Synthesis of nanostructured materials and subsequent processing into nanometer-thick films for supercapacitor applications

Synthesis of nanostructured materials and subsequent processing into nanometer-thick films for supercapacitor applications Beatriz Mendoza Sánchez Wolfson College A thesis submitted for the degree of Doctor of Philosophy University of Oxford Trinity term 2012 Abstract The present thesis work concerns the synthesis and processing of nanostructured materials for supercapacitor applications. Various nanostructured materials were synthesized and subsequently processed into thin film electrodes at a laboratory and/or semi-industrial scale which were then investigated for their potential application as supercapacitor electrodes. In the first part of the work, a spray deposition technique for the manufacture of 1,500 cm2 area electrodes with thickness from 10-25 nm to typically 1-2 µm was developed, optimized and validated. First films of multi-walled carbon nanotubes with demonstrated uniformity of thickness, microstructure, and electrochemical properties across the 1,500 cm2 area were manufactured. Next, molybdenum trioxide was investigated for its potential application as supercapacitor electrode. α-MoO3 nanobelts were synthesized using a hydrothermal method and spray deposited electrodes were tested in various electrolytes. Cyclic voltammetry of α-MoO3 nanobelt electrodes in aqueous electrolytes showed the greatest charge storage in 1 M H2 SO4 with a complex redox activity in a 0 to 1 V (vs Ag/AgCl) electrochemical window. The oxido-reduction v processes were then investigated by a combination of X-ray photoelectron spectroscopy and electrochemical characterization methods. The charge storage properties of α-MoO3 /SWNTs (75 %/ 25 % w/w) (MOSC) electrodes were investigated in LiClO4 /propylene carbonate in a 1.5 to 3.5 V vs Li/Li+ (half cell). Cyclic voltammetry was performed at scan rates from 0.1 mV s−1 to 50 mV s−1 and the separate contributions to charge storage of capacitive and diffusion-controlled 3D ion intercalation processes were estimated. Graphene electrodes were manufactured by a surfactant-water based exfoliation method followed by spray-deposition. Cyclic voltammetry and galvanostatic charge-discharge experiments revealed a combination of electric double layer and pseudocapacitive behavior that was maintained to unusually high scan rates of 10,000 mV s−1 . Finally, the high power density of graphene and high energy density of single-walled carbon nanotubes were combined in a hybrid electrode manufactured following a layer by layer approach (LBL). vi Contents 1 Introduction 1 2 Literature Review 7 2.1 2.2 Basic concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.1.1 Electrochemical energy storage systems . . . . . . . . . 7 2.1.2 Double layer capacitance . . . . . . . . . . . . . . . . . 8 2.1.3 Electrochemical double layer capacitor configuration . . 10 2.1.4 Energy and power density . . . . . . . . . . . . . . . . 12 2.1.5 Pseudocapacitance . . . . . . . . . . . . . . . . . . . . 14 2.1.6 Charge and energy storage 2.1.7 Thermodynamics and kinetics . . . . . . . . . . . . . . 19 . . . . . . . . . . . . . . . 18 Key variables in the optimization of the performance of supercapacitors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.3 2.4 Materials for Supercapacitors . . . . . . . . . . . . . . . . . . 30 2.3.1 Carbon-based materials . . . . . . . . . . . . . . . . . 30 2.3.2 Redox-based electrochemical capacitors . . . . . . . . . 41 Hybrid supercapacitors . . . . . . . . . . . . . . . . . . . . . . 53 2.4.1 Electrolyte considerations . . . . . . . . . . . . . . . . 53 vii 2.4.2 Electrochemical window and charge balance . . . . . . 55 2.4.3 Hybrid supercapacitors using redox positive electrodes 2.4.4 Hybrid supercapacitors using Li-ion intercalation elec- 57 trodes . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 2.5 Optimization of supercapacitor design . . . . . . . . . . . . . . 60 2.5.1 Electrolytes . . . . . . . . . . . . . . . . . . . . . . . . 60 2.5.2 Current collectors . . . . . . . . . . . . . . . . . . . . . 61 2.5.3 Electrode internal resistance . . . . . . . . . . . . . . . 63 2.5.4 Time constant and electrode manufacturing methods . 64 2.6 Supercapacitor applications . . . . . . . . . . . . . . . . . . . 66 2.7 Summary and key opportunities . . . . . . . . . . . . . . . . . 68 3 A scalable spray deposition route for the manufacture of large area nanostructured supercapacitor electrodes. Part 1: optimization of spray deposition variables and development of a numerical model for the optimization of the spraying dynamics. 86 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 3.2 Description of the spray deposition equipment . . . . . . . . . 90 3.3 Variable Optimization . . . . . . . . . . . . . . . . . . . . . . 94 3.3.1 Spray distance z . . . . . . . . . . . . . . . . . . . . . 94 3.3.2 Atomizing pressure P . . . . . . . . . . . . . . . . . . . 95 3.3.3 Mass Distribution . . . . . . . . . . . . . . . . . . . . . 99 3.3.4 Dynamics of spraying . . . . . . . . . . . . . . . . . . . 101 viii 3.4 Summary of optimized spray deposition variables and implications of the experimental set up . . . . . . . . . . . . . . . . 108 4 A scalable spray deposition route for the manufacture of large area nanostructured supercapacitor electrodes. Part 2: validation of the optimization of the spray deposition procedure with the manufacture and investigation of properties of first large area electrodes. 115 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 4.2 Experimental details . . . . . . . . . . . . . . . . . . . . . . . 117 4.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . 120 4.4 4.3.1 XPS characterization of surface functionalized MWNTs 120 4.3.2 Spray deposition of large area uniform thin films . . . . 120 4.3.3 Thickness Analysis . . . . . . . . . . . . . . . . . . . . 121 4.3.4 Surface morphology and electrochemical properties . . 123 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 5 Investigation of nanostructured thin film α-MoO3 based supercapacitor electrodes in an aqueous electrolyte 130 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 5.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 5.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . 137 5.3.1 Material characterization . . . . . . . . . . . . . . . . . 137 5.3.2 Cyclic voltammetry . . . . . . . . . . . . . . . . . . . . 139 5.3.3 XPS surface characterization of the electrodes . . . . . 140 ix 5.3.4 Optimization of electrochemical window and cycling behaviour . . . . . . . . . . . . . . . . . . . . . . . . . 147 5.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 6 An investigation of nanostructured thin film α-MoO3 /singlewalled carbon nanotube electrodes in LiClO4 /propylene carbonate for supercapacitor/battery applications 160 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 6.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 6.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . 165 6.3.1 Material characterization . . . . . . . . . . . . . . . . . 165 6.3.2 Electrochemical characterization: cyclic voltammetry . 165 6.3.3 Electrochemical characterization: galvanostatic discharge curves . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 6.3.4 6.4 Electrochemical characterization: EIS analysis . . . . . 181 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193 7 Scaleable ultra-thin and high power density graphene supercapacitor electrodes manufactured by aqueous exfoliation and spray deposition. 201 7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202 7.2 Experimental Details . . . . . . . . . . . . . . . . . . . . . . . 205 7.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . 208 7.3.1 Graphene characterization . . . . . . . . . . . . . . . . 208 7.3.2 Electrochemical characterization of graphene electrodes 212 7.3.3 Scaleability . . . . . . . . . . . . . . . . . . . . . . . . 222 x 7.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223 8 Engineering hybrid electrodes combining the properties of graphene and SWNTs using a layer-by-layer manufacturing method 232 8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232 8.2 Experimental details . . . . . . . . . . . . . . . . . . . . . . . 234 8.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . 235 8.4 Conclusions and future work . . . . . . . . . . . . . . . . . . . 240 9 Conclusions and future work 244 xi List of Figures 1.1 Energy conversion, storage and conservation technologies . . . 2 2.1 Schematic of an EDLC . . . . . . . . . . . . . . . . . . . . . . 11 2.2 Ragone plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 2.3 Charging and discharging scheme of a battery and an EDLC . 19 2.4 Diagram illustrating energy storage of a battery and an EDLC 20 2.5 Reversibility of capacitor versus irreversibility of batteries . . . 23 2.6 Equivalent electrical circuit of a supercapacitor 2.7 Deposition of pseudocapacitive materials onto 3D nanostructures 46 2.8 Comparative graph of pseudocapacitive materials . . . . . . . 52 2.9 Schematic of a hybrid supercapacitor . . . . . . . . . . . . . . 54 . . . . . . . . 25 2.10 Optimization of the working electrochemical window in a hybrid supercapacitor . . . . . . . . . . . . . . . . . . . . . . . . 57 2.11 Supercapacitor applications . . . . . . . . . . . . . . . . . . . 67 3.1 Optical images of the spray deposition equipment . . . . . . . 92 3.2 Optimization of spray distance z 3.3 Arithmetic average roughness Ra of a PFA film as function of . . . . . . . . . . . . . . . . 95 atomizing pressure . . . . . . . . . . . . . . . . . . . . . . . . 96 xii 3.4 Thickness of a PFA film as function of atomizing pressure . . . 98 3.5 Mass distribution of spray footprint . . . . . . . . . . . . . . . 100 3.6 Matlab computed spraying patterns . . . . . . . . . . . . . . . 103 3.7 Matlab computed spray-deposited films . . . . . . . . . . . . . 105 3.8 Dynamics of spraying . . . . . . . . . . . . . . . . . . . . . . . 106 3.9 Spacings xpar (i,j) and their frequency as spray deposition progresses up to revolution r . . . . . . . . . . . . . . . . . . . . 107 3.10 Shortest xpar at revolution r . . . . . . . . . . . . . . . . . . . 108 3.11 Matlab computed spraying pattern obtained by data logging experimental vy . . . . . . . . . . . . . . . . . . . . . . . . . . 110 3.12 Spacings xpar (i,j) and their frequency as spray deposition progresses up to revolution r for experimental vy . . . . . . . . . 111 4.1 XPS of carboxyl-functionalized MWNTs . . . . . . . . . . . . 121 4.2 Optical images of flexible 100 cm x 15 cm MWNTCOOH electrode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 4.3 Thickness measurements of a 100 cm x 15 cm MWNTCOOH thin film electrode 4.4 . . . . . . . . . . . . . . . . . . . . . . . . 122 Optical images of 2.5 cm x 2.5 cm MWNTCOOH films onto ITO coated glass substrates . . . . . . . . . . . . . . . . . . . 123 4.5 SEM images of the 100 cm x 15 cm MWNTCOOH electrode . 124 4.6 Cyclic voltammetry of the 100 cm x 15 cm MWNTCOOH electrode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 4.7 EIS of the 100 cm x 15 cm MWNTCOOH electrode . . . . . . 126 xiii 4.8 Cyclic voltammetry analysis of MWNTCOOH electrodes with increasing number of spray-deposited layers . . . . . . . . . . 127 5.1 SEM images and XRD pattern of α-MoO3 electrodes . . . . . 138 5.2 TEM images of α-MoO3 electrodes . . . . . . . . . . . . . . . 139 5.3 Cyclic voltammetry of α-MoO3 electrodes . . . . . . . . . . . 140 5.4 Cyclic voltammetry of α-MoO3 electrodes in neutral electrolytes140 5.5 XPS spectrum in the Mo 3d binding energy region of as made α-MoO3 nanobelts . . . . . . . . . . . . . . . . . . . . . . . . 141 5.6 XPS spectra in the Mo 3d binding energy region of α-MoO3 electrodes polarized at (a) E = 1 V, (b) E = 0 V . . . . . . . 143 5.7 Plot showing percentage concentration of Mo oxidation states calculated from spin-orbit components Mo 3d5/2 . . . . . . . . 144 5.8 XPS spectra in the O 1s binding energy region of α-MoO3 electrodes polarized at (a) E = 1 V, (b) E = 0 V . . . . . . . 145 5.9 Electrochemical window showing decrease of charge storage . . 148 5.10 Optimized electrochemical window . . . . . . . . . . . . . . . 149 5.11 Variation of current with scan rate for optimized electrochemical window . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150 5.12 A negative electrochemical window . . . . . . . . . . . . . . . 151 5.13 A SWNT/α-MoO3 composite . . . . . . . . . . . . . . . . . . 152 6.1 SEM image of a 75 %/25 % w/w α-MoO3 /SWNTCOOH electrode 6.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165 Cyclic voltammograms of a MOSC electrode at 0.1 mV s−1 . . 167 xiv 6.3 Cyclic voltammograms of a MOSC electrode at different scan rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 6.4 Capacitance and charge storage of a MOSC electrode . . . . . 170 6.5 Power dependence (p) on scan rate νs of cathodic current Ic ∝ νsp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 6.6 Cyclability of a MOSC electrode at 10 mV s−1 . . . . . . . . . 173 6.7 Discharge curves of a MOSC electrode at 10 mA g−1 . . . . . 176 6.8 Capacity retention during galvanostatic discharge of a MOSC electrode at 500 mA g−1 . . . . . . . . . . . . . . . . . . . . . 178 6.9 EIS spectra of a MOSC electrode at various depth of discharge points. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186 6.10 Nyquist plots of a MOSC electrode at various depth of discharge points . . . . . . . . . . . . . . . . . . . . . . . . . . . 187 6.11 Equivalent electrical circuit used to fit the EIS experimental data of a MOSC electrode. . . . . . . . . . . . . . . . . . . . . 189 6.12 Equivalent series resistance Rs versus depth of discharge . . . 191 6.13 Charge transfer at a MOSC electrode. . . . . . . . . . . . . . 192 7.1 TEM images of graphene flakes . . . . . . . . . . . . . . . . . 209 7.2 Raman spectrum of a spray-deposited graphene electrode . . . 210 7.3 X-ray photoelectron spectroscopy of a spray-deposited graphene electrode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211 7.4 Cyclic voltammograms of graphene electrodes before annealing 213 7.5 Charge storage of a graphene electrode and comparison with a SWNT electrode . . . . . . . . . . . . . . . . . . . . . . . . 215 xv 7.6 Cyclic voltammetry of graphene electrodes before and after annealing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 7.7 Galvanostatic charge-discharge curves of graphene electrodes . 219 7.8 EIS analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222 7.9 Cycling testing of a graphene electrode and a 100 cm x 15 cm flexible graphene electrode . . . . . . . . . . . . . . . . . . . . 223 8.1 TEM images of spray deposited graphene at increasing times . 236 8.2 TEM images of spray deposited SWNTCOOH at increasing times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237 8.3 Cyclic voltammetry of a LBL7 electrode . . . . . . . . . . . . 239 8.4 Capacitance versus scan rate for a composite electrode graphene/SWNTCOOH (50 %/ 50 % w/w) . . . . . . . . . . . . . . . . . . . . . . . . 240 A1 Equivalent electrical circuit used to fit the high frequency EIS data of a graphene electrode . . . . . . . . . . . . . . . . . . . 249 A2 Galvanostatic charge-discharge cycling tests of graphene electrodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252 A3 EIS of a 40 nm graphene electrode . . . . . . . . . . . . . . . 253 xvi List of Tables 2.1 Comparative table of properties of a battery, electrostatic capacitor and EDLC . . . . . . . . . . . . . . . . . . . . . . . . 14 2.2 Comparative characteristics of battery and supercapacitor behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 3.1 Summary of optimized spray deposition variables . . . . . . . 109 5.1 XPS binding energies (eV) for α-MoO3 electrodes polarized at various potentials E 6.1 . . . . . . . . . . . . . . . . . . . . . . . 142 A summary of the performance of MoO3 electrodes in organic electrolytes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179 6.2 Major composition and structural changes of MoO3 during Liion insertion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182 6.3 Fitting parameters of equivalent electrical circuit 1 . . . . . . 190 6.4 Fitting parameters of equivalent electrical circuit 2 . . . . . . 190 A1 Summary of best-fitting parameters to high frequency EIS data of a graphene electrode . . . . . . . . . . . . . . . . . . . 251 xvii Chapter 1 Introduction Current worldwide energy consumption is approximately 14 TW and forecast to be 25-30 TW by 2050. The main energy use sectors are: industry (37 %), commercial and personal transportation (20 %), residential homes (11 %), and losses during energy generation and transmission (26 %). The largest percentage of this energy (80 %) currently comes from CO2 emitting fossil fuels (oil, coal, and natural gas) [1] which is a serious environmental threat causing global warming and pollution with concomitant negative consequences on disruption of natural habitats, water resources, agriculture and human health. Moreover fossil fuels are non-renewable energy sources with oil, and natural gas predicted to be exhausted in 70-80 years [2] . Alternative carbon dioxide-free and renewable energy sources are therefore actively being sought including: nuclear power, hydrothermal power, geothermal power, wind power, and solar power [1, 3] . The use of these alternative sources of energy demands the development of new materials and technologies for energy conversion, storage and conservation, and the development of supporting in1 frastructure at a large scale and in a short time enabling the efficient use of these new technologies at the lowest cost and in a environmentally responsible way. Energy storage systems are of paramount importance to push forward the deployment of intermittent energy sources such as wind and solar power otherwise poorly exploited [2] . In order to make wind, tidal and solar to electrical energy conversion sustainable and economically competitive, primary generation (consisting for instance of photovoltaic and thermophotovoltaic cells) needs to be coupled to high capacity energy storage systems such as batteries and electrochemical capacitors [1] . Figure 1.1 shows this concept in a flow chart of renewable technologies falling in broad categories of energy conversion, storage and conservation. Figure 1.1: Solar/thermal/electrochemical energy conversion, storage and conservation technologies highlighting the role of electrochemical energy storage systems. Reproduced from reference [1] . Electrical energy storage systems include solid state capacitors, fuel cells, 2 batteries, and electrochemical capacitors. Electrochemical capacitors (ECs) are essentially high power density energy storage systems which, unlike batteries, are able to fully charge and discharge in seconds and provide a long cycle life. However, the energy density of ECs is lower than that of batteries, and ECs are considered energy storage devices whose characteristics complement those of batteries. The main niche markets include hybrid electrical vehicles (HEV) and portable electronics [4, 5] where ECs allow battery size and weight reduction. Applications making use of the high power density and long cycle life of electrochemical double layer capacitors are described in Chapter 2. In order to meet energy storage demands for current and future technological applications, the performance of ECs needs to be improved. A great part of the efforts to improve energy density of ECs are focused on the design of new nanostructured materials serving as