of B shown in Equation (C.2). The interaction between AB is expressed in Equation (C.3).  192  𝐴 =  ?̅?𝐴+ − ?̅?𝐴− (C.1) =12𝑛[𝑎𝑏 + 𝑎 − 𝑏 − (1)]  𝐵 =  ?̅?𝐵+ − ?̅?𝐵− (C.2) =12𝑛[𝑎𝑏 + 𝑏 − 𝑎 − (1)] 𝐴𝐵 =12𝑛[𝑎𝑏 + (1) − 𝑎 − 𝑏] (C.3) To determine the significant variables in the process, ANOVA (analysis of variance) is used to examine the magnitude and direction of the effects. The analysis involves calculating sum of squares, mean squares, F0, and P-value for the factors A,B, AB, and etc. For example, sum of squares for A, B, and AB is expressed in Equation (C.4), (C.5), and (C.6). The total sum of squares and error sum of squares is expressed in Equations (C.7) and (C.8).  𝑆𝑆𝐴 =[𝑎𝑏 + 𝑎 − 𝑏 − (1)]2𝑛 × 4 (C.4) 𝑆𝑆𝐵 =[𝑎𝑏 + 𝑏 − 𝑎 − (1)]2𝑛 × 4 (C.5) 𝑆𝑆𝐴𝐵 =[𝑎𝑏 + (1) − 𝑎 − 𝑏]2𝑛 × 4 (C.6) 𝑆𝑆𝑇 = ∑ ∑ ∑ 𝑦𝑖𝑗𝑘2𝑛𝑘=12𝑗=12𝑖=1−𝑦⋯24𝑛 (C.7) SSE = SST – SSA – SSB - SSAB (C.8) 193   Figure C-10: Example contour plot for fiber diameter, factor B and C are fixed at 230 °C and 1100 °C, respectively To verify which effects are statistically significant, an F-test is performed. The higher the F ratio, the more likely that the variance is significantly larger than random error. The F value is defined as the mean square divided by mean square of the error (C.9). If F0 > Fα,k,n−k-1 the hypothesis is rejected and the effect is statistically significant. Alternatively, the hypothesis is rejected if the P-value is less than α. A summary of all the ANOVA computations discussed are generated using Design-Expert and presented in Table C-2. Once the significant main effects and interaction terms are identified, a regression model is fitted to the data. The first-order regression model is shown in Equation (C.10); were x is the coded variables and β are the regression coefficients. This can be used to predict the response at any point in the space spanned by the factors in the design. To validate this model, Design-Expert offers a number of diagnostic tools such as normal plot of residuals, residuals versus predicted response, and residuals versus run plot. An example of residuals (difference between actual vs. predicted) versus run is presented in Figure 194  C-9, here the plot shows reasonable fluctuations about the zero line. If a point falls beyond the red lines (95% confidence controls) then it should be flagged as an outlier. Using the optimization function in Design-Expert, predicted solutions were generated according to input constraints shown in Table 3-6. An example of the software predictions are illustrated in contour and 3D plots, shown in Figure C-10. Generally, this first-order method is used to determine an appropriate direction of potential improvement [152]. A more formal method for the process optimization, method of steepest ascent, may be used to conduct additional experiments and refine the model.     𝐹0 =𝑆𝑆𝑅𝑘⁄𝑆𝑆𝐸(𝑛 − 𝑘 − 1)⁄=𝑀𝑆𝑅𝑀𝑆𝐸 (C.9) 𝑦 =  𝛽0 + 𝛽1𝑥1 + 𝛽2𝑥2 … + 𝐶 (C.10) 195  D Appendix  Appendix D provides procedures for how oxygen transport, ohmic, and ionic resistances were extracted and interpreted from the polarization and EIS data sets. A brief description of EIS theory is included, a more in-depth discussion is detailed in the literature [207,225].   D.1 Resolving oxygen transport resistance from limiting current data   Figure D-1: Limiting current as a function of dry mole fraction of oxygen for four different total pressures in a cell As mentioned in Chapter 6.3.2, oxygen transport resistance was determined using the method proposed by Baker et al. [206]. The method relates limiting current to oxygen transport resistance by Equation (D.1). Limiting currents were extracted from polarization curves measured at four different outlet pressures. The collected data should lie on a straight line passing through the origin