electrodes where pseudocapacitance and double layer capacitance effects are combined [1, 6–8] . A better understanding of the interaction between electrode and electrolytes at the nanoscale would also greatly help to design new nanostructured materials. The concept of hybrid electrochemical capacitors is currently considered the most promissing approach to improve the energy density performance of ECs where a key element is the conjunct design of electrode, electrolyte, and other manufacturing aspects. Technologically, the most important consideration is the cost of materials and scalability of manufacturing methods. Research work on ECs has yielded results that suggest a wide array of different materials and ECs configurations that show a better laboratory performance than that of the already excellent and commercially available 3 ”symmetric activated carbon” EC. However, either little or no account has been given to the scalability of these newer approaches and commercial impact has been negligible. The objectives of this work are: • To commission a new spray deposition process that scales up the manufacture of laboratory scale ECs’ electrodes to large area (1500 cm2 ) nanostructured electrodes. • To optimize variables of the spraying process in order to produce safely and reproducibly thin film, large area nanostructured electrodes. • To validate the optimization of the spray deposition process by manufacturing and characterizing large area thin film nanostructured electrodes made of carbon nanotubes. • To investigate new nanostructured materials either as a single material or in a composite format for supercapacitor applications including: electroactive material synthesis and characterization, electrode manufacturing by spray deposition, electrode testing and improvement of performance. The particular emphasis is on materials and methods to synthesize and then process electroactive materials to manufacture electrodes in a way that offers scope for technological impact. • To investigate the fundamental energy storage mechanism of the electroactive materials by making use of standard electrochemical characterization methods and spectroscopy techniques in order to judge its 4 suitability for supercapacitor applications and elucidate opportunities for improvement of performance. The thesis is structured as follow. The present introduction (Chapter 1) places the reader in the broad context of the need for efficient, economic and sustainable energy storage systems where ECs play a key role. Next, in the literature review (Chapter 2) the reader is provided with all the fundamental concepts about ECs discussed in comparison with batteries and a broad overview about the current state of the art in the research field and opportunities is given. Next, a spray deposition technique is developed, optimized, and validated for the fabrication of thin film large 1,500 cm2 nanostructured films (Chapter 3 and 4). Then, materials for electrochemical capacitors are investigated including molybdenum oxide in an aqueous electrolyte (Chapter 5), and in a organic electrolyte (Chapter 6), graphene (Chapter 7), and a graphene-single walled carbon nanotube combination in a layer by layer manufacturing approach (Chapter 8). Finally overall conclusions and future work are addressed (Chapter 9). 5 Bibliography [1] J. Baxter, Z. Bian, G. Chen, D. Danielson, M. S. Dresselhaus, A. G. Fedorov, T. S. Fisher, C. W. Jones, E. Maginn, U. Kortshagen, A. Manthiram, A. Nozik, D. R. Rolison, T. Sands, L. Shi, D. Sholl and Y. Wu, Energy Environ. Sci., 2009, 2, 559–588. [2] V. Arunachalam and E. Fleischer, MRS Bull., 2008, 33, 261–261. [3] M. Z. Jacobson, Energy Environ. Sci., 2009, 2, 148–173. [4] J. Tollefson, Nature, 2008, 456, 436–440. [5] M. Armand and J. M. Tarascon, Nature, 2008, 451, 652–657. [6] X. Zhao, B. M. Sánchez, P. J. Dobson and P. S. Grant, Nanoscale, 2011, 3, 839–855. [7] A. S. Arico, P. Bruce, B. Scrosati, J. M. Tarascon and v. W. Schalkwijk, Nat. Mater., 2005, 4, 366–377. [8] D. R. Rolison, J. W. Long, J. C. Lytle, A. E. Fischer, C. P. Rhodes, T. M. McEvoy, M. E. Bourg and A. M. Lubers, Chem. Soc. Rev., 2009, 38, 226–252. 6 Chapter 2 Literature Review 2.1 2.1.1 Basic concepts Electrochemical energy storage systems Energy can be stored in different ways: (1) chemically such as in fuels, and biofuels where energy is converted directly into thermal (heat) or mechanical (combustion engines) energy, (2) electrostatically such as in solid state capacitors where electrical energy is stored in a electrostatic field created by a separation of positive and negative charges in a dielectric medium that fills the gap between two plates, and (3) electrochemically where chemical energy is converted into electrical energy such as in fuels cells, batteries and electrochemical capacitors [1] . Electrochemical energy storage systems share the fundamental principle of having an electrode-electrolyte interface where energy-providing processes take place. However, the mechanisms for energy storage are different resulting in different performance characteristics [2] . 7 • Fuel cells. Electrical energy is released upon oxidation-reduction processes between a continuously supplied fuel such as hydrogen, natural gas or ethanol, and an oxidant such oxygen, air or hydrogen peroxide. • Batteries. Electrical energy is released upon oxidation-reduction and/or ion intercalation processes of electrochemically active materials deposited at conductive substrates (anode and cathode). The performance of batteries is controlled mainly by reaction kinetics, mass transport and phase transformations of the electrochemically active materials. • Electrochemical capacitors. Electrical energy is stored at an electrical double layer or charge separation at the electrode-electrolyte interface. Fundamental principles of electrochemical capacitors share similarities and contrasts with batteries and thus, the understanding of the basic functioning of electrochemical capacitors is often addressed in the literature by comparison with batteries. The next sections describe fundamental principles of electrochemical capacitors along with differences, advantages and disadvantages when compared with batteries. 2.1.2 Double layer capacitance Electrostatic or double layer capacitance takes place when charge separation occurs upon polarization at an electrode-electrolyte interface. The concept of a double layer corresponds to a model first proposed by Helmholtz in 1853 for colloid interfaces where two layers of opposite charges are separated by 8 a short distance of atomic dimensions [3] . The model was later adapted to electrode-electrolyte interfaces where, as shown in Figure 2.1a, on the electrode (metal) side, an excess or deficiency of electrons gives rise to a net positive or negative surface charge density which in turn affects the concentration of solvated ions on the solution side of the electrode interface leading to the formation of a double layer. The Helmholtz model was successively refined by Gouy and Chapman who included thermal fluctuations that lead to a diffuse layer of ions, and then by Stern who introduced the concept of an inner layer consisting of adsorbed ions according to Langmuir’s adsorption isotherm beyond which an external diffuse region of ionic charge could be considered. Finally, Grahame took into account the size of hydrated ions populating inner and outer Helmholtz layers. The Grahame model is depicted in Figure 2.1b showing the inner and outer Helmholtz layers, and the diffuse ion distribution region extended beyond the outer Helmholtz layer. The thickness of the diffuse region is < 10 nm for solutions with ionic concentrations greater than 10−2 M The double layer capacitance C is mathematically formulated as: C = κr A/d (2.1) where κr = ǫr κ0 is the electrolyte permittivity with ǫr being the electrolyte dielectric constant, and κ0 the permittivity of the free space, d is the effective thickness of the double layer, and A is the electrode surface area. 9 2.1.3 Electrochemical double layer capacitor configuration Electrochemical double layer capacitors (EDLCs) store energy via the electrochemical double layer. As in a battery, it requires two electrodes, one of which is charged negatively with respect to the other. Delivery of energy occurs upon reversible formation and release of electrical double layers resulting in movement of electrons in a external circuit. The basic components of an EDLC are shown in Figure 2.1a. Because the capacitance of an EDLC is proportional to surface area, ideal EDLC electrode materials must have a high specific surface area (SSA), for which materials with high porosity are generally preferred (see later). Thus, porous electrochemically active materials deposited on current collectors usually constitute both the cathode and the anode. Cathode and anode can consist of the same (symmetric configuration) or different (asymmetric configuration) electrochemically active materials. An electrolyte at the interface with the electrochemically active material, which provides purely ionic conductivity, allows for the formation of the electrical double layer. A separator between cathode and anode provides a physical barrier to prevent electrical shorting. The separator is permeable to ions and can be a gelled electrolyte or a microporous plastic film. The separator must be inert in the EDLC environment. Changes of potential due to flow of charge in the external circuit and within the solution only charge the double layer with no charge passing across the double layer. The double layer formation time is approximately 10−8 s having the capability of responding rapidly to electrode potential changes [2] . A typical value 10 of the double layer capacitance for carbons and metals is 10-40 µF cm−2 [2] . Figure 2.1: (a) Schematic of a double layer capacitor [4] , and (b) schematic of the Grahame model for the electrical double layer [5] . IHP and OHP stand for inner and outer Helmholtz plane respectively; σ i is the charge density of specifically adsorbed ions in the IHP and σ d is the excess charge density in the diffuse layer; φ1 , and φ2 are the potentials at distance x1 and x2 respectively; φM and qM are the potential and charge respectively of the metal electrode. 11 2.1.4 Energy and power density The terms specific energy (watt-hours per kilogram, Wh kg−1 ) and energy density (Watt-hours per litre, Wh L−1 ) are used to compare the energy contents of energy storage systems (see section 2.1.6 and Equation 2.12). The rate capability of a system is expressed as specific power (W kg−1 ) and power density (W L−1 ) (see Equation 2.18). A comparison of power and energy capabilities of different energy storage devices is commonly done in a Ragone plot [2, 6] . Figure 2.2 shows a Ragone plot indicating the approximate positions of various energy storage technologies. Electrochemical capacitors fill the gap between conventional capacitors [3] and batteries. Electrochemical capacitors store more energy than conventional capacitors (20-200 times larger capacitance) due to the higher surface area available for charge storage in the electrical double layer [7, 8] . However, the electrostatic surface charging mechanism provides a energy density that is inferior to that of batteries. A fundamental difference between EDLCs and batteries is that the energy storage mechanism of EDLCs is a surface process whereas in a battery bulk Faradaic reactions are involved. Major implications include: (1) energy delivery and uptake in EDLCs is faster than in batteries; (2) EDLCs do not have limitations imposed by electrochemical kinetics through a polarization resistance (kinetic hindrance of the charge-transfer reaction taking place [2] ), and diffusion constraints inherent to bulk processes of batteries; and (3) charging and discharging of an EDLC does not involve chemical phase and composition changes that causes a characteristic swelling of active materials in batteries. As a consequence, EDLCs have a higher power density than batteries and a 12 much higher cyclability, i.e. upon repeated charge/discharge cycles EDLCs charge/discharge within a shorter time and keep their charge capacity and energy storage characteristics for a larger number of cycles than batteries. The charge-discharge time of an EDLC is typically 60-120 s, whereas in batteries, which are designed primarily as energy storage devices, the charge-discharge time is of at least 10-15 min. EDLCs can sustain millions of cycles whereas batteries survive a few thousands at best. A carbon-carbon EDLC for instance can survive 500,000 cycles whereas the cyclability of a battery ranges between 500 and 2,000 cycles [9] . A comparison of the properties of batteries, EDLCs and electrostatic capacitors, as presented by Zhang et al., is shown in Table 2.1 [10] . Figure 2.2: Ragone plot showing energy storage domains for various energy storage technologies. Reproduced from [2] . 13 Table 2.1: Comparative table of properties of a battery, electrostatic capacitor and EDLC. Reproduced from [10]. Discharge time Charge time Energy density (Wh kg−1 ) Specific power (W kg−1 ) Charge-discharge efficiency Cycle life (cycle number) 2.1.5 Battery 0.3-3 h 1-5 h 10-100 50-200 0.7-0.85 500-2000 Electrostatic capacitor 10−3 -10−6 s 10−3 -10−6 s < 0.1 > 10,000 ≈1 > 500,000 EDLC 0.3-30 s 0.3-30 s 1-10 ≈ 10,000 0.85-0.98 > 100,000 Pseudocapacitance In order to increase the energy density of EDLCs, advanced electrochemical capacitors make use of additive pseudocapacitance effects. Pseudocapacitance takes place when Faradaic charge is transfered across an electrodeelectrolyte interface in a quasi-two-dimensional, fast and reversible process. The term pseudocapacitance is used to distinguish it from double layer capacitance where a true electrostatic capacitance occurs. The electrons in a Faradaic process are transferred to or from valence electron orbitals of electrochemically active material whereas the electrons involved in a electrostatic double layer are conduction band electrons of a conductive material. Pseudocapacitance was first studied in underpotential deposition of H or Pt systems and in RuO2 redox systems by Conway et al. in 1975 and the term supercapacitor was then coined [11] . The term “super” or “ultra” makes reference to the superior energy density storage of electrochemical capacitors when compared with conventional solid state capacitors [12, 13] . In this sense, EDLCs only, and EDLCs adding pseudocapacitance are both super- or ultracapacitors, although the term “supercapacitor” is more commonly used for the latter case. Here we adopt this convention and use the term supercapac14 itor to refer to electrochemical capacitors where double layer capacitance is accompanied by Faradaic pseudocapacitive processes so that the total capacitance is the sum of the double layer capacitance plus the pseudocapacitance. Pseudocapacitance has been given two kinds of theoretical treatment: a thermodynamic equilibrium approach and a kinetic approach for processes that exhibit pseudocapacitance under conditions displaced from equilibrium [3] . From a thermodynamic point of view, pseudocapacitance originates when a property y, proportional to charge, is related to electrical potential V by an equation of the form, y/(1 − y) = Kexp(V F/RT ) (2.2) where K is an equilibrium constant, R is the gas constant, T is temperature and F is the Faraday constant. The quantity y is defined according to four distinguishable pseudocapacitive processes involving chemical changes of state of reactant species as a result of charge transfer: 1. Electrosorption of a given species on a substrate where y is the extent of a two-dimensional (2D) fractional coverage θ of the substrate surface. A well-studied electrosorption process is the underpotential deposition of adatoms such as H at Pt or metals adatoms such as Pb at Au or Ag: H3 O + + M + e ⇀ ↽ M Hads + H2 O For this process, Equation 2.2 becomes, 15 (2.3) θ/(1 − θ) = KcH + exp(V F/RT ) (2.4) where θ= MHads is the fractional electrode surface coverage of H3 O+ onto metal M=1- θ, and cH + is the concentration of H3 O+ . 2. Intercalation of a species into a layer-lattice host, e.g. Li+ intercalated into layer-lattice hosts such as TiS2 , MoS2 , CoO2 , and V6 O13 where y is defined as the site fraction occupancy X of the intercalated guest atom: X/(1 − X) = Kexp(V F/RT ) (2.5) This Li-intercalation pseudocapacitive process is considered to be a transition case between a three dimensional (3D) ion intercalation in a battery and a purely 2D pseudocapacitive process. Although Liintercalation into a host crystal is essentially a 3D process, in the case of layered hosts such as TiS2 , the charge and discharge curves, and the cyclic voltammograms resemble a 2D electrosorption process. Also, 3D sorption isotherms of Li intercalated into TiS2 have a similar form than isotherms of A true 2D electrosorption process [3, 11, 13] . Chapter 6 deals with this type of pseudocapacitive process for Li intercalation into α-MoO3 layered crystal. 3. Redox processes where oxidation-reduction of species take place, which can be expressed in the general form: 16 Ox + e− ⇀ ↽ Red (2.6) where Ox and Red refer to the oxidized and reduced species respectively. Here y = ℜ = [Ox]/([Ox] + [Red]), ℜ/(1 − ℜ) = Kexp(V F/RT ) (2.7) 4. Specific adsorption is another type of pseudocapacitance where ions, usually anions, not only interact electrostatically with the electrode surface, as in a electrostatic double layer, but an interaction of valence electrons of the ions and the electrode’s surface orbitals also occurs. This phenomenon is also known as chemisorption and is described by the following quasi-chemical equation where partial charge transfer occurs from an anion A− at an electrode M : M + A− → M/A(1−δ) + δe − (2.8) where δ is an electron charge fraction in M. The kinetics of pseudocapacitive processes is described in Section 2.1.7. 17 2.1.6 Charge and energy storage During charge of a battery, ideally a thermodynamic potential exists independent of the extent of the charge Q added as long as the oxidized and reduced forms of electrochemically active material remain constant, i.e. battery potential VB = constant. In contrast, in an EDLC, every additional charge added has to do electrical work against the charge density already accumulated progressively increasing the interelectrode potential difference so that: VC = Q C (2.9) where VC is the EDLC potential, Q is charge and C is the double layer capacitance. From this equation it follows that 1 dV = dQ C (2.10) A comparison of charge and discharge curves for an ideal battery and an ideal EDLC is shown in Figure 2.3 where voltage maintains constant during charging and discharging of a battery whereas there is a linear variation with slope 1/C of voltage as a function of charge for an ideal EDLC. The energy stored by a battery EB is given by: EB = Z VB dQ = QVB whereas the energy stored by an EDLC EC is given by: 18 (2.11) EC = Z VC dQ = Z Q 1 Q2 1 1 dQ = = QVC = CVC2 C 2 C 2 2 (2.12) Therefore, EC = 21 EB , which is illustrated in Figure 2.4 where the line and squared shaded areas represent EB and EC respectively. Figure 2.3: Comparison of charge-discharge processes in a battery and an EDLC: potential as a function of charge. Reproduced from [3] . In the case of a pseudocapacitor, two characteristics of the charge and discharge behavior distinguish it from a battery. Resembling an EDLC: (1) the variation of voltage with charge is approximately linear and, (2) upon linear variation of potential with time, s = dV/dt, a constant or almost constant current arises: ±I = ±sC 2.1.7 (2.13) Thermodynamics and kinetics The thermodynamic and kinetic behaviour of pseudocapacitor and battery systems are fundamentally different [11, 14] . Pseudocapacitance processes are 19 Figure 2.4: Energy storage of a battery at ideal constant charging voltage VB and for an EDLC with progressively changing voltage VC . Reproduced from [3] . essentially 2D thermodynamically reversible processes in which a monolayer or quasi-monolayers of electrochemical reactive species can be electrosorbed or electrodesorbed with charge transfer. In a pseudocapacitor, the degree of conversion of a given species to the other is a continuous function of potential as described in Equation 2.2. In contrast, in a battery, reagents are well defined 3D phases which have unique chemical potentials or Gibbs energy. Their conversion from one state to the other proceeds ideally at a singular potential until one of the phases is converted to another. For the electrosorption pseudocapacitive process described in Equation 2.3, a Langmuir-type electrosorption isotherm is expressed in Equation 2.4. If the charge for the formation of a monolayer of H on M is q, then the pseudocapacitance Cφ can be expressed as: 20 qF Kc+ H exp(V F/RT ) Cφ = q(dθ/dV ) = · 2 RT [1 + Kc+ H exp(V F/RT )] (2.14) where K is the equilibrium constant in the electrosorption process described in Equation 2.3, cH + is the concentration of H3 O+ , and the other variables are defined as for Equation 2.2. In terms of θ: Cφ = qF · θ(1 − θ) RT (2.15) The ideal equilibrium state described in Equation 2.3 is not practically realized, and deviation from equilibrium for 2D pseudocapacitance occurs. Under non equilibrium conditions, a Faradaic current i arises: i/zF = q K1 cH + exp[−V (t)F/RT ] ·s · RT {K 1 cH + + exp[−V (t)F/RT ]}2 (2.16) where K1 = k1 /k−1 , k1 and k−1 are the kinetic rate constants for the forward and backwards directions of the electrosorption process described in Equation 2.3, z is the charge transfer, and s = dV(t)/dt for V(t) = Vt=0 + st for a linear voltage sweep. The same kinetic treatment explained here for an electrosorption process applies for the other pseudocapacitive processes described in Section 2.1.5. When the pseudocapacitive process in Equation 2.3 is subject to an increasing linear voltage sweep s, the kinetic reversibility is eventually lost for s > s0 where s0 is the transitional voltage sweep rate. Realization of kinetic reversibility distinguishes batteries from supercapacitors in a fundamental way. When analyzed by cyclic voltammetry (a 21 electroanalytical method commonly used to study electrochemical systems) ordinary batteries exhibit irreversibility during anodic and cathodic potential sweeps. The irreversibility arises when a phase is converted to another having different structures during the charge-discharge process. A supercapacitor gives complete reversibility for a range of potentials at sufficient small sweeping rate dV /dt while the battery never or rarely shows reversibility [3, 11] . For a reversible response to positive and negative sweeps, the voltammogram for one direction of the sweep is the mirror image of the negative-moving sweep. Irreversibility of a process is manifested by an asymmetric voltammogram with non-mirror images of positive and negative potential sweeps. A comparison of cyclic voltammogams of a RuO2 redox pseudocapacitor and a Pb/PbCl2 battery are shown in Figure 2.5. As explained in section 2.1.4, differences in kinetic reversibility have important implications in the power density, and cyclability behaviour of batteries and supercapacitors. Table 2.2 summarizes the characteristics of batteries and supercapacitors. 22 Figure 2.5: Cyclic voltammograms of (a) RuO2 in 1M H2 SO4 showing mirrorimage symmetry, and (b) Pb-PbCl2 battery showing typical irreversible behaviour. Reproduced from [11] . 23 Table 2.2: Comparative characteristics of battery and supercapacitor behaviour. Reproduced from [11] . Battery Ideally has single-valued free energies of components. EMF is ideally constant with degree of charge/discharge, except for non-thermodynamic incidental effects, or phase change during discharge. Behaviour is not capacitive, except in very general sense. Irreversibility is usual behaviour (materials and kinetic irreversibility). Response to linear modulation of potential gives rise to i vs. V irreversible profile with nonconstant currents. Discharge at constant current arises at a more less constant potential. Supercapacitor Has continuous variation of free energy with degree of conversion of materials. Potential is thermodynamically related to state of charge through log(X/1-X) factor, in a continuous manner. Behaviour is capacitive. High degree of reversibility is common (104 -106 cycles with RuO2 ). Response to linear modulation of potential gives more or less constant charging current. profile. Discharge at constant current gives linear decline of potential with time, as with a capacitor. 24 2.2 Key variables in the optimization of the performance of supercapacitors Important variables to observe in order to improve supercapacitor energy density are evident from Equation 2.12: capacitance C and potential V . Capacitance, in turn is optimized by the use of high SSA and conductive materials which favours charge-transfer processes. In the case of power density, further insight into the impedance behaviour of a supercapacitor is needed. The impedance behaviour of an electrochemical system can be best interpreted by consideration of equivalent electrical circuits. An electrochemical capacitor can be modeled, in its simplest form, by the equivalent-circuit model shown in Figure 2.6 where Cdl is the double layer capacitance, Rs represents the internal resistance of the capacitor coming from the electrolyte and electrode-current collector resistance (often referred to as equivalent series resistance (ESR) and calculated from electrochemical impedance spectroscopy (EIS) studies), RF is the Faradaic (potential dependent) resistance, also called polarization resistance or charge transfer resistance. Figure 2.6: Equivalent electrical circuit of an electrochemical capacitor. The dynamics of coupled resistor-capacitor systems is determined by the 25 so called time constant τ = RC, where R determines the rate (current I ) at which charge can be delivered for a given voltage V, across a resistor; C determines the extent at which charge q can be accommodated on the double layer capacitor brought to a voltage V . In other words, the RC product determines the effective duration of the charging or discharging processes for some initial applied voltage on the capacitor. In the circuit in question there are two time constants: Rs C is the time constant for charging and discharging the circuit, and RF C is the self-discharge constant. Because self-discharge processes are undesirable in a energy storage system, the self-discharge time constant RF C must be maximized whereas to achieve high power density, Rs C must be minimized. This effect is evident when considering the maximum power delivery of a supercapacitor P : P = V 2 /4Rs (2.17) The estimation of Rs including ion resistance and effects of current transients in the electrodes is not simple but a frequently used approximation is Rs = (2/3)t x r’/A where t is the thickness of the electrode, r’ is the resistivity of the electrolyte and A is the geometric area of the electrode. Therefore, reduction of the electrode thickness contributes to the reduction of the ESR in a supercapacitor electrode. As further discussed in Chapter 2, electrode thickness and accurate control of mass load constitute critical elements of electrode design for improved performance. Cyclability tests of a supercapacitor evaluate the degree to which the system keeps its charge capacity and energy storage characteristics upon 26 repeated charge-discharge cycles. For a two-electrode configuration supercapacitor, a single charge-discharge test at constant current is used to calculate specific power and specific density as follows: ∆V I m (2.18) E(W h/kg) = P t (2.19) P (W/kg) = where ∆V = (Vmax + Vmin )/2. Vmax and Vmin are the potential at, respectively, the beginning and end of discharge, I is the charge-discharge current, t is the discharge time and m is the mass of the electrode in cathode and anode [9, 15] . For an optimum energy density, 100 % of the electrode mass should be electrochemically active contributing to charge storage. In practice electrochemical utilization is less than a 100 % with a mass fraction of the electrode remaining electrochemically inactive and only adding up to electrode weight and thus decreasing the energy density per unit mass. Maximum electrochemical utilization is achieved by careful electrode design; thin electrode films favour electrochemical utilization though other electrode design aspects have to be considered. The capacitance of a supercapacitor is calculated using two electroanalytical methods: cyclic voltammetry and galvanostatic charge-discharge tests. In a typical cyclic voltammetry experiment, a three-electrode configuration is used: the working electrode which is the electrode under study, a reference electrode such as saturated calomel, and a counter electrode such as platinum. The working electrode potential is ramped linearly from a given potential V1 27 up to a set potential V2 , and then back to V1 . This process can be repeated over several cycles [16] . The current at the working electrode is plotted versus potential to give a voltammogram trace. The specific capacitance Cs is then calculated as: Cs = Ia + |Ic | 2ms (2.20) where Ia and Ic are the anodic and cathodic currents, s is the voltage sweeping rate dV/dt, and m is the mass of the electrode. Alternatively: Cs = qa + |qc | 2m∆V (2.21) where qa and qc are the anodic and cathodic charges, m is defined as before, and ∆V = V2 - V1 is the working electrochemical window. In most cases, the stored charge is more accurately calculated as the integral of the Rt R0 current over time in a cyclic voltammogam with qa = 0 Ia dt and qc = t Ic dt where t is the scanning time from V1 to V2 . By means of galvanostatic charge discharge tests at constant current density, in a three electrode configuration, the specific capacitance Cs can be calculated as: Cs = Itd m∆V (2.22) where I is the current, td is the discharge time, ∆V = V2 - V1 where V2 is the potential at the beginning of discharge and V1 is the potential at the end of discharge - often these potential limits are the same as in cyclic 28 voltammetry -, and m is defined as before. In evaluating the capacitance of a supercapacitor electrode, tests must be done at current densities at least up to 300 mA cm−2 since measurements at lower current densities may lead to over-optimistic results. Another parameter to take into account while considering the design and performance of a supercapacitor is the efficiency with which energy applied is effectively stored and released. Losses of a system include unproductive reactions, short circuit of the system, or in the case of supercapacitors, hydrogen and oxygen evolution. The Coulombic efficiency η of a supercapacitor must be optimized: η = qd /qc (2.23) where qd and qc are the total amount of discharge and charge of the device tested at voltage V in a galvanostatic experiment. 29 2.3 2.3.1 Materials for Supercapacitors Carbon-based materials In order to enhance capacitance by charging a double layer, high surface area and electronically conductive materials are desirable. Graphitic carbon in its different forms is by far the most widely used supercapacitor electrode material. Many forms of engineered carbons have been used as electrodes for supercapacitors including activated carbons, carbon fibers, and carbon aerogels. Attractive characteristics of these forms of carbons are: high conductivity, high surface-area range (up to 3000 m2 g−1 for activated carbons), good corrosion resistance, high temperature stability, controlled pore structure to some extent, processability and compatibility in composite materials, well established methods of activation, relatively low cost, and friendly environmental nature [12] . Other forms of carbon with an improved pore-size control are templated carbons and carbide-derived carbons. Carbon nanostructures such as carbon nanotubes, nanohorns, carbon onions and more recently graphene have emerged as alternative materials for EDLC electrodes [6, 17, 18] . Here, activated carbon is reviewed in order to provide an insight into basic concepts about processing and study of carbons, and due to the scope of this work, attention is then concentrated on carbon nanotubes and graphene. The reader is directed to excellent reviews about other forms of carbon and their application in supercapacitors [6, 9, 12, 19–21] . 30 Activated carbon Graphite is an allotrope of carbon with a sp2 bond configuration. Most current commercial carbons consist of amorphous carbon with a more or less disordered microstructure based on that of graphite [12] . Amorphous carbons are sections of hexagonal layers randomly distributed. The process of graphitisation consists of the ordering and stacking of theses layers and it is achieved by thermal treatment at high temperatures ( > 2,500 ◦ C). The characteristics of carbon can be tailored according to different parameters during synthesis. Most carbons are derived from carbon-rich precursors by carbonization (heat treatment in an inert atmosphere). Thermal decomposition of precursors during carbonization eliminates volatile components. A subsequent temperature increase causes condensation reactions, graphitic units (microcrystallites) start to grow and to align to form stacks of graphenelike sheets. The carbon precursors and processing conditions determine the size and number of formed graphene sheets and the orientation of crystallites which determines the texture and electrical conductivity of the synthesized carbon. Graphitised carbons are characterized by a highly ordered graphitic structure and are produced from precursors which pass through a fluid stage allowing for alignment of large aromatic structures. Non-graphitised carbons that consist of rigid amorphous structure of randomly oriented graphene layers are synthesized from precursors that retain a solid phase during carbonization: biomass (wood, nut shells), non-fusing coals, and thermosetting polymers such as polyvinylidene chloride (PVDC) [22] . The process of increasing the SSA and pore volume of carbonized carbon 31 is given the generic name of “activation”. The carbon form thus obtained receives the generic name of “activated carbon”. Thermal activation consists of controlled gasification of carbonized carbon at temperatures between 700 ◦ C to 1100 ◦ C in the presence of oxidising gases such as steam, carbon dioxide, air, or mixtures of these gases. The oxidising environment opens or etches pores and increases the SSA. Chemical activation involves the dehydration action of chemical agents such as phosphoric acid, zinc chloride, and potassium hydroxide at temperatures from 400-700 ◦ C. A post-washing procedure eliminates activating reactant residues and inorganic species coming from precursors [23–25] . The surface areas of porous carbons and electrodes are most commonly measured by gas adsorption (usually nitrogen at 77 K) and use BET (BrunauerEmmett-Teller) theory to convert adsorption data into an estimate of apparent surface area. Alternative techniques use sorption of other gases such as Ar and CO2 [26] . The surface area of porous carbons arises from a interconnected network of pores. The IUPAC classifies pores according to size into three classes: micropores ( < 2 nm), mesopores (2-50 nm) and macropores (> 50 nm). In most activated carbons, the pore size distribution is not optimum because of poor size control in the activation process. Activated carbons can exhibit high surface areas ranging from 500 to 3,000 m2 g−1 [12] . However, because of the poor control size, the usable surface area is in the range 1,000-2,000 m2 g−1 [19] . The capacitance exhibited by activated carbon strongly depends on morphology, electrical conductivity, degree and type of porosity, electrolyte used, and the presence of surface functionalities. The capacitance of activated carbons ranges from 20- 160 F g−1 [9] . 32 Pores of different dimensions in carbon-based materials contribute differently to total surface area and to double layer capacitance. Macropores make a negligible contribution to the surface area and act mainly as transport vias for electrolyte ions to get to the interior of carbon particles [12] . Mesopores contribute to the surface area and provide wide transport vias for diffusion. Micropores have a high surface area to volume ratio and thus, are a major contributor to measured surface area. Carbon Nanotubes Carbon nanotubes (CNTs), including single walled carbon nanotubes (SWNTs) and multi-walled carbon nanotubes (MWNTs), have been applied as electrode materials for electrochemical energy storage systems such as batteries, fuel cells, and EDLCs [27] . Among properties that make CNTs suitable for energy storage systems are their hollow cylindrical nanostructure with chirality-dependent semiconductor to conductor properties [28, 29] , large surface area, porosity of nanometer size, corrosion resistance, and high temperature stability [27, 30] . The specific capacitance of CNTs depends on their morphology (porosity, dimensions, surface area), and purity which in turn depends on the fabrication process and further processing [31] . The capacitance of purified (free from other forms of carbon an catalyst particles) CNTs varies from 15-80 F g−1 , and their surface area from 120 to 400 m2 g−1 [32] . The surface area of as produced CNT bundles is considered to be mainly mesoporous due to voids associated with the CNTs tangling nature and nanometer sized inner diameters. The surface area of CNTs is increased upon chemical activation, which is 33 commonly done by heating in KOH under an argon atmosphere [33, 34] . Upon activation, the degree of wall defects is increased and length shortening occurs giving raise to microporosity. As in the case of other carbon forms, there is no consensus about the pore size that optimizes capacitance. Frackowiak et al. doubled the surface area of MWNTs by KOH activation obtaining a maximum of 1050 m2 g−1 . The capacitance increased from 10-15 F g−1 to 90 F g−1 in alkaline media and to 95 F g−1 in acidic media, which was attributed to an increase in microporosity, whereas mesoporosity was thought to play an important role providing adsorption sites and ion transportation channels. In order to improve the capacitance of CNT based electrodes, the electric double layer storage mechanism can be combined with a pseudocapacitive effect. CNTs are chemically modified to introduce surface functionalities which at the same time modify the texture and morphology. A common practice is to introduce carboxylic groups on the CNTs surface by chemical treatment with HNO3 [35–37] . Besides introducing a pseudocapacitive effect, carboxylic groups increase the hydrophilicity of CNTs enhancing wetability in aqueous solvents which in turn favours double layer formation. Acid treatments shorten the length and open the tips of CNTs, and introduce wall defects that result in an enhanced surface area [38–40] . Zhao et al. carboxyl-functionalized CNTs to increase specific capacitance from 38 F g−1 to 155 F g−1 for SWNTs and from 23 F g−1 to 77 F g−1 for MWNTs [37] . A purification process before CNTs are functionalized is key to achieve an efficient