as shown in Figure D-1. The slope of this line is proportional to the oxygen transport resistance RMT. This calculation was repeated for each cathode pressure. The result is a list of values of RMT for each cathode pressure tested. The computed RMT was plotted as a function of 196  outlet pressure in Figure D-2. The slope of the linear best fit line corresponds to the slope of RMT with respect to total pressure p. The y-intercept from this plot represents RPI (Figure D-2). Finally, the transport resistance can be further deconvoluted into pressure dependent resistance RPD by Equation (D.2), where Pcathode is the cathode pressure.     𝑅𝑀𝑇 =4𝐹𝑥𝑜𝑖𝑙𝑖𝑚𝑝 − 𝑝𝑤𝑅𝑇 (D.1) 𝑅𝑀𝑇 = 𝑅𝑃𝐼 + 𝑅𝑃𝐷(𝑃) = 𝐼𝑛𝑡𝑒𝑟𝑐𝑒𝑝𝑡 + 𝑆𝑙𝑜𝑝𝑒 · 𝑃𝑐𝑎𝑡ℎ𝑜𝑑𝑒 (D.2)  Figure D-2: Dry transport resistance as a function of total pressure in a cell. The line represents the best linear fit to the four data points   D.2 Introduction to EIS theory EIS is widely used in research areas such as corrosion, electrode kinetics, membranes, batteries, and fuel cells. It is a non-destructive method used to characterize the electrochemical properties of a cell. Electrochemical impedance is usually measured by applying an AC potential to an electrochemical cell and then measuring the current through the cell. For an ideal resistor, 197  the resistance may be calculated by Ohm’s law (R = V/I). Ohm’s law assumes resistance is independent of frequency and that AC currents and voltages through a resistor are in phase with each other. In the real world however, circuit elements exhibit much more complex behavior. As a result, the simple concept of resistance is replaced by a more general circuit parameter—impedance. Unlike resistance, impedance is not limited by the aforementioned assumptions for ideal resistors. Applying a sinusoidal potential excitation results in an AC current signal which can be analyzed as a sum of sinusoidal functions. The electrochemical impedance is measured using a small excitation signal to observe the cell’s response in the pseudo-linear range (Figure D-3). The response shown in Figure D-4, is a sinusoidal current at the same frequency but shifted in phase.     Figure D-3: Current versus Voltage curve showing pseudo-linearity[226]  Figure D-4: Presentation of an AC sinusoidal potential signal (E) with a resulting current response (I) [226] 198  The sinusoidal perturbation voltage applied to the system is expressed as a function of time shown in Equation (D.1), where Et is the potential at time t, E0 is the amplitude of the voltage signal, and ω (rad/sec) is the radial frequency. The relationship between radial frequency ω and frequency f (hertz) is shown in Equation (D.4).  In a linear system, the response signal, It, is shifted in phase (Φ) and has a different amplitude than I0, as shown in Equation (D.5). The impedance of the system then can be expressed by Equation (D.6).   (D.3)  (D.4)  (D.5)  (D.6)  (D.7)  (D.8)  (D.9)  (D.10) 𝑍′ = 𝑅𝑒(𝑍) = |𝑍| cos 𝜙 (D.11) 𝑍" = 𝐼𝑚(𝑍) = |𝑍| sin 𝜙 (D.12)  Because of the complex set of differential equations that is required to relate the electrochemical system properties to the voltage or current excitation responses, a simplified mathematical formula, Euler’s relationship in Equation (D.7), is generally used to express the impedance as a complex function. With this relationship, the AC potential signal and current response can be expressed as Equation (D.8) and Equation (D.9); where j is the imaginary number (√-1). The impedance is then represented as a complex number as expressed in Equation (D.10). The expression for Z is composed of a real and an imaginary part, and can be presented in a 199  complex impedance notation shown in Equation and Equation; where Z = Z’ + jZ”. If the real part is plotted on the