functionalization. An efficient method to purify CNTs was demonstrated by Tobias et al. where steam is used as a mild oxidizing agent to remove amorphous carbon and 34 graphitic shells encapsulating catalyst particles [38, 39] . Catalytic particles are subsequently removed by washing with an acid solution. The steam treatment in itself increases surface area and amount of available reactive sites by opening the tips of CNTs. A non-covalent approach to increase capacitance of CNT based electrodes is the manufacture of composite electrodes containing CNTs, conductive polymers and/or other redox active nanomaterials [41] . Can carbon nanotubes replace carbon in supercapacitor applications? In general, untreated CNTs or nanofibers have lower capacitance than activated carbons in organic electrolytes [35] . The specific capacitance of CNTs in aqueous electrolytes can be increased by covalent functionalization [37] . Acid treatment of MWNTs to introduce carboxylic groups increased electron transfer kinetics, which exceeds that of carbons, and gave a lower ESR [42] . However, covalent functionalization was detrimental to cyclability [12] . The carbons used in EDLCs are pretreated to remove moisture and surface functionalities to improve stability during cycling. Specifically, oxygen functionalities in carbon increased ESR and thus, deteriorated capacitance [12] . The conductivity of SWNTs varies according to chirality. Current methods to synthesize SWNTs generally obtain a mixture of metallic, semiconductive and insulating SWNTs, and there is no established method to synthesize or separate only conductive SWNTs. Although better performance in EDLCs was reported for SWNTs than for MWNTs, the manufacturing cost of SWNTs is higher and their processing tends to be more difficult due 35 to increased hydrophobicity [37] . The entangled predominantly mesoporous nature of CNTs is considered to favour better accessibility of surface area for double layer formation as compared with activated carbons with coexisting “unaccessible ” microporosity [30, 35] which leads to a reduces ESR and a high power density > 8 kW kg−1 . The entangled CNT network provides mechanical stability allowing the manufacture of binder-free CNT electrodes which is not possible in the case of carbons with a characteristic mechanically unstable particulate microstructure. CNT electrodes have been manufactured as free standing mats, bucky papers, or directly grown CNTs on current collectors avoiding an increase of ESR imposed by electrochemically inactive binders (which also reduce accessible area of electrochemically active materials to electrolytes) [43–45] . CNT networks have been used as 3D scaffolds to support redox nanostructured materials increasing greatly the electrochemical utilization [46, 47] . Without doubt, current manufacturing cost of CNTs when compared with the various carbon forms is the main factor undermining the use of CNTs in EDLCs. Other significant barriers include perceived but uncertain health and safety issues surrounding nano-scale size materials. Graphene Graphene consists of a one-atom thick sp2 -bonded carbon sheet forming a honeycomb two dimensional (2D) nanostructure with unique electronic properties: ambipolar electric field effect with high charge carrier (massless Dirac fermions) mobilities independent of temperature, and a room temperature 36 quantum Hall effect [48–52] . These electronic properties and a theoretical surface area of 2,600 m2 g −1 suggest that graphene may be a promising candidate material for supercapacitor applications. However the realization of this potential in practice demands cost-effective and scaleable synthesis methods that preserve the key properties. Major challenges include: (1) obtaining defect-free graphene that preserves its intrinsic high electrical conductivity, (2) synthesis of high surface area graphene with a high yield of mono- and bi-layer flakes, and (3) inhibition of graphene’s natural tendency to restack. Current methods to produce graphene include pyrolysis of suitable precursors, arc evaporation of SiC, thermal exfoliation of graphitic oxide at high temperatures (1050 ◦ C), chemical vapor deposition (CVD), and chemical reduction of graphene oxide using different reducing agents, typically hydrazine either in aqueous solution or as gas [53–58] . The latter is a popular method that produces “graphene” that contains a high fraction of residual oxides and structural defects that undermine electronic conductivity [55, 59, 60] . Processing times are reduced to minutes by microwave assisted methods that reduce graphene oxide either directly or by making use of polar solvents (solvothermal methods) [61, 62] . The material obtained by this method has an improved electrical conductivity (274 S m−1 ) over chemically reduced graphene oxide (≈ 100 S m−1 ) [62] . Exfoliation of graphite in aqueous or organic solvents produces a purer graphene with higher electrical conductivity and a lower degree of defects than graphene-oxide derived graphene [63–68] . Exfoliation in liquid-phase using organic solvents such as N-methyl pyrrolidone has proved to be effective producing < 5 layer graphene flakes with conductivity up to 18,000 S m−1 [64] . Drawbacks of this method include the use of toxic, envi37 ronmentally unfriendly, and high boiling points solvents which make difficult processability and scaleability. Alternatively, graphite has been exfoliated in aqueous media making use of surfactants such as sodium dodecylbenzene sulfonate (SDBS) and sodium cholate (NaC), achieving yields up to 0.3 mg ml−1 where the presence of surfactant, however, lowers the electrical conductivity [65, 66] . A general practice to remove the surfactant consists of annealing in inert atmospheres such Ar/N2 achieving higher electrical conductivities, e.g. 1,500 S m−1 [66] . CVD methods produced vertically aligned graphene with a high degree of purity, and high surface area, although the main concern with this method is the reproducibility and high manufacturing associated cost [58] . The capacitances for graphene electrodes reported to date vary greatly according to synthesis method, degree of functionalization, degree of structural defects, texture, and orientation of the graphene sheets respect to the current collector [18, 58, 69] . The capacitances reported for graphene oxide-derived graphene (GOG) are in the range 100-200 F g−1 [53–57] . Because lower capacitances have been obtained for purer graphene with higher surface area such as graphene produced by CVD methods [58] , it is likely that a large percentage of the capacitance reported for GOG comes actually from residual oxygen functionalities of impure “graphene”. The degree of mesoporosity and microporosity of otherwise physically integral graphene sheets greatly affects supercapacitor electrode performance. Recent findings indicate that microporosity of graphene sheets favours an enhanced energy density by increasing the surface area available for charge storage but undermines power density performance because of the inherently 38 increased ionic resistance. The largest energy density (70 Wh kg−1 in ionic liquid and 0-3.5 V electrochemical window) reported to date has been for chemically “activated graphene” with a high degree of porosity in the range 0.6 to 5 nm, and a surface area of 3,100 m2 g−1 [70] , whereas CVD grown graphene with no sign of porous behaviour (as determined by electrochemical impedance spectroscopy) showed a high power density performance but a low capacitance [58] . In the former case, the graphene sheets are in an in-plane orientation with respect to the current collector whereas in the latter case, graphene sheets are grown vertically onto the current collector. Vertically oriented graphene sheets allow a better exposure of charge storage surfaces to the electrolyte favouring a fast ionic transport and thus enhancing power density. However, the amount of charge storage will greatly depend on the number of graphene sheets per unit area which might be the major area of improvement for CVD manufacturing methods in order to increase the energy density. In the case of chemically activated graphene, the high energy density was attributed to the 3D microporous texture providing with high surface area. Overall, these findings seem to indicate that pure non-defective graphene is not suitable to provide a high energy density, being however ideal for high power density applications. A further insight into the performance of graphene for supercapacitor applications is addressed in Chapter 7. In order to prevent restacking of graphene sheets in electrodes where graphene lies in an in-plane geometry with respect to the current collector, several strategies have been followed including the intercalation of metal and metal oxide nanoparticles. This method has been an approach followed not only for supercapacitors but also for battery applications in which case Li 39 could be stored on both sides of the graphene sheet doubling the theoretical capacity of the graphite of ≈ 372 mAh g−1 [71] . Platinum nanoparticles have been used as spacer in between graphene sheets to avoid aggregation in a Pt nanoparticle-graphene composite with a increased surface area of 862 m2 g−1 as compared to dried graphene with a surface area of 44 m2 g−1 [72] . Graphene sheets have been re-assembled in the presence of carbon nanotubes and fullerenes allowing for tuning of the interlayer space of graphene sheets [73] . An increase of capacity storage of Li ions was correlated with an increase of interlayer spacing controlled by the intercalating carbon species. The highest capacity of 600 mAh g−1 (versus 290 mAh g−1 of graphene) was found for C60 intercalated graphene layers with a graphene interlayer spacing of 0.4 nm. Interestingly, the graphene interlaying spacing increased as the number of stacking layers was reduced. Graphene has been combined with Sn and SnO2 nanoparticles in 3D structures where intercalation of nanoparticles into graphene layers helped to prevent graphene aggregation while graphene layers prevented aggregation of SnO2 nanoparticles, and electrode volume expansion during cycling preventing capacity fading [71, 74, 75] . One of the main attractive features of graphene is that it offers the possibility of manufacturing free standing flexible electrodes [76, 77] . Graphene papers have a high tensile modulus of 35 to 41.8 GPa, and a room temperature electrical conductivity of 7,200 S m−1 [78, 79] . Graphene paper has served as substrate of pseudocapacitive materials resulting in a enhanced energy density of the construct. Polyaniline has been anodically electrodeposited into free standing graphene papers improving capacitance from 147 F g−1 in 40 graphene paper to 233 F g−1 in the polyaniline/graphene paper composite [80] . In situ polymerization of aniline monomers on a graphene suspension produced a composite with high capacitance of 1,035 F g−1 in KOH although the cyclability performance was poor with 53 % capacitance fading after 1,000 cycles [81, 82] . A summary of the capacitance performance of nanocomposites of graphene with pseudocapacitive materials is provided in reference [18] . 2.3.2 Redox-based electrochemical capacitors Materials with redox activity are used to enhance capacitance of EDLCs via a charge transfer pseudocapacitive effect. Transition metal oxides are considered the best candidates as pseudocapacitive materials because they often have a variety of oxidation states available for redox charge transfer. However, most of them suffer from poor electrical conductivity, and redox or intercalation/deintercalation pseudocapacitive processes can lead to large mechanical strains that undermine cyclability. Ruthenium oxide Among the transition metal oxides, the most investigated in the last decades is RuO2 due to excellent capacitive and power performance which is attributed to high electronic conductivity, availability of three oxidation states (II-IV), and the possibility to append proton conductivity. The presence of structural water in amorphous RuO2 ·xH2 O provides pathways for proton conduction. The pseudocapacitance is maximized when both proton and electron conduction are optimized to provide a “double injection” of protons 41 and electrons in the hydrous RuO2 [6, 83] : RuOx (OH)y + δH + + δe− ⇀ ↽ RuOx−δ (OH)y+δ (2.24) The electronic conductivity of RuO2 increases with ordering of the crystal structure. While the proton conductivity increases with the content of structural water, the crystal ordering is reduced, ultimately leading to an amorphous structure. Therefore, synthesizing methods must optimize the water content of RuO2 ·xH2 O in order to balance electron and proton transport [84] . RuO2 supercapacitors use acidic aqueous electrolytes, commonly sulfuric acid as a source of protons for redox activity within an electrochemical window of 0-1 V. A maximum capacitance of 720-900 F g−1 has been reported for RuO2 ·0.5H2 O [84] , generally synthesized by a sol-gel method where a hydrothermal treatment (annealing in water) restricts condensation of hydroxyl groups and restricts crystallization maintaining a high water content of RuO2 ·xH2 O nanoparticles and thus enhancing capacitance [85] . Sugimoto et al. synthesized a layered ruthenic acid where crystalline RuO2 is interleaved with layers of water and thus behaves as a proton-electron conductor achieving a capacitance of 390 F g−1 [83] . High capacitances and power performance have been achieved by anodic deposition of RuO2 ·xH2 O nanotubes on graphite or Ti substrates [86] . Upon annealing in air, the capacitance was as high as 1300 F g−1 and the power density and energy density at 0.8 V and 4 kHz were 4320 kW kg−1 and 7.5 Wh kg−1 respectively. The nanotubular structure provided paths for elec- 42 trolyte penetration facilitating proton transport whereas electron transport was promoted by uniformly distributed thin walls of RuO2 ·xH2 O. A similar performance was reported for mesoporous crystalline RuO2 which was attributed to a simultaneous enhancement of proton and electron transport in the mesoporous structure giving a power density of 2600 kW kg−1 at 4.6 Wh kg−1 energy density [87] . Although the specific capacitance value was low, 84 F g−1 , there was only a 47 % capacitane loss for a very high potential sweep rate of 10,000 mV s−1 . Despite the excellent performance of RuO2 ·xH2 O in various nanostructured forms, the high manufacturing and material cost prevents its use for technological applications. Attempts to lower the cost of ruthenium oxide based EDLCs includes the addition of ternary solid solutions, e.g. a capacitance of 249.6 F g−1 was achieved for amorphous 23 % w Ru1−y Cry O2 loaded in TiO2 tubes [88] . Similarly, composites of RuO2 ·xH2 O with carbon nanofibers, carbon black, and carbon nanotubes have been synthesized in order to reduce cost and increase conductivity [89–93] . Manganese oxide Alternative lower cost oxides for EDLC applications include MnO2 , NiO, Fe3 O4 and V2 O5 . There has been particular interest in MnO2 because of its low cost, low toxicity and high electrochemical activity. The pseudocapacitive behavior of MnO2 involves three different mechanisms [94–99] : 1. A surface adsorption of electrolyte cations (C+ = Na+ , K+ ,Li+ ): 43 (M nO2 )surf ace + C + + e− ⇀ ↽ (M nO2− C + )surf ace (2.25) 2. A reduction process to MnOOH: M nO2 + H + + e− ⇀ ↽ M nOOH (2.26) 3. An intercalation-deintercalation of cations: M nO2 + C + + e− ⇀ ↽ M nOOC (2.27) The electrochemical performance of MnO2 as a pseudocapacitive material is determined by several variables including particle size, crystal structure which is in turn determined by the synthesis method (sol-gel routes, electrodeposition, and sputter deposition), and post-synthetic heat treatment used to produce the oxide [100, 101] . The specific capacitance and cyclability of MnO2 electrodes is strongly influenced by morphology, microstructure, and electrode thickness. A common practice to manufacture a supercapacitor electrode is to mix a high surface area powder of electrochemically active material, in this case MnO2 , with a form of carbon (carbon black commonly as conductive additive) and a polymer “binder” (poly-(vinylidene fluoride)) to obtain a paste which is deposited onto a current collector by rolling and pressing. MnO2 electrodes produced by this method have a micrometric thickness delivering a specific capacitance of 150-250 F g−1 [102, 103] . There was a sharp increase in 44 capacitance (700-1380 F g−1 ) for nanometer-thick electrodes electrodeposited directly onto current collectors [100] . Nanometer-thick electrodes enable maximum electrochemical utilization and short diffusion lengths for electrolyte ions increasing the specific capacitance per unit mass [94, 96, 104] . Also, in thin electrodes, a higher percentage of the electrochemically active material is in direct contact with current collectors minimizing the overall ESR. The idea of deposition of nanometer-thick layers of pseudocapacitive materials directly onto current collectors in order to increase specific capacitance and thus energy density has been extended to 3D nanostructures. Pseudocapacitive materials are deposited onto highly conductive nanostructured materials such as carbon and carbon nanotubes that serve as current collectors as shown in Figure 2.7a and Figure 2.7b. In the case of MnO2 , thin layers have been deposited onto carbon nanotubes, templated mesoporous carbon, carbon aerogels and carbon foams [103, 105–110] . Fan et al. synthesized an electrode where γ-MnO2 was dispersed on an array of vertically aligned CNTs by an electrochemically induced deposition method as shown in Figure 2.7c and Figure 2.7d [105] . The specific capacitance of the composite, normalized to MnO2 content, was 784 F g−1 . This value is still far from the theoretical MnO2 capacitance of 1,110 F g−1 . The cyclability of this electrode was reported only up to 800 cycles at 1 mA cm−2 when the capacitance had decreased only by 5 %. Other attempts to increase capacitance and lower ESR of the electrode include the synthesis of composites with conductive and redox active polymers such as polypyrrole, polyaniline, and polythiophene. The capacitance of the composites is higher than that of the components due to enhanced 45 Figure 2.7: Deposition of pseudocapacitive materials onto high SSA, conductive, 3D-nanostructured materials. Schematics of (a) carbon nanotubes or rods grown onto current collectors, and (b) pseudocapacitive material deposited onto carbon nanotubes, SEM images of (c) carbon nanotubes directly grown on a graphite substrate, and (d) γ-MnO2 dispersed onto carbon nanotubes. Reproduced from [6] and [105] . conductivity and additive redox activity of the polymer [111] . However, major drawbacks of polymers as supercapacitor materials include: (1) although the polymeric matrix is expected to enhance SSA of the electrode, often the polymer blocks surface area available for redox activity of the oxide (or any other pseudocapacitive component of the composite), and (2) polymers have limited stability during cycling [6] . The working electrochemical window of MnO2 is limited having a maximum achievable energy density according to Equation 2.12. The upper cutoff voltage is limited by oxygen evolution and the lower cut-off voltage is determined by the onset of the Mn+4 irreversible reduction and subsequent manganese dissolution. In order to expand the limited voltage window, supercapacitors are built using a MnO2 electrode combined with an electrode 46 whose electrochemical window is complementary to that of MnO2 . These devices are called asymmetric or hybrid supercapacitors and are dealt with in section 2.4. Iron oxide Iron oxides are an alternative even lower cost materials recently investigated for supercapacitors. Magnetite (Fe3 O4 ) exhibits redox-type pseudocapacitance with alkali sulfite and sulfate electrolytes [112] . Hematite (α-Fe2 O3 ) and MnFe2 O4 show Li ion intercalation/deintercalation in lithium based electrolytes [113, 114] . The specific capacitance reported for iron oxides electrodes is inferior to that of manganese oxides, typically less than 100 F g−1 in aqueous and organic electrolytes. The main drawback of iron oxides is the low semi-conductivity of hematite (10−9 Ω−1 cm−1 ), and maghemite (10−9 Ω−1 cm−1 ), although magnetite (102 -103 Ω−1 cm−1 ) display an almost metallic conductivity. For this reason, in order to increase capacitance and reduce ESR when using these materials for supercapacitor applications, composites with carbon and carbon nanotubes have been synthesized [113, 115] . The cyclability reported for iron oxides ranges from 300 to 500 cycles with 18 to 50 % capacitance loss [113, 115] . Vanadium oxide Vanadium oxides are an alternative material for supercapacitor applications due a relative low cost as compared to ruthenium oxide, availability of oxidation states (II-V) for pseudocapacitance activity, and a layered crystal structure in case of V2 O5 suitable for ion intercalation [116, 117] . Crystalline 47 V2 O5 and the hydrated compound V2 O5 ·xH2 O have been extensively studied for Li-ion intercalation materials for application as cathodes in Li-ion batteries and more recently for supercapacitor applications both in aqueous and organic electrolytes. The main limitations for ion intercalation is the moderate conductivity of V2 O5 (10−2 -10−3 Ω−1 m−1 ), and low diffusion coefficient of ions into V2 O5 , 10−12 - 10−13 cm2 s−1 for lithium. In order to overcome these limitations, thin layers of vanadium oxide have been deposited onto conductive, high surface area nanostructured materials such as carbon nanotubes, and graphite enhancing electrochemical utilization and minimizing diffusion lengths achieving capacitances up to 1,230 F g−1 for V2 O5 ·xH2 O [117–120] . Another key aspect to enhance pseudocapacitance is the content of water on vanadium oxides. The interlayer distance of V2 O5 crystals was increased from 4.56 Å to 11.52 Å by water intercalation in the compound V2 O5 ·xH2 O facilitating ion insertion and providing paths for proton exchange when tested in aqueous electrolytes [116] . Huang et al. showed how by annealing of an electrodeposited mix of V2 O5 , V2 O5 ·1.6H2 O, and V6 O13 onto graphite, the content of V2 O5 ·1.6H2 O can be tuned to optimize K+ ion and proton exchange, and thus pseudocapacitance [117] . The maximum capacitance obtained was 150-250 F g−1 at a scan rate of 250 mV s−1 in a electrochemical window from -0.2 to 0.8 V. Nanostructuring of vanadium oxides in form of nanorods, nanotubes, nanocables, and nanowires is another approach to enhance electrochemical utilization by increasing surface area of electrochemically active material, shortening of diffusion length distances and reduction of internal electrical resistance [116, 121] . Nanorods arrays are typically grown via electrodeposi48 tion methods into a template directing material such as porous polycarbonate membranes. V2 O5 nanorods, V2 O5 ·xH2 O nanotubes, and core-shell Ni/V2 O5 ·xH2 O grown with this method have achieved an enhanced Li-ion storing capacity when compared with sol-gel films [116] . Unfortunately templating methods are expensive and the scaleability for the manufacture of large area electrodes is unlikely. In another approach, vanadium pentoxide nanowires have been synthesized by a hydrothermal method and combined with carbon nanotubes in a composite forming a interpenetrating network structure that enhances surface area and electrical conductivity. The specific capacitance achieved was 440 F g−1 in aqueous electrolyte Na2 SO4 at a charge-discharge current density of 0.25 A g−1 . The capacitance retention of the composite at 10 A g−1 was improved by 30 % with respect to V2 O5 nanowires only [122] . Alternatively, V6 O13 has been investigated as a mixed valence (IV an V) oxide for supercapacitor applications [123] . A specific capacitance of 215 F g−1 at 0.2 A g−1 was achieved with 96 % retention after 300 cycles. Vanadium nitride Vanadium nitride nanocrystallites (VN) delivered an impressive specific capacitance comparable with that of RuO2 of 1,340 F g−1 at 2 mV s−1 which was reduced to 554 F g−1 at 100 mV s−1 [124] . The observed high capacitance was attributed to the formation of thin oxide layers onto highly conductive VN nanocrystals ( 1.67 x 106 Ω−1 m−1 ), and to the fact that vanadium oxide thin layers undergo a cascade of redox reactions passing through II, III, and IV oxidation states upon adsorption of OH− groups from the electrolyte 49 KOH. At a scan rate of 50 mV s−1 the capacitance decreased to 400 F g−1 being degraded by 10 % after 1000 cycles. Nickel Oxide Nickel oxides are another low cost material with potential use as electrode in supercapacitors. However, NiO is a poorly conductive material (10−4 Ω−1 cm−1 ) with a small electrochemical window from -0.2 V to 0.5 V. The pseudocapacitive activity of NiO and Ni(OH)2 is based on the redox reactions: N i(OH)2 + OH − ⇀ ↽ N iOOH + H2 O + e− (2.28) N iO + OH − ⇀ ↽ N iOOH + e− (2.29) NiO and Ni(OH)2 films have been electrodeposited as nanoporous structures in order to enhance specific capacitance [125–129] . The specific capacitance was below 200 F g−1 for most cases and the cyclability performance was in general poor around 400-500 cycles. Zhao et al. electrodeposited a hexagonal nanoporous Ni(OH)2 film giving a maximum capacitance of 578 F g−1 with a poor cyclability with 4.5 % loss of capacitance after 400 cycles [126] . A better performance has been achieved when nickel oxides are combined with cobalt oxides deposited on conductive CNTs or carbon [130–134] . Fan et al. deposited nickel and cobalt oxides on CNTs by thermal decomposition. The specific capacitance was 569 F g−1 with 3.6 % loss after 2,000 cycles. Hu et al. synthesized mesoporous Cox Ni1−x layered double hydroxides using a chemical co-precipitation route achieving a capacitance of 1,809 F g−1 at 1 A 50 g−1 with 9.8 % loss after 1000 cycles at 10 A g−1 . The good performace was attributed to the interlayered nanostructure with content of water molecules in a thin-walled (10 nm) cell configuration that provided a mesoporous structure accesible to electrolyte [135] . In order to enlarge the electrochemical window (up to ≈ 1.8 V), nickel oxides have been combined with manganese oxides [136, 137] . Attempts to produce 3D nanostructures of NiO include nanospherical structures (730 F g−1 , 200 cycles with 2 % capacitance loss) and 3D NiO/silicon microchanel plates (586.4 F g−1 , 400 cycles with 4.8 % capacitance loss) [138, 139] . Other oxides Due to its low cost, layered orthorhombic crystal structure, and a broad range of valence states (+2 to +7), α-MoO3 is an attractive material for supercapacitor applications offering both redox and ion intercalation pseudocapacitive properties. α-MoO3 has been extensively studied as an electrode in Li-ion batteries, and only recently its application to supercapacitors has been studied both in organic and aqueous electrolytes. An overview of previous work and advantages and disadvantages of this material is covered in Chapters 5 and 6. Other materials largely explored for application in supercapacitors are conductive p- and n- dopable polymers such as poly(3-arylthiopene), p-doped poly(pyrrole), poly(3-methylthiophene), or poly(1,5-diaminoanthraquinone). These materials have been tested for supercapacitor applications and have shown high gravimetric and volumetric capacitance in non-aqueous electrolytes. However, swelling and shrinking upon cycling is a well-known detri51 mental characteristic [140] . In summary, the performance of several metal oxides and carbons for supercapacitor applications has been reviewed. Figure 2.8 shows a comparative graph of pseudocapacitive materials as presented by Naoi K. et al. [141] where other data points have been added to account for some of the materials here reviewed. There is no ideal metal oxide for supercapacitor applications. The use of best performance noble metal oxides is prevented by high cost issues. Other metal oxides present problems of low conductivity, narrow working electrochemical windows or limited cyclability. Although, nanostructured design and composite manufacturing approaches have helped to overcome some of these problems, the high cost of some synthesis procedures prevents widespread application and scaleability. Efforts must be directed to investigate cheap and scaleable synthesis procedures using low cost materials. Figure 2.8: Comparative graph of pseudocapacitive materials. Reproduced from reference [141] with some data points added for graphene, V2 O5 ·xH2 O, and MoO3 . 52 2.4 Hybrid supercapacitors In order to achieve high energy density, a battery-like electrode (energy source) is combined with an EDLC electrode (power source) in an asymmetric hybrid system configuration. According to Equation 2.12, the energy density is proportional to capacitance and the square of the voltage. The battery-like electrode introduces pseudocapacitive processes which enhance capacitance and therefore energy density. The operating voltage, which depends on the electrochemical stability of the electrolyte, can be increased upon combination of an appropriate hybrid system with an appropriate electrolyte [142] . Common hybrid systems are: (1) pseudo-capacitive metal oxides with a capacitive carbon electrode, and (2) lithium-insertion electrodes with a carbon electrode [6] . A hybrid supercapacitor configuration is shown in Figure 2.9. 2.4.1 Electrolyte considerations An important part in the design of hybrid supercapacitors is the choice of an appropriate electrolyte [143] . Most hybrid systems use either an aqueous or an organic electrolyte. Aqueous electrolytes can work in a voltage window of maximum 1 V to avoid the regime that leads to electrolysis (the thermodynamic window of water is 1.23 V), and have a high ion mobility and a relatively low ESR. Neutral aqueous electrolytes are preferred to acidic ones for technological applications due to their more environmentally friendly nature. Organic electrolytes decompose at voltages ranging from 3 to 5 V but usually have a low ionic conductivity (usually Li ion), and high ESR that prevents 53 Figure 2.9: Schematic of a hybrid supercapacitor using carbon as negative electrode combined with a MnO2 positive electrode. Cyclic voltammograms are shown in insets in 0.1 M K2SO4 electrolyte. Reproduced from [100] . their use for high power applications. The use of pure organic electrolytes needs manufacture, assembly, and operation in water-free atmosphere which increases overall electrolyte and manufacturing cost. Another drawback of organic electrolytes is safety-related. Thermal runaway (a cascade of events leading to continuous increase of temperature) of the electrolyte leads to vaporization, fire, and explosion of the device. Thermal runaway problems are not uncommon in cells of supercapacitors. 54 2.4.2 Electrochemical window and charge balance The optimization of the performance of a hybrid supercapacitor involves the optimization of the working electrochemical window. The first developed hybrid systems consisted of an activated carbon (AC) electrode combined with a pseudocapacitive metal oxide electrode such as MnO2 , Ni(OH)2 or PbO2 in aqueous media. When designing a hybrid supercapacitor, it is desirable (as previously described) that the working voltages of negative and positive electrodes are as as widely spaced as possible. Limits are imposed by pH dependent irreversible processes such as hydrogen evolution at the negative electrode, oxygen evolution at the positive electrode, and dissolution of electrochemically active materials. The optimized electrochemical window of a hybrid supercapacitor is the sum of complementary working voltages of positive and negative electrodes where no irreversibilities occur. AC and metal oxides show complementary electrochemical windows leading to enhancement of the overall voltage of the resulting hybrid supercapacitor. AC shows a unique behaviour under negative polarisation with a high hydrogen overpotential, i.e. hydrogen evolution shifts to lower potentials than the nominal thermodynamic value. This behaviour is attributed to an increased pH (decreased pOH) on the electrode surface as the concentration of OH− anion increases due to confinement within the carbon micropores resulting on a slow kinetics of water reduction: 2H2 O + e− ⇀ ↽ H2 + 2OH − 55 (2.30) where the nascent H2 is also confined within the carbon micropores and electrooxidized during the subsequent anodic scan, obviated by an increase of anodic current, adding some pseudocapacitive charge storage to the usual double layer charge storage of the AC anode [142] . Likewise, metal oxides such as MnO2 show a high oxygen overpotential due to a lower pH value near the electrode surface under positive polarisation caused by H+ generated by redox activity of the oxide. Therefore AC and MnO2 are suitable as negative and positive electrodes respectively in a hybrid supercapacitor configuration. In order to get the optimum cell voltage in an asymmetric hybrid supercapacitor, the mass of the electrodes has to be balanced so that charge is equal in positive and negative electrodes, i.e. C1 ∆V1 m1 = C2 ∆V2 m2 where C is capacitance, ∆V is the electrochemical window, m is mass, and the subscripts 1 and 2 indicate negative and positive electrode respectively [142] . Figure 2.10a shows voltammograms of individual MnO2 and AC electrodes where the potential range before irreversible processes (observed in a voltammogram as deviations from a square shape) occur in each electrode is indicated. Figure Figure 2.10b shows voltammograms of the corresponding AC/MnO2 hybrid supercapacitor at a different working electrochemical windows where electrolyte pH and electrode mass have been optimized. The hybrid capacitor shows ideal capacitive behaviour up to 2.2 V. 56 Figure 2.10: Optimization of the working electrochemical window in a hybrid supercapacitor. (a) Comparative cyclic voltammograms in a three electrode cell configuration using activated carbon or MnO2 , (b) Cyclic voltammograms of the optimized AC/MnO2 supercapacitor in 2 M KNO3 at pH = 6.4, and mass ratio MnO2 /AC= 2.1. Reproduced from [142] . 2.4.3 Hybrid supercapacitors using redox positive electrodes A high energy density at high power density, minimum ESR and cycling stability (> 100,000 cycles) are requirements for real supercapacitor applications. Hybrid systems such as AC/MnO2 in aqueous electrolytes have a 0 - 2.2 electrochemical window with delivered performance of 21 Wh kg−1 57 at 123 kW kg−1 (MnO2 -CNT composite as positive electrode [142] ), 17 Wh kg−1 at 2 kW kg−1 with 6 % capacitance loss after 23,000 cycles (MnO2 nanorods as positive electrode [143] ). In order to improve the cyclability of AC/MnO2 , the voltage window can be reduced at the expense of reduction of energy and power density [144] . AC/Ni(OH)2 systems in KOH showed poor performance upon cycling and presented problems of electrolyte depletion (OH− in the electrolyte is consumed in redox processes in Equations 2.28 and 2.29) [145, 146] . Other systems such as AC/PbO2 delivered high power but weight and environmental issues prevented its application [141] . Other hybrid systems replaced AC as the negative electrode with redox active materials such as β-FeOOH nanorods which was then tested in combination with a MnO2 positive electrode in Li2 SO4 [147] . The performance, however, was not much better than the AC/MnO2 aqueous system, 12 Wh kg−1 at 3.7 kW kg−1 with a 15 % capacitance loss after 2,000 cycles. 2.4.4 Hybrid supercapacitors using Li-ion intercalation electrodes Various Li-ion intercalation anodes have been combined with AC cathodes including Li4 Ti5 O12 /AC, first developed by Amatucci et al. delivering an energy density greater than 10 Wh kg−1 for the first time where the titanate electrode ensured high power capacity and long-life cyclability [148] . Other Liion intercalation hybrid systems include AC/LiMn2 O4 in Li2 SO4 that overcame problems of electrolyte depletion and delivered an energy density of 35 Wh kg−1 at 100 W kg−1 , reducing to 10 Wh kg−1 at 2 kW kg−1 . The system 58 could be cycled up to 20,000 cycles with 5 % capacity loss [146] . More recently, phosphate-based materials have been used as negative electrodes due to high lithium-ion mobility, and good thermal properties. A LiTi2 (PO4 )3 /MnO2 system in Li2 SO4 delivered a capacity of 36 mAh g−1 and energy density of 43 Wh kg−1 at 200 W kg−1 with 20 % energy density loss after 1000 cycles [15] . Hybrid supercapacitors using pre-lithiated AC or polyacene as negative electrode and AC as positive electrode have achieved cell voltages of 3.8 V with energy densities up to 25 Wh kg−1 [141] . A main problem in Li-ion intercalation systems is degradation upon cyclability due to phase transformation and generated strain after repeated intercalation/deintercalation. The use of nanostructured materials such as nanoparticles and nanorods as Li-ion intercalation host has proved to improve cyclability. TiO2 -B nanorods have been used as Li-ion intercalation anodes showing superior capacity retention upon cycling compared to anatase and TiO2 -B nanoparticles of dimensions similar to the diameter of the nanowires [149] . Nanoscale-sized materials such as hematite nanoparticles showed enhanced Li-ion intercalation as compared to the micron-sized particles [149] . Zhao et al. developed a hybrid supercapacitor where α-Fe2 O3 100 nm particles are uniformly intercalated in a MWNTs mesoporous network achieving a energy density of 72 Wh kg−1 at a power density of 1 kW kg−1 . The high energy density was attributed to a decrease in internal resistance and an increase of ion-mobility through the MWNT interconnected network [113] . 59 2.5 Optimization of supercapacitor design In order to improve energy and power density of supercapacitors the resistance of supercapacitors must be minimized as well as the weight of inactive materials in electrode and cell packaging. The main contributors to ESR in a capacitor are the electrolyte resistance, current collector-electrode contact resistance and internal electrode resistance. Manufacturing methods of supercapacitor electrodes, and assembly of a supercapacitor cell must observe such key parameters. 2.5.1 Electrolytes The overall resistance of a supercapacitor is dependent on the ionic resistivity of the electrolyte, and electrolyte ion size, which directly influences diffusion proceses involved in ion transport into the electrochemically active material [150] . Aqueous electrolytes such as H2 SO4 and KOH have a lower resistivity (1-2 Ω cm) than organic electrolytes (20-60 Ω cm). Propylene carbonate and acetonitrile are organic solvents of salts such as LiClO4 providing the ions. The resistivity of propylene carbonate (52 Ω cm) is much higher than the resistivity of acetonitrile (18 Ω cm). Therefore, acetonitrile is preferred to achieve better performance but environmental and safety issues prevent its use due to its toxicity and flammability. A broad description of the properties of different electrolytes is provided in references [9, 19] . Ionic liquids are room temperature-solvent free electrolytes with stability over electrochemical windows of up to 4 V and temperatures up to 300 ◦ C with zero vapor pressure and no flammability [151–154] . However, the ionic conductivity 60 of ionic liquids is poor at room temperature and thus, they are mainly used at higher temperatures which reduces the usable electrochemical window from 4 V at 25 ◦ C to 3.25 V at 100 ◦ C. In practice, supercapacitors are mainly used at temperatures ranging from -30 ◦ C to +60 ◦ C. Ionic liquids are currently designed to improve conductivity at this temperature range by strategically combining cations and anions but more progress is still required. 2.5.2 Current collectors The function of a current collector in a supercapacitor is to serve as a substrate that transfers charge to and from the electrode. Therefore, it must be a high surface area conductive material providing with an efficient electrical contact with the electrochemically active material in order to minimize interfacial resistance and maximize utilization of the electrochemically active material. The resistance of current collectors is a major concern in high power devices due to the ohmic heating caused upon frequent charge-discharge cycles that can eventually lead to fire and failure of the device. In addition, the current collector must be inert in the electrolyte media, and have a minimum mass, thickness, and flexibility to allow for maximum energy density (energy per unit mass), and maximum volumetric efficiency in a variety of device designs suitable for a specific application. The current collector-electrode contact resistance of supercapacitors must be reduced down to at least 0.1 Ω cm−2 [4] . A robust electrical contact between electrochemically active materials and current collectors has been achieved by designing conductive nanostructures directly grown or deposited on current 61 collectors, for instance, carbon nanotubes directly grown on metallic substrates [155] where pseudocapacitive materials are then deposited as thin films to enhance energy density. Conductive carbonaceous powders have been deposited on Al foil current collectors via a sol-gel route providing better electrical contact between the current collector and a carbon electrode [156] . Moreover, carbon substrates have been used directly as current collectors, e.g., thin films of electrochemically active materials have been deposited on ultraporous carbon aerogels that serve as current collector [103] ; and flexible films of carbon nanotubes (bucky paper), and graphene serve as electrodes and current collectors at the same time [43–45, 157] . High surface area and light weight current collectors include stainless steel meshes, and 3D nickel foams where electrochemically active materials are deposited by rolling and pressing. The porosity of the foams facilitates impregnation and enhances utilization of the electrochemically active material [158, 159] . Miniaturization of devices such as electromechanical systems (MEMS) demands miniaturized power sources, microsupercapacitors are manufactured using diverse micro/nanofabrication techniques where 3D metallic current collectors have been fabricated using for instance UV lithography. The design of the current collectors consist of a interdigitized architecture with a higher surface area than their planar-design counterpart [160, 161] . However, high cost and high throughput production with a constant quality assurance are the main concern with these methods. 62 2.5.3 Electrode internal resistance The internal electrical resistance of an electrode in a supercapacitor should ideally be below 1 mΩ cm [4] . In order to improve the contact between electrochemically active material particles, conductive additives such as acetylene black are commonly added to the electrode. However this practice is detrimental to energy density and power density due to increased mass load of inactive material, and decrease of accessible surface area. Approaches to improve conductivity of electrochemically active materials without the use of conductive additives include 3D-nanostructure design. A 3D mesoporous array of nanotubes of metallic-conductive materials such as RuO2 ·H2 O provides quickly accessible pathways for fast electron hoping. This approach eliminates the problem of having a large fraction of electrochemically active material remaining inactive in the bulk of thick films and thus increasing ESR [86] . Mesoporosity and crystalline structure of electrochemically active materials are the key design parameters to promote electron as well as proton conductivity [87] . However, cost related issues must be observed when considering synthesis methods of sophisticated nanostructures. Although excellent performance can be obtained, application is generally far from reality. Research efforts must be focused on cheap synthesis methods procuring the use of low cost materials. Hydrothermal synthesis methods constitute an environmentally friendly and low cost method to synthesize a variety of nanostructures including nanowires and nanoparticles [162] . Insight into this type of synthesis methods constitutes a largely exploitable research area for the design of high 63 performance supercapacitors. 2.5.4 Time constant and electrode manufacturing methods In order to achieve high power densities meeting the needs for vehicle applications, the time constant t = Rs C, where Rs is the ESR, must be reduced to values less than 1 s [4] . Because Rs increases with electrode thickness, the development of processing techniques that allow for the manufacture of thin electrodes may become critical for some applications. Current electrode manufacturing techniques consist of mixing of electrochemically active materials with a conductive additive, and a fluoropolymer to maintain mechanical stability of the electrode, followed by application of high mechanical pressures of the mixture for deposition onto a current collector. Electrodes manufactured with this widely used technique tend to have a high ESR due to the addition of electrochemically inactive materials and poor control of electrode thickness which lies within the micrometric scale. Other methods such as rolling, slicing and dip-coating present similar problems. More recently, spray deposition techniques allow for accurate control of electrode thickness at the nanometric scale excluding the use of inert materials. In addition, a spray deposition technique can promote a homogeneous dispersion of electrochemically active materials in solution prior to deposition facilitating the manufacture of thin electrodes with high uniformity in thickness, morphology, and electrochemical properties. Because large areas can be spray-deposited, a spray deposition technology offers the possibility to scale-up the manufacture of current su- 64 percapacitor electrodes developed at a laboratory scale. Chapters 3 and 4 are dedicated to this electrode manufacturing technique. The commercialization of supercapacitors implies assembly of electrodes in cells. Manufacturing issues to be considered include to minimize production cost, ensure quality and uniformity of fabrication, minimize packaging weight (the presence of inactive weight is a main reason for low energy densities of supercapacitor cells), ensure long cycle and shelf life by minimizing leakage currents which are commonly caused by poor purity of electrode and electrolytes. 65 2.6 Supercapacitor applications Because supercapacitors are sources with high power delivery and long cycle life, they are used in situations of high power demands “buffering” the energy supply provided by batteries or fuels cells [140, 163, 164] . The main market for supercapacitors is the transportation sector with hybrid electric cars and trains, see Figure 2.11e. Hybrid electrical vehicles (HEV) are powered with petrol and use a battery module to improve efficiency (Prius, Toyota) or instead the vehicle almost completely runs on electrical power using petrol only to extend its range (plug-in hybrid Volt Chevrolet, General Motors) [165, 166] . When combined with batteries or fuel cells, EDLCs provide peak power when needed, for example, at moments of acceleration and deceleration in a HEV, whereas the battery provides the average power needed at non-accelerating times [6, 164] . This enables a dramatic increase of battery cycle life and prevents oversizing of either batteries or EDLCs avoiding costrelated issues, although more sophisticated power engineering circuits and control are needed. Other EDLC applications include serving as memory back-ups in cameras, computers, mobiles, and video recorders. EDLCs are becoming rapidly commercialized in cordless applications such as screw drivers and electric cutters where cycle life of EDLCs exceeds that of the application. EDLCs have also been used to supply quick power demands in emergency doors, for example the A380 superjumbo (eight supercapacitors per door) shown in Figure 2.11d. Figure 2.11a shows some commercially available supercapacitors and their 66 application in toys (Figure 2.11b), memory back up in portable computers (Figure 2.11c), and in a sea-port hybrid diesel/electric gantry crane (Figure 2.11f) where a supercapacitor stores energy during load lowering allowing for 40 % energy savings [163, 167] . Figure 2.11: (a) Some commercially available supercapacitors and their application in (b) toys, (c) memory back up in a laptop, (d) emergency door of a A380 superjumbo plane (schematic), (e) power buffering in a hybrid electrical car (schematic), and (e) a rubber tired gantry crane [163, 167] . 67 2.7 Summary and key opportunities Considerable research effort has been devoted to the search for suitable materials for supercapacitor applications. Activated carbon remains an excellent low cost electrode material with acceptable performance and is currently being commercialized. Carbon nanotubes constitute an alternative high surface area, and highly conductive material that offers the possibility of manufacturing free standing electrodes but their current high manufacturing cost and processing difficulties so far have prevented their use for commercial applications. However, costs are reducing and know-how is increasing. The “natural” tendency of CNTs to form flexible thin films with intrinsic porosity remains highly attractive. Graphene is a carbon form with exceptional properties and currently its suitability for supercapacitors and battery applications is intensively investigated including: synthesis, texture design of graphene sheets for optimum charge storage, electrode design to optimize accessibility of electrolytes, and to prevent restacking of graphene sheets, and its combination with pseudocapacitive materials to enhance energy density. Several transition metal oxides and some nitrides are currently being investigated as pseudocapacitive materials with a scientific focus on the design of 3D nanostructures (including nanoparticles, nanotubes, nanorods, and interleaved nanometer-sized layers), and the synthesis of composites with carbon materials. Although, outstanding performance is observed for some nanoarchitectures, manufacturing costs are prohibitively expensive with no possibility for scale-up. A serious rival to AC symmetric supercapacitors has yet to emerge. Hybrid supercapacitors offering high energy densities and 68 long cycle life appear to offer the most scope for technological innovation. The overall performance optimization of supercapacitors needs to observe key manufacturing elements related to the different components of the device seeking for a high energy density, power density and optimum cyclability using materials of low cost and environmental friendly nature. Manufacturing methods must allow an accurate and reproducible way to control mass load and thickness; in this respect, spray deposition methods constitute a scalable manufacturing alternative to produce large area, thin and uniform supercapacitor electrodes without the use of polymeric binders which helps reducing ESR and thus improving energy and power density. In this thesis a spray deposition technique is used for electrode manufacturing and it is the focus of the next two chapters. 69 Bibliography [1] J. Baxter, Z. Bian, G. Chen, D. Danielson, M. S. Dresselhaus, A. G. Fedorov, T. S. Fisher, C. W. Jones, E. Maginn, U. Kortshagen, A. Manthiram, A. Nozik, D. R. Rolison, T. Sands, L. Shi, D. Sholl and Y. Wu, Energy Environ. Sci., 2009, 2, 559–588. [2] M. Winter and R. J. Brodd, Chem. Rev., 2004, 104, 4245–4270. [3] B. E. Conway, Electrochemical Supercapacitors: Scientific Fundamental and Technological Applications, Kluwer Academic/ Plenum Publishers, New York, USA, 1999. [4] A. Burke, J. Power Sources, 2000, 91, 37–50. [5] A. J. Bard and F. L., Electrochemical Methods: Fundamentals and Applications, John Wiley & Sons, Inc., Unites States of America, 2nd edn., 2001. [6] P. Simon and Y. Gogotsi, Nat. Mater., 2008, 7, 845–854. [7] J. H. Chen, W. Z. Li, D. Z. Wang, S. X. Yang, J. G. Wen and Z. F. Ren, Carbon, 2002, 40, 1193–1197. 70 [8] B. Zhang, J. Liang, C. L. Xu, B. Q. Wei, D. B. Ruan and D. H. Wu, Mater. Lett., 2001, 51, 539–542. [9] A. Burke, Electrochim. Acta, 2007, 53, 1083–1091. [10] Y. Zhang, H. Feng, X. Wu, L. Wang, A. Zhang, T. Xia, H. Dong, X. Li and L. Zhang, Int. J. Hydrogen Energy, 2009, 34, 4889–4899. [11] B. E. Conway, J. Electrochem. Soc., 1991, 138, 1539–1548. [12] A. G. Pandolfo and A. F. Hollenkamp, J. Power Sources, 2006, 157, 11–27. [13] B. E. Conway, V. Birss and J. Wojtowicz, J. Power Sources, 1997, 66, 1–14. [14] A. Shukla, S. Sampath and K. Vijayamohanan, Curr. Sci. India, 2000, 79, 1656–1661. [15] J. Luo, J. Liu, P. He and Y. Xia, Electrochim. Acta, 2008, 53, 8128–8133. [16] R. Compton and C. Banks, Understanding Voltammetry, World Scientific Publishing Co. Pte. Ltd., Oxford, UK, 2007. [17] C. Portet, G. Yushin and Y. Gogotsi, Carbon, 2007, 45, 2511–2518. [18] D. A. C. Brownson, D. K. Kampouris and C. E. Banks, J. Power Sources, 2011, 196, 4873–4885. [19] P. Simon and A. Burke, Electrochem. Soc. Interface, 2008. [20] E. Frackowiak and F. Béguin, Carbon, 2001, 39, 937–950. 71 [21] H. Pröbstle, M. Wiener and J. Fricke, J. Porous Mater., 2003, 10, 213– 222. [22] Y. Gogotsi, S. Dimovski and J. A. Libera, Carbon, 2002, 40, 2263–2267. [23] H. Pierson, Handbook of Carbon, Graphite, Diamond and Fullerenes, Noyes Publications, NJ, USA, 1993. [24] R. Bansal, J. B. Donnet and F. Stoeckli, Active carbon, Marcel Dekker, New York, 1988. [25] D. Lozano-Castell, D. Cazorla-Amors, A. Linares-Solano, S. Shiraishi, H. Kurihara and A. Oya, Carbon, 2003, 41, 1765–1775. [26] J. Chmiola, G. Yushin, Y. Gogotsi, C. Portet, P. Simon and P. L. Taberna, Science, 2006, 313, 1760–1763. [27] T. Katakabe, T. Kaneko, M. Watanabe, T. Fukushima and T. Aida, J. Electrochem. Soc., 2005, 152, A1913–A1916. [28] T. W. Ebbesen, H. J. Lezec, H. Hiura, J. W. Bennett, H. F. Ghaemi and T. Thio, Nature, 1996, 382, 54–56. [29] J. E. Fischer, H. Dai, A. Thess, R. Lee, N. M. Hanjani, D. L. Dehaas and R. E. Smalley, Phys. Rev. B, 1997, 55, R4921–R4924. [30] J. M. Boyea, R. Camacho, S. Turano and W. Ready, NLB, 2007, 4, 585–593. [31] E. Frackowiak, K. Metenier, V. Bertagna and F. Béguin, Appl. Phys. Lett., 2000, 77, 2421–2423. 72 [32] E. Frackowiak, K. Jurewicz, K. Szostak, S. Delpeux and F. Bguin, Fuel Process. Technol., 2002, 77-78, 213–219. [33] E. Frackowiak, S. Delpeux, K. Jurewicz, K. Szostak, D. Cazorla-Amoros and F. Béguin, Chem. Phys. Lett., 2002, 361, 35–41. [34] Q. Jiang, M. Z. Qu, G. M. Zhou, B. L. Zhang and Z. L. Yu, Mater. Lett., 2002, 57, 988–991. [35] C. Niu, E. K. Sichel, R. Hoch, D. Moy and H. Tennent, Appl. Phys. Lett., 1997, 70, 1480–1482. [36] C. G. Hu, W. L. Wang, S. X. Wang, W. Zhu and Y. Li, Diamond Relat. Mater., 2003, 12, 1295–1299. [37] X. Zhao, B. T. T. Chu, B. Ballesteros, W. L. Wang, C. Johnston, J. M. Sykes and P. Grant, Nanotechnology, 2009, 20, 065605. [38] G. Tobias, S. Lidong, C. Salzmann, Y. Huh and M. Green, J. Phys. Chem., 2006, 110, 22318–22322. [39] B. Ballesteros, G. Tobias, L. Shao, E. Pellicer, J. Nogues, E. Mendoza and M. Green, Small, 2008, 4, 1501–1506. [40] J. Bahr and J. Tour, J. Mater. Chem., 2002, 12, 1952–1958. [41] X. Zhao, B. Mendoza-Sánchez, P. J. Dobson and P. S. Grant, Nanoscale, 2011, 3, 839–855. [42] P. Papakonstantinou, R. Kern, L. Robinson, H. Murphy, J. Irvine, 73 E. McAdams, J. McLaughlin and T. McNally, Fuller. Nanotub. Car. N., 2005, 13, 91–108. [43] V. L. Pushparaj, M. M. Shaijumon, A. Kumar, S. Murugesan, L. Ci, R. Vajtai, R. J. Linhardt, O. Nalamasu and P. M. Ajayan, Proc. Natl. Acad. Sci. U.S.A., 2007, 104, 13574–13577. [44] D. Z. Wang, S. Pengcheng, L. Changong, W. Wu and S. Fan, Nanotechnology, 2008, 19, 075609(1–6). [45] C. Meng, C. Liu and F. Shoushan, Electrochem. Commun., 2009, 11, 186–189. [46] R. K. Das, B. Liu, J. R. Reynolds and A. G. Rinzler, Nano Lett., 2009, 9, 677–683. [47] K.-W. Nam, K.-H. Kim, E.-S. Lee, W.-S. Yoon, X.-Q. Yang and K.-B. Kim, J. Power Sources, 2008, 182, 642–652. [48] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, M. I. Katsnelson, I. V. Grigorieva, S. V. Dubonos and A. A. Firsov, Nature, 2005, 438, 197–200. [49] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva and A. A. Firsov, Science, 2004, 306, 666–669. [50] A. K. Geim and K. S. Novoselov, Nat. Mater., 2007, 6, 183–191. [51] Y. Zhang, Y.-W. Tan, H. L. Stormer and P. Kim, Nature, 2005, 438, 201–204. 74 [52] S. Pisana, M. Lazzeri, C. Casiraghi, K. S. Novoselov, A. K. Geim, A. C. Ferrari and F. Mauri, Nat. Mater., 2007, 6, 198–201. [53] S. Vivekchand, C. Rout, K. S. Subrahmanyam, A. Govindaraj and C. N. R. Rao, J. Chem. Sci., 2008, 120, 9–13. [54] K. S. Subrahmanyam, S. R. C. Vivekchand, A. Govindaraj and C. N. R. Rao, J. Mater. Chem., 2008, 18, 1517–1523. [55] S. Stankovich, D. A. Dikin, R. D. Piner, K. A. Kohlhaas, A. Kleinhammes, Y. Jia, Y. Wu, S. T. Nguyen and R. S. Ruoff, Carbon, 2007, 45, 1558–1565. [56] Y. Wang, Z. Shi, Y. Huang, Y. Ma, C. Wang, M. Chen and Y. Chen, J. Phys. Chem. C, 2009, 113, 13103–13107. [57] M. D. Stoller, S. J. Park, Y. W. Zhu, J. H. An and R. S. Ruoff, Nano Lett., 2008, 8, 3498–3502. [58] J. R. Miller, R. A. Outlaw and B. C. Holloway, Science, 2010, 329, 1637–1639. [59] X. Li, G. Zhang, X. Bai, X. Sun, X. Wang, E. Wang and H. Dai, Nat. Nanotechnol., 2008, 3, 538–542. [60] S. De, P. J. King, M. Lotya, A. O’Neill, E. M. Doherty, Y. Hernandez, G. S. Duesberg and J. N. Coleman, Small, 2010, 6, 458–464. [61] A. V. Murugan, T. Muraliganth and A. Manthiram, Chem. Mater., 2009, 21, 5004–5006. 75 [62] Y. Zhu, S. Murali, M. D. Stoller, A. Velamakanni, R. D. Piner and R. S. Ruoff, Carbon, 2010, 48, 2118–2122. [63] Y. Hernandez, V. Nicolosi, M. Lotya, F. M. Blighe, Z. Sun, S. De, I. T. McGovern, B. Holland, M. Byrne, Y. K. Gun’Ko, J. J. Boland, P. Niraj, G. Duesberg, S. Krishnamurthy, R. Goodhue, J. Hutchison, V. Scardaci, A. C. Ferrari and J. N. Coleman, Nat. Nanotechnol., 2008, 3, 563–568. [64] U. Khan, A. O’Neill, M. Lotya, S. De and J. N. Coleman, Small, 2010, 6, 864–871. [65] M. Lotya, P. J. King, U. Khan, S. De and J. N. Coleman, ACS Nano, 2010, 4, 3155–3162. [66] M. Lotya, Y. Hernandez, P. J. King, R. J. Smith, V. Nicolosi, L. S. Karlsson, F. M. Blighe, S. De, Z. Wang, I. T. McGovern, G. S. Duesberg and J. N. Coleman, J. Am. Chem. Soc., 2009, 131, 3611–3620. [67] J. S. Ronan, M. Lotya and J. N. Coleman, New J. Phys., 2010, 12, 125008. [68] J. N. Coleman, Adv. Funct. Mater., 2009, 19, 3680–3695. [69] J. J. Yoo, K. Balakrishnan, J. Huang, V. Meunier, B. G. Sumpter, A. Srivastava, M. Conway, A. L. Mohana Reddy, J. Yu, R. Vajtai and P. M. Ajayan, Nano Lett., 2011, 11, 1423–1427. [70] Y. Zhu, S. Murali, M. D. Stoller, K. J. Ganesh, W. Cai, P. J. Ferreira, A. Pirkle, R. M. Wallace, K. A. Cychosz, M. Thommes, D. Su, E. A. Stach and R. S. Ruoff, Science, 2011, 332, 1537–1541. 