X-axis and the imaginary part is plotted on the Y-axis of a chart, we get a "Nyquist Plot" (Figure D-5). Notice that in this plot the Y-axis is negative and that each point on the Nyquist Plot is the impedance at one frequency. For example in Figure D-5, low frequency data are on the right side of the plot and higher frequencies are on the left. The frequencies cannot be plotted on the Nyquist plot, they are usually represented in the Bode plot. Thus, as shown in the figure, the cell impedance is expressed as Equation (D.13) and Equation (D.14). In the Bode plot, the impedance is plotted with log frequency on the X-axis and both the absolute values of the impedance (|Z|= Z0) and the phase-shift on the Y-axis.   Figure D-5: Nyquist Plot with Impedance Vector [226] |𝑍| = (𝑍′2 + 𝑍"2)1/2 (D.13) tan 𝜙 =𝑍"𝑍′ (D.14) An equivalent electrical circuit model is normally used to interpret the EIS data. The circuit model consists of common electrical elements such as resistors, capacitors, and inductors. It is important to use the elements that are based on the physical electrochemistry of the system. The method could be misleading since more than one circuit model could potentially fit the data. Therefore, it is important to understand the electrochemical system and fit the electrical elements 200  which are logically applicable to the actual system. Some example circuits and the corresponding Nyquist plot curve is shown in Figure D-7.    Figure D-6: Example of a Bode plot showing the relationship between log impedance, phase angle and applied frequency[226]  Figure D-7: Nyquist plots for (a) a capacitor, (b) a capacitor in series with a resistor, (c) a capacitor in parallel with a resistor, and (d) a resistor in series with a parallel RC-circuit (Reproduced from [227] with permission of The Royal Society of Chemistry)   201  D.3 Procedure for analyzing EIS data  EIS measurements in H2/N2 consist of measuring amplitude and phase variations of the fuel cell voltage and current when an AC sinusoidal voltage perturbation is applied to the fuel cell. The measurement is repeated over a specified frequency range of 1 Hz – 50 kHz. MEAs were conditioned at the desired RH for at least 20 minutes along with three subsequent cleaning cycles to remove possible contaminants from the cathode surface. The EIS data was collected using CorrWare and analyzed by ZView software. The EIS data is presented as a Complex (Nyquist) plot and Bode plot shown in Figure D-8.   Figure D-8: Example of EIS data visualization in the ZView software, complex plot and bode plot are used to resolve impedances in the fuel cell components  202   Figure D-9: Screenshot of model fitting using Randles circuit in ZView 𝑍 =  𝜔−12√𝑅𝑖𝑜𝑛𝑖𝑐𝐶𝑑√22(1 − 𝑗); where j = √-1 (D.15)  Figure D-10: Example plot of |Z| (magnitude) as the function of ω−1/2 (frequency) for the cell; the slope can be used to determine Rionic Model fitting using equivalent circuits gives results for Rohm and Rionic, which is the total ohmic resistance (resistance to ion migration within the electrolyte, resistance to electron transport within the cell components, and contact resistances) and catalyst protonic resistance, respectively. The particular equivalent circuit used for this data was a Randles circuit with Warburg element (Figure D-9). The Rohm was simply found as the high frequency intercept while Rionic was found 203  by following the computations proposed by Hou et al. [208]. According to Hou, the frequency-dependent impedance is only determined by the ratio of Rionic to Cd as shown in Equation (D.15) (Figure D-10). This step was repeated for all samples at the various operating conditions. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0372790/manifest

Comment

Related Items