76 [71] G. Wang, B. Wang, X. Wang, J. Park, S. Dou, H. Ahn and K. Kim, J. Mater. Chem., 2009, 19, 8378–8384. [72] Y. Si and E. T. Samulski, Chem. Mater., 2008, 20, 6792–6797. [73] E. Yoo, J. Kim, E. Hosono, H.-s. Zhou, T. Kudo and I. Honma, Nano Lett., 2008, 8, 2277–2282. [74] S.-M. Paek, E. Yoo and I. Honma, Nano Lett., 2008, 9, 72–75. [75] L. Fenghua and et al., Nanotechnology, 2009, 20, 455602. [76] D. A. Dikin, S. Stankovich, E. J. Zimney, R. D. Piner, G. H. B. Dommett, G. Evmenenko, S. T. Nguyen and R. S. Ruoff, Nature, 2007, 448, 457–460. [77] C. Wang, D. Li, C. O. Too and G. G. Wallace, Chem. Mater., 2009, 21, 2604–2606. [78] C. Haiqun, B. M. Marc, J. G. Kerry, G. W. Gordon and L. Dan, Adv. Mater., 2008, 20, 3557–3561. [79] D. Li, M. B. Muller, S. Gilje, R. B. Kaner and G. G. Wallace, Nat. Nanotechnol., 2008, 3, 101–105. [80] D.-W. Wang, F. Li, J. Zhao, W. Ren, Z.-G. Chen, J. Tan, Z.-S. Wu, I. Gentle, G. Q. Lu and H.-M. Cheng, ACS Nano, 2009, 3, 1745–1752. [81] J. Yan, T. Wei, B. Shao, Z. Fan, W. Qian, M. Zhang and F. Wei, Carbon, 2010, 48, 487–493. 77 [82] J. Yan, T. Wei, Z. Fan, W. Qian, M. Zhang, X. Shen and F. Wei, J. Power Sources, 2010, 195, 3041–3045. [83] W. Sugimoto, H. Iwata, Y. Yasunaga, Y. Murakami and Y. Takasu, Angew. Chem. Int. Ed., 2003, 42, 4092–4096. [84] D. McKeown, P. Hagans, L. Carette, K. Russell and D. Rolison, J. Phys. Chem. B, 1999, 103, 4825–4832. [85] K. H. Chang, C. Hu and C. Y. Chou, Electrochim. Acta, 2009, 54, 978– 983. [86] C. C. Hu, K. H. Chang, M. C. Lin and Y. T. Wu, Nano Lett., 2006, 6, 2690–2695. [87] K. M. Lin, K. Chang, C. C. Hu and Y. Y. Li, Electrochim. Acta, 2009, 54, 4574–4581. [88] G. Bo, Z. Xiaogang, Y. Changzhou, L. Juan and Y. Long, Electrochim. Acta, 2006, 52, 1028–1032. [89] F. Pico, E. Morales, J. A. Fernandez, T. A. Centeno, J. Ibaez, R. M. Rojas, J. M. Amarilla and J. M. Rojo, Electrochim. Acta, 2009, 54, 2239– 2245. [90] X. J. He, Y. J. Geng, S. Oke, K. Higashi, M. Yamamoto and H. Takikawa, Synth. Met., 2009, 159, 7–12. [91] F. Pico, J. Ibanez, M. A. Lillo-Rodenas, A. Linares-Solano, R. M. Rojas, J. Amarilla and J. Rojo, J. Power Sources, 2007, 176, 417–425. 78 [92] C. Yuan, L. Chen, B. Gao, L. Su and X. Zhang, J. Mater. Chem., 2009, 19, 246–252. [93] K. Naoi, S. Ishimoto, N. Ogihara, Y. Nakagawa and S. Hatta, J. Electrochem. Soc., 2009, 156, A52–A59. [94] S. C. Pang, M. A. Anderson and T. W. Chapman, J. Electrochem. Soc., 2000, 147, 444–450. [95] S. F. Chin, S. C. Pang and M. A. Anderson, J. Electrochem. Soc., 2002, 149, A379–A384. [96] M. Toupin, T. Brousse and D. Bélanger, Chem. Mater., 2004, 16, 3184– 3190. [97] M. Chigane and M. Ishikawa, J. Electrochem. Soc., 2000, 147, 2246– 2251. [98] S. L. Kuo and N. L. Wu, J. Electrochem. Soc., 2006, 153, A1317–A1324. [99] L. Athouel, F. Moser, R. Dugas, O. Crosnier, D. Bélanger and T. Brousse, J. Phys. Chem. C, 2008, 112, 7270–7277. [100] D. Bélanger, T. Brousse and J. W. Long, Electrochem. Soc. Interface, 2008. [101] W. Wei, X. Cui, W. Chen and D. G. Ivey, J. Phys. Chem. C, 2008, 112, 15075–15083. [102] J. Li, X. Wang, Q. Huang, S. Gamboa and P. J. Sebastian, J. Power Sources, 2006, 160, 1501–1505. 79 [103] A. E. Fischer, K. A. Pettigrew, D. R. Rolison, R. M. Stroud and J. W. Long, Nano Lett., 2007, 7, 281–286. [104] J. N. Broughton and M. J. Brett, Electrochim. Acta, 2004, 49, 4439– 4446. [105] Z. Fan, J. Chen, B. Zhang, B. Liu, X. Zhong and Y. Kuang, Diamond Relat. Mater., 2008, 17, 1943–1948. [106] S. L. Chou, J. Z. Wang, S. Y. Chew, H. K. Liu and S. X. Dou, Electrochem. Commun., 2008, 10, 1724–1727. [107] S. Ma, K. Ahn, E. Lee, K. Oh and K. Kim, Carbon, 2007, 45, 375–382. [108] S. Ma, K. W. Nam, W. S. Yoon, X. Yang, K. Ahn, K. Oh and K. B. Kim, J. Power Sources, 2008, 178, 483–489. [109] R. K. Sharma, H. S. Oh, Y. G. Shul and H. Kim, Physica B, 2008, 403, 1763–1769. [110] X. Dong, W. Shen, J. Gu, L. Xiong, Y. Zhu, H. Li and J. Shi, J. Phys. Chem. B, 2006, 110, 6015–6019. [111] R. K. Sharma, A. C. Rastogi and S. B. Desu, Electrochim. Acta, 2008, 53, 7690–7695. [112] S. Y. Wang, K. C. Ho, S. L. Kuo and N. L. Wu, J. Electrochem. Soc., 2006, 153, A75–A80. [113] X. Zhao, C. Johnston and P. S. Grant, J. Mater. Chem., 2009, 19, 8755–8760. 80 [114] S. L. Kuo and N. L. Wu, Electrochem. Solid-State Lett., 2007, 10, A171–A175. [115] X. Du, C. Wang, M. Chen, Y. Jiao and J. Wang, J. Phys. Chem. C, 2009, 113, 2643–2646. [116] Y. Wang, K. Takahashi, K. H. Lee and G. Z. Cao, Adv. Funct. Mater., 2006, 16, 1133–1144. [117] C. M. Huang, C. C. Hu, K. H. Chang, J. M. Li and Y. F. Li, J. Electrochem. Soc., 2009, 156, A667–A671. [118] I. H. Kim, J. Kim, B. Cho and K. Kima, J. Electrochem. Soc., 2006, 153, A1451–11458. [119] I. H. Kim, J. Kim, B. Cho, Y. H. Lee and K. Kima, J. Electrochem. Soc., 2006, 153, A989–A996. [120] W. C. Fang, J. Phys. Chem. C, 2008, 112, 11552–11555. [121] A. C. Santulli, W. Xu, J. B. Parise, L. Wu, M. C. Aronson, F. Zhang, C. Y. Nam, C. T. Black, A. L. Tiano and S. S. Wong, Phys. Chem. Chem. Phys., 2009, 11, 3718–3726. [122] Z. Chen, Y. Qin, D. Weng, Q. Xiao, Y. Peng, X. Wang, H. Li, F. Wei and Y. Lu, Adv. Funct. Mater., 2009, 19, 3420–3426. [123] H. M. Zeng, Y. Zhao, Y. J. Hao, Q. Y. Lai, J. H. Huang and X. Y. Ji, J. Alloys Compd., 2009, 477, 800–804. 81 [124] D. Choi, G. Blomgren and P. Kumta, Adv. Mater., 2006, 18, 1178– 1182. [125] W. Xing, F. Li, Z. F. Yan and G. Q. Lu, J. Power Sources, 2004, 134, 324–330. [126] D. D. Zhao, S. J. Bao, W. J. Zhou and H. L. Li, Electrochem. Commun., 2007, 9, 869–874. [127] M. S. Wu, Y. A. Huang, C. H. Yang and J. J. Jow, Int. J. Hydrogen Energy, 2007, 32, 4153–4159. [128] M. S. Wu, Y. A. Huang, J. J. Jow, W. D. Yang, C. Y. Hsieh and H. M. Tsai, Int. J. Hydrogen Energy, 2008, 33, 2921–2926. [129] J. Chang, M. Park, D. Ham, S. B. Ogale, R. S. Mane and S. H. Han, Electrochim. Acta, 2008, 53, 5016–5021. [130] Y. Zheng, M. Zhang and P. Gao, Mater. Res. Bull., 2007, 42, 1740– 1747. [131] G. H. Yuan, Z. H. Jiang, A. Aramata and Y. Z. Gao, Carbon, 2005, 43, 2913–2917. [132] J. Y. Lee, K. Liang, K. H. An and Y. H. Lee, Synth. Met., 2005, 150, 153–157. [133] D. W. Wang, F. Li and H. M. Cheng, J. Power Sources, 2008, 185, 1563–1568. 82 [134] C. C. Hu and C. Y. Cheng, Electrochem. Solid-State Lett., 2002, 5, A43–A46. [135] Z. A. Hu, Y. L. Xie, Y. X. Wang, H. Y. Wu, Y. Y. Yang and Z. Y. Zhang, Electrochim. Acta, 2009, 54, 2737–2741. [136] E. H. Liu, W. Li, J. Li, X. Y. Meng, R. Ding and S. T. Tan, Mater. Res. Bull., 2009, 44, 1122–1126. [137] O. A. Shlyakhtin, A. M. Skundin, S. J. Yoon and Y. J. Oh, Mater. Lett., 2009, 63, 109–112. [138] C. Yuan, X. Zhang, L. Su, B. Gao and L. Shen, J. Mater. Chem., 2009, 19, 5772–5777. [139] F. Miao, B. Tao, P. Ci, J. Shi, L. Wang and P. K. Chu, Mater. Res. Bull., 2009, 44, 1920–1925. [140] R. Kötz and M. Carlen, Electrochim. Acta, 2000, 45, 2483–2498. [141] K. Naoi and P. Simon, Electrochem. Soc. Interface, 2008. [142] V. Khomenko, E. Raymundo-Piero and F. Béguin, J. Power Sources, 2006, 153, 183–190. [143] Q. Qu, P. Zhang, B. Wang, Y. Chen, S. Tian, Y. Wu and R. Holze, J. Phys. Chem. C, 2009, 113, 14020–14027. [144] T. Brousse, P. L. Taberna, O. Crosnier, R. Dugas, P. Guillemet, Y. Scudeller, Y. Zhou, F. Favier, D. Blanger and P. Simon, J. Power Sources, 2007, 173, 633–641. 83 [145] J. H. Park, O. O. Park, K. H. Shin, C. S. Jin and J. H. Kim, Electrochem. Solid-State Lett., 2002, 5, H7–H10. [146] Y. Wang and Y. Y. Xia, Electrochem. Commun., 2005, 7, 1138–1142. [147] W. H. Jin, G. T. Cao and J. Y. Sun, J. Power Sources, 2008, 175, 686–691. [148] G. Amatucci, F. Badway and A. D. Du Pasquier, Electrochemical Society, 2000, pp. 344–359. [149] A. S. Arico, P. Bruce, B. Scrosati, J. M. Tarascon and v. W. Schalkwijk, Nat. Mater., 2005, 4, 366–377. [150] C. Largeot, C. Portet, J. Chmiola, P. L. Taberna, Y. Gogotsi and P. Simon, J. Am. Chem. Soc., 2008, 130, 2730–2731. [151] T. Tsuda and C. Hussey, Electrochem. Soc. Interface, 2007, 16, 42–49. [152] A. Balducci, R. Dugas, P. L. Taberna, P. Simon, D. Ple, M. Mastragostino and S. Passerini, J. Power Sources, 2007, 165, 922–927. [153] A. Balducci, W. Henderson, M. Mastragostino, S. Passerini, P. Simon and F. Soavi, Electrochim. Acta, 2005, 50, 2233–2237. [154] A. Balducci, F. Soavi and M. Mastragostino, Appl. Phys. A, 2006, 82, 627–632. [155] S. Talapatra, S. Kar, S. K. Pal, R. Vajtai, L. Ci, P. Victor, M. M. Shaijumon, S. Kaur, O. Nalamasu and P. M. Ajayan, Nat. Nanotechnol., 2006, 1, 112–116. 84 [156] C. Portet, P. L. Taberna, P. Simon and C. Laberty-Robert, Electrochim. Acta, 2004, 49, 905–912. [157] D. N. Futaba, K. Hata, T. Yamada, T. Hiraoka, Y. Hayamizu, Y. Kakudate, O. Tanaike, H. Hatori, M. Yumura and S. Iijima, Nat. Mater., 2006, 5, 987–994. [158] J. Li, Q. M. Yang and I. Zhitomirsky, J. Power Sources, 2008, 185, 1569–1574. [159] J. R. McDonough, J. W. Choi, Y. Yang, F. La Mantia, Y. Zhang and Y. Cui, Appl. Phys. Lett., 2009, 95, 243109–243109. [160] W. Sun and X. Chen, Microelectron. Eng., 2009, 86, 1307–1310. [161] W. Sun and X. Chen, J. Power Sources, 2009, 193, 924–929. [162] K. Byrappa and Y. Massahiro, Handbook of hydrothermal technology. A technology for crystal growth and materials processing, Noyes Publications, New Jersey, USA, 2001. [163] J. R. Miller and P. Simon, Science, 2008, 321, 651–652. [164] F. Gassman, Europhys. News, 2003, 34, 176–180. [165] J. Tollefson, Nature, 2008, 456, 436–440. [166] M. Armand and J. M. Tarascon, Nature, 2008, 451, 652–657. [167] J. R. Miller and A. Burke, Electrochem. Soc. Interface, 2008, 53–57. 85 86 Chapter 3 A scalable spray deposition route for the manufacture of large area nanostructured supercapacitor electrodes. Part 1: optimization of spray deposition variables and development of a numerical model for the optimization of the spraying dynamics. 87 3.1 Introduction The optimization of the performance of supercapacitors involves of a large number of variables related to different key components: electrodes, electrolyte, current collectors, the electrode-electrolyte and electrode-current collector interfaces. Areas of investigation therefore include fundamental research on the performance of electrochemically active materials - including nanostructure design, porosity, orientation respect the current collector, study of electron and ion conducting properties -, the study of the composition and stability of electrolytes, the assembly and testing of the device as a full cell taking care of optimum energy and power density performance as well as cycling stability, and a follow up study of performance for a specific application. In order to achieve a high capacitance and therefore high energy density, the design of supercapacitor electrodes pursues the maximization of surface area available for charge storage and maximization of the utilization of electrochemically active materials. In order to achieve a high power density, among other variables to observe, it is necessary to minimize the equivalent series resistance of the supercapacitor which varies inversely with the electrode thickness and is improved by providing a good electrical contact between the electrochemically active material and the current collector. Traditional electrode fabrication methods derive from the battery industry and include cold rolling, pressing and molding, doctor blading and calendering of a paste or slurry that is typically prepared by mixing electrochemically active materials with conductive fillers to maintain in-plane conductivity and electrical integrity, and inert “binders” such as poly-(vinylidene fluo- 88 ride) (PVDF) to provide sufficient mechanical strength to the electrodes [1, 2] . The main drawbacks of these fabrication methods are: (1) a restricted control of mass loading and electrode thickness, and (2) the presence of electrochemically inactive materials. The use of binders decreases the surface area available for charge storage events: charge and mass transport, double layer formation, and pseudocapacitive activity. For instance, the presence of binders prevents electrolyte access to pores of activated carbon, and ion diffusion to the bulk of nanostructured pseudocapacitive materials [2] . Thin film electrodes maximize the utilization of electrochemically active materials as demonstrated for MnO2 : a capacitance of 1380 F g−1 , near to the theoretical capacitance, was obtained for thin film (< 2 µm) MnO2 electrodes where the redox activity was restricted to the surface of the electrochemically active material. A slight increase of mass load (µg range) led to a reduced coulombic efficiency showing that the bulk of the electrode remained electrochemically inactive or inaccessible [3] . Several methods have been used to prepare binder free electrodes: vacuum filtration, spin coating, colloidal templating, electrophoretic deposition, electrochemical deposition, inkjet printing, dip coating, and drop-drying [4–12] . These methods poorly control deposited mass, film thickness, morphology, and uniformity [13–15] . Other problems include poor reproducibility, long processing times, and low possibility for scalability. Here, the potential of a new spray deposition technique for the manufacture of binder-free thin film electrodes at a large scale has been investigated. Previous laboratory scale research has proved the capability of this technique to produce binder-free electrodes of 1 cm2 area and thickness in the range 0.5-1 µm with accurate 89 control of mass loading and uniformity [16] . In order to scale-up laboratory research findings, a semi-industrial scale spraying equipment has been assembled capable of producing 1500 cm2 area electrodes. In this chapter, the spraying equipment was commissioned and the variables of the spraying process were optimized in order to produce safely and reproducibly large area thin film nanostructured electrodes in a variety of materials. 3.2 Description of the spray deposition equipment Scaled up spray deposition equipment (SDE) has been assembled for the fabrication of 1500 cm2 area nanostructured films. Scheme 3.1 shows the main components and images of the equipment are shown in Figure 3.1. The SDE is comprised of the following key components: • Spraying Nozzles (B): ViscoMist atomizing nozzles (Lechler, Germany) of different sizes provide the possibility of spraying colloidal suspensions of different particle size, concentration and viscosity. Each nozzle possesses two liquid-feed ports enabling the mixing of two different suspensions immediately prior to atomization. The suspension is atomized by compressed air to form a defined cone shape that can be expanded to a flat fan shape via the use of secondary gas jets integrated in the spray head. • A software programmable linear drive (C) enables a reciprocating linear displacement of one or more securely attached spraying nozzles at a 90 velocity of 0.3 to 300 cm s−1 . • A metallic cylinder (D) of a 15 cm x 100 cm external surface area rotates beneath the reciprocating spray at rotational speeds up to 60 rpm. • Infrared preheating ceramic units, shown in Scheme 3.1, provide uniform heating to the polished aluminium cylinder interior surface while rotating so that a fugitive liquid component or “carrier” of a suspension is driven-off from a spray-deposited surface leaving solid components. • An ultraviolet irradiation source (not shown) enables the controlled in-situ UV-polymerization of a UV-polymerizable component of a suspension in order to produce a nanocomposite film (not used in this work). • A control panel (E) provides a user interface for the control of temperature, atomizing pressure, flow access to spraying nozzles, displacement of linear drive, rotational speed of the cylinder, and timing of the spraying. • A syringe pump (F) feeds suspensions to the spraying nozzle at 1 to 9 ml min−1 . Alternatively a peristaltic pump (not shown) is used for the spraying of larger volumes (> 1 L) of suspension. The general operation of the SDE can be described as follows: 1. The cylinder is covered with a flexible web substrate, for instance an Al coated polymer, and preheated to a desired temperature. The rotation 91 Scheme 3.1: Spray deposition equipment. Figure 3.1: Spray deposition equipment (A) and its components: nozzle (B), linear drive (C), metallic cylinder (D),control panel (E), pump(F). of the cylinder is started together with the preheating so that its surface is heated uniformly. 2. The ventilation extraction is started. 3. A previously prepared suspension is fed to the nozzle. The atomizing air is started at a pre-determined atomizing pressure. 92 4. The linear drive reciprocating movement is started. 5. The unit is closed during the progress of the spray deposition. 6. Upon completion of spray deposition, the suspension flow to the nozzles and atomizing pressure are turned off; the heating elements are turned off and the cylinder continues rotating until cool down to room temperature. 7. The film is then collected while maintaining the ventilation on. 8. The ventilation and the main power switch of the machine are turned off. The cylinder has been machined to create a set of 4 flat surface sections of 26 mm x 150 mm distributed across the 100 cm external circumference to hold several flat and stiff substrates for the simultaneous spray deposition of several laboratory-scale flat samples. In this case one or more stiff substrates are fixed to any of the flat surfaces; the cylinder is kept at a static position exposing the substrates to the spraying flow while the linear drive reciprocates. The machined flat surfaces were designed to hold substrates of the size of a standard microscope glass slide (26 mm x 75 mm) but for typical supercapacitor electrode testing several substrates (up to 12) of 1 cm x 2 cm were simultaneously spray deposited, each over a 1 cm2 area. Alternatively, the substrates were fixed instead to the flat surface of a hot plate allowing simultaneous spray deposition of up to 20 1 cm x 2 cm substrates. 93 3.3 Variable Optimization The several variables associated with the spray-deposition process required optimization for the manufacture of a thin while uniform large area nanostructured films of 1500 cm2 with consistent electrochemical properties. The optimized process variables were: (1) distance between atomizing point and target (spray distance z ), (2) atomizing pressure P, (3) the spatial mass distribution within the spray cone which under static conditions produces a deposited mass profile ( the “footprint”), (4) dynamics of spraying determined by both the linear drive speed vy and the rotational speed of the cylinder wx . 3.3.1 Spray distance z In order to assess the effect of spray distance on film properties, a fluoropolymer aqueous suspension, PFA (perfluoroalcoxy) 20 w/w % concentration was sprayed onto glass substrates at P =165.47 kPa atomizing pressure, room temperature, and a flow rate F = 2 ml min−1 . The spray distance z was varied across the operationally available range from a maximum z max = 14.8 cm to a minimum z min = 8.8 cm in 1 cm steps under static spray conditions to produce the dried ‘footprint’ samples A, B, C, D, E, F, G shown in Figure 3.2. The ‘quality’ of the films in terms of uniformity, assessed by eye, decreased as z decreased. The quantification of roughness or surface topography was not possible due to the presence of cracks on all samples. At short spraying distances, convection forces due to the impinging gas stream became important and disrupted the formation of a uniform film; the spraying cone was also 94 narrower confining a given amount of sprayed suspension into a smaller area leading to accumulation of suspension at the target where in-situ drying was overwhelmed by the incoming flux of suspension with consequent formation of cracks. Therefore, z max was chosen as the optimum spray distance for subsequent experiments. Figure 3.2: Optimization of spray distance (z ). PFA 20 w/w % aqueous suspension sprayed onto glass substrates at decreasing z from A to G (P = 165.47 kPa, flow rate F = 2 ml/min, room temperature). 3.3.2 Atomizing pressure P Glass substrates were again sprayed with PFA 20 w/w % aqueous suspension at increasing atomizing pressures from 68.94 kPa to 344.73 kPa. All other conditions were kept constant: F = 2ml min−1 , room temperature, z max =14.8 cm. The spraying time t was 0.5 s. The arithmetic average roughness of the resulting uncured polymer films Ra was measured for all samples using a profile-meter (scan length = 5,000 µm) as shown in Figure 3.3a. The roughness decreased as the atomizing pressure increased as further shown in the 3D surface roughness maps obtained by confocal microscopy in Figure 3.3b and Figure 3.3c. 95 Figure 3.3: (a) Arithmetic average roughness Ra (5000 µm scan length) measured across the spray-deposited PFA films every ∆x = 0.5 cm starting from the center of the footprint at x = 0, and at different atomizing pressures; (b) and (c) 3D surface roughness maps obtained by confocal imaging of PFA films spray deposited at 68.94 kPa and 344.73 kPa respectively. The variation of PFA film thickness as a function of atomizing pressure was also evaluated using a profile-meter, and the measurements taken at the centre and periphery - localized at a distance x from the centre of the footprint which slightly varied with atomizing pressure but thickness measurements were taken at x = 4.0 cm - of each film are shown in Figure 3.4. Three layers were sprayed for each film. The spraying time for each layer was 0.5 s allowing each layer to dry before the next one was sprayed. At the center of the footprint the thickness decreased as atomizing pressure increased whereas at the periphery the thickness remained approximately constant. The spray deposition process is a complex series of events with key interdependent variables such as droplet size and velocity, spray shape, mass and 96 energy fluxes which are function of the various process parameters previously outlined. Mi et al. provides a description of the key underlying physics of the spray forming of alloys with which the current spray process shares many features: the atomizing pressure produces a large number of droplets (spray mist) with each droplet traveling to the substrate with a specific momentum; at the point of contact with the substrate the droplet may “stick ” or deposit (primary mass deposition) or bounce/splash to another position (secondary mass deposition or redeposition) or be lost from the system as “overspray”. Mi et al. considered two mechanisms of redeposition, a bouncing mechanism of solid particles (elastic scattering) and a “splashing” mechanism of large liquid droplets [17] . Under the assumption that the spray deposition of aqueous suspensions can be described by similar events, an approximate qualitative explanation for the film thickness variation with atomizing pressure can be given. High atomizing pressures produced smaller, lower mass droplets which were more easily accelerated by the atomizing gas involving an overall increase in kinetic energy which lead to: (1) a thinner film produced by primary deposition of small droplets (small mass), and (2) a high degree of redeposition due to splashing at positions far from the centre of the spray footprint, both of which promoted a tendency of formation of thinner films. On the other hand, small atomizing pressures produced larger droplets traveling at a reduced velocity to the substrate and therefore bouncing or splashing less energetically producing a thicker film due to primary deposition of those larger drops. Overspray causing ‘loss’ of material from the system was increased at high pressures where thinner films were spray deposited. At the same time 97 the distribution of droplets within the spray cone, which is ultimately reflected on the mass distribution of the spray-deposited film, was also affected by the atomizing pressure. As a first approximation the mass distribution of the spray footprint could be considered to follow a Gaussian mass distribution - in fact confirmed in next section- according to which the film thickness would be maximum at the centre of the footprint and increasingly reduced towards the periphery. According to the thickness variations shown in Figure 3.4, as the atomizing pressure increased the mass distribution tended to ‘even out’ , with the thickness at the centre decreasing and reaching similar values than in the periphery. Therefore, the overall result of increasing atomizing pressure is the formation of thinner films with a flatten out mass distribution. Therefore, for the thinnest and most uniform film at a constant pressure of 344.73 kPa was used for all subsequent optimization. Figure 3.4: PFA film thickness at centre and periphery (x = 4 cm measured from the centre of the footprint) of the footprint as function of atomizing pressure. 98 3.3.3 Mass Distribution In order to determine the spatial mass distribution of spray-deposited films, a PFA 20 w/w % aqueous suspension was sprayed onto a flat glass substrate for t= 0.5 s x 5 at z = z max , P = 344.73 kPa, F = 2 ml min−1 , and room temperature. The film thickness was measured along x and y axes as indicated in Figure 3.5 where the central circle represents the plan view of the footprint with a radius Rf = 3.2 cm - set as the radius where thickness was > 0.1 µm -. The measurements were taken at a constant ∆x = 0.25 cm steps in x direction at constant y= 0 cm, y= -0.5 cm, and y= 0.25 cm as shown in Figure 3.5a. Similarly, measurements in y direction were taken every y= 0.25 cm steps at constant x = 0, x = 0.25 cm and x = -0.25 cm as shown in Figure 3.5b. Figure 3.5a and Figure 3.5b show near Gaussian distributions of thickness in x and y directions. After normalization, the data was fitted to a Gaussian distribution using a least squares fitting method. As shown in Figure 3.5c and Figure 3.5d, the best fit, corresponding to the minimum sum of squared residuals (s), for x at y = 0 and y at x = 0 was for Gaussian distributions with standard deviations of σx = 1.1 cm and σy = 1.1 cm, respectively. The best fit to the thickness data collected across the full footprint was a 3D Gaussian distribution with σx = 1.0 cm and σy = 0.6 cm where the sum of squared residuals calculated for component 2D Gaussians along x (sx ) and y (sy ) direction ranged as 0.0062 < sx < 0.24 and 0.0081 < sy < 0.11 respectively. Repeating the fitting procedure for an experimental data set corresponding to a film spray deposited at P = 68.94 kPa also showed a 99 best fit to a 3D Gaussian distribution suggesting that the spray cone mass distribution has this near-symmetrical Gaussian generic shape. Figure 3.5: Mass distribution determination: thickness of a PFA film sample measured at different locations in (a) x and (b) y directions. Best-fit Gaussian distribution for (c) x at y = 0 cm and (d) y at x = 0 cm showing standard deviations (σx , σy ) and sum of squared residuals (s), and (e) 3DGaussian fit (black grid) to thickness measurements (coloured surface) of a PFA film. 100 3.3.4 Dynamics of spraying The linear drive shown in Figure 3.1 provides displacement in y direction while the cylinder rotates providing tangential displacement in x. The resulting displacement that the axis of the spray cone makes over the cylinder surface, i.e. the resultant vector of the two motions, at constant linear velocities in x and y directions for each revolution is a zigzag-like spray pattern (from now on called the “path”). If after one revolution of the cylinder, the spray cone returns to the same starting point, the second path will deposit on top of the first one, and upon repetition of this process for several revolutions a “stripe” of deposition would develop. In this case the relative motions can be described as “in phase”. For an “out of phase” case, during the second revolution, the second path is displaced in the x direction relative to the start point of the first path, upon continued rotation of the cylinder subsequents paths will be displaced from each other until a spray pattern starts to develop covering the spray surface area. This concept is illustrated in Figure 3.8a and Figure 3.8c. The number of cylinder revolutions required to cover the entire cylinder surface depends on the size and mass distribution of the footprint, and the relative motions of linear and rotary drives. Furthermore, once suspension has been deposited by a single path, time must be allowed for drying in order to avoid post-deposition flow caused when a second and subsequent paths with a degree of overlapping deposit suspension over a non-dried surface. Thus, the time interval before adjacent paths are sprayed and the distance between nearest temporal neighbour paths must be maximized. 101 This optimization problem was addressed by using a numerical methods approach. The movement of the linear drive was idealized as produced by a rotor with circular uniform motion where the linear drive displacement y, and with it the spray cone, was that of a ficitious body traversing the circular path C or circumference of the rotor at constant angular speed ωy and constant linear speed vy = ωy ry where ry is the radius of the idealized rotor. The maximum displacement of the linear drive y = 26.1 cm was considered to be produced at half turn (180 ◦ ) of the idealized rotor so that y = C/2 = 2πry /2 = πry giving ry = y/π. Upon completion of half turn, the idealized rotor changed direction to give a negative displacement -y of the linear drive. Acceleration and deceleration effects upon change of direction where considered to be zero. In practice reversals in direction of the linear drive take place beyond the edge of the substrate width therefore acceleration and deceleration effects on the spray deposition must be minimum. In order to build up a uniform film, the ratio of angular speeds of the idealized linear drive rotor (ωy ) and the cylinder (ωx ) must be a coprime rational number, i.e. a number that can be expressed as a ratio of two integers with no common divisor. A coprime number assures out of phase and equally spaced spray paths, thus achieving a uniform film after a certain number of revolutions. However, setting the ratio of angular speeds to a coprime number results in adjacent spray paths being consecutively sprayed. In order to avoid post-deposition overflow, the spray of adjacent paths must be also temporally spaced. The solution to both constraints of film uniformity and maximum temporal spacing of adjacent paths was found to be the ratio of consecutive elements of the Fibonacci series [18] , ωy /ωx = F(n+1)/F(n) 102 where n is an integer. Figure 3.6a shows and example of a computed sprayed pattern using Matlab software for an idealized linear and rotary drive arrangement where ωy /ωx = F(10)/F(9)= 55/34. A uniform, complete and interleaved spray pattern is completed for a number of cylinder revolutions r = F(n). As n is increased in the Fibonacci ratio, F(n+1)/F(n), the distance between adjacent spray paths in a complete spray pattern, x p , logarithmically decreased producing a denser pattern as observed by comparing the patterns in Figure 3.6a and Figure 3.6b. The parameter x p is critical to achieving uniformity at minimum film thickness as explained next. Figure 3.6: Matlab computed spraying patterns for (a) ωy /ωx = F(10)/F(9)= 55/34, xp = 1.82 cm and (b) ωy /ωx = F(7)/F(6) = 13/8, xp = 7.71 cm. In the previous section, the mass distribution of the footprint was shown 103 to approximate to a Gaussian distribution. As the spray nozzle follows the zigzag-like paths according to Figure 3.6, mass will accumulate resulting in formation of a film as shown in Figure 3.7a. The distance xp determines the macro-scale waviness or degree of evenness of the surface. For the surface in Figure 3.7b, the Fibonacci ratio was F(10)/F(9)= 55/34 giving a xp =1.82 cm which for the particular Gaussian distribution based on experiments resulted in a surface of low waviness. For the surface shown in Figure 3.7c, the Fibonacci ratio has been decreased to F(8)/F(7)= 21/13 which gave a xp =4.77 cm resulting in a wavy and uneven surface, but a thinner film. Further simulations showed that for Fibonacci ratios with n > 9, the surfaces become smooth as x p further decreases but, for the purpose of thin film supercapacitor manufacturing, the film becomes unnecessarily thick. Therefore to achieve a uniform film of minimum thickness ωy /ωx = F(10)/F(9)= 55/34 was selected as the optimum ratio. The progress of the spray deposition until the achievement of a complete uniform spray pattern is described by several key variables fully determined by the ratio of the velocities of the spray deposition system set as ωy /ωx = F(10)/F(9)= 55/34. Figure 3.8 shows that as the spray progresses, the number of spray paths increases in 1 per each revolution r having a spacing relationship with one another described by xpar , the orthogonal distance between parallel segments of a pair of compared spray paths (i, j). As the number of revolutions r increases, there are N = r!/2(r-2)! pairwise combinations xpar (i, j). Figure 3.9 shows that at revolution r, a number Nu = r-1 of unique spacings xpar (1, j) appear at different points of the sprayed area with a frequency that is increased by 1 per increasing revolution r. The red 104 Figure 3.7: Spray deposition process model: (a) Gaussian distributions building up a surface, (b) a uniform and even spray deposited film for F(10)/F(9) = 55/34, x p = 1.82 cm, and (c) a uniform but uneven spray deposited thinner film for F(8)/F(7) = 21/13, x p = 4.77 cm. bars in Figure 3.9 indicate xpar (i, j) < 2Rf , where Rf = 3.2 cm as described previously in section 3.3.3, in which case overlapping of sprayed paths occurs giving rise to film formation as illustrated in Figure 3.7a. In order to prevent 105 post-deposition flow the frequency of those spacings should be minimized and the time interval between the spray deposition of spray paths i and j, t(i, j), should be maximized. Figure 3.9c shows that for r = 34, the minimum t(i, j) = 9.6 s, and therefore the carrier or fugitive liquid in a suspension must be removed at times td < 9.6 s. Figure 3.8: Spraying paths for ωy /ωx = F(10)/F(9)= 55/34. First sprayed paths (numbered in colour) deposited for a rotational displacement up to cylinder revolution (a) r = 3 , and (c) r = 5; (b) and (d) show the spacings xpar (i, j) (indicated with spacing lines in (a)) as spray progresses up to revolution r. Figure 3.10 shows that the minimum xpar (i, j) per revolution decreases as the number of r revolutions increases - the spray pattern becomes denser tending to be a constant and unique value xpar = 1.17 cm from r = 14 to r 106 Figure 3.9: Spraying dynamics of the pattern in Figure 3.6 at ωy /ωx = F(10)/F(9)= 55/34. Spacings xpar (i,j) and their frequency as spray deposition progresses up to (a) r = 10, (b) r = 11, and (c) r = 34. The red bars indicate xpar (i,j) < 2Rf . The times ti,j are indicated at the top of bars of (c). = F(9)= 34 achieving thus a uniform pattern. 107 Figure 3.10: Spraying dynamics for a set up ωy /ωx = F(10)/F(9)= 55/34. Shortest xpar at revolution r as spray progresses up to r = F(9)= 34. The diameter of the spray footprint (2Rf ) is indicated by the blue line. 3.4 Summary of optimized spray deposition variables and implications of the experimental set up The optimization of the spray deposition variables and the evaluation of the spraying dynamics by a numerical method has been used to guide the selection of spray deposition parameters summarized in Table 3.1. The velocity of the linear drive in the SDE set as vy = 35.2 cm s−1 was not a constant in practice but followed a noisy pattern that involved microsecond-periods of acceleration and deceleration at the turning end points of the traverse as shown in Figure 3.11a. The tangential velocity of the cylinder vx = 41.8 cm s−1 was not monitored in real time and was considered to be a constant on the basis of a continuously rotating cylinder in a single direction. Figure 3.11b shows a spray pattern obtained by data logging the real vy shown in Figure 3.11a in the model previously developed with ωy /ωx = F(10)/F(9)= 108 55/34; the spray paths had more of a “sinusoidal” shape rather than a “zigzag” shape resulting from a displacement y ≈ 0 cm at vy ≈ 0 cm s−1 at the turning end points of the linear drive. The non-idealized linear motion lead to irregularities in the spacing of spray paths indicated further in the inset in Figure 3.11b. As shown in Figure 3.12, the spraying dynamics described by the frequency of spacings xpar (i, j) and its spray deposition timing t(i, j) was also deviated from the predicted behaviour for a constant vy . The spray pattern was nearly uniform with 0.60 cm < xpar (i, j) < 5.3 cm for the nearest spacial neighbours (i, j) at the end of spray deposition, but for a larger number of cylinder revolutions r = 41 rather than r = F(9 ) = 34 as predicted by the model, and for a total spray time of tspray = 1.64 min. The suspension carrier must be removed at td < t(i, j) = 7.2 s rather than t(i, j) < 9.6 s to avoid post-deposition flow. Table 3.1: Summary of optimized spray deposition variables. Symbols indicate: (∗ ) set in the numerical model, (∗∗ ) experimental values, and († ) theoretical value ≈ experimental value. Parameter Suspension flow (F ) Spray distance (z ) Atomizing pressure (P ) Angular velocities ratio (ωy /ωx ) Angular velocity of linear drive (ωy ) Angular velocity of cylinder (ωx ) Linear velocity of linear drive (vy ) Linear velocity of cylinder (vx ) Footprint radius (Rf ) Cylinder revolutions (r ) Time for removal of suspension carrier (td ) Spray time (tspray ) 109 Value 2 ml min−1 14.8 cm 344.8 kPa F(10)/F(9)= 55/34 ∗ 4.2 rad s−1 ∗ 2.6 rad s−1 † 35.18 cm s−1 ∗ 41.8 cm s−1 † 3.2 cm 34∗ , 41∗∗ 9.6 s ∗ , 7.2 s ∗∗ 1.36 min ∗ , 1.64 min ∗∗ Figure 3.11: (a) Velocity of the linear drive vy as a function of time showing a noisy pattern oscillating about the set point value of vy = 35.2 cm s−1 , (b) Matlab computed spraying pattern obtained by data logging vy in (a) at ωy /ωx = F(10)/F(9)= 55/34. The inset in (b) shows an enlarged view of an area of the spraying pattern showing some non-uniform spacings between spray paths. The deviation from the ideal behaviour of the linear drive velocity vy = constant prevents the achievement of an ideally uniform pattern. The acceleration and deceleration is intrinsic to the change of direction of the movement of the linear drive and therefore a condition that in practice can only be ameliorated. In order to diminish this effect, the variations of velocity of the linear drive (noise) were minimized as much as possible by suitable adjustments in the performance of the actuator that provides the motion of the linear drive resulting in the “smoothed” velocity profile shown in Figure 3.11a. 110 Figure 3.12: Spraying dynamics of the pattern in Figure 3.11a with ωy /ωx = F(10)/F(9)= 55/34 and an vy in Figure 3.11a, indicating spacings xpar (i, j) and their frequency as spray deposition progresses up to cylinder revolution r = 41. The red bars indicate xpar (i, j) < 2Rf . The times t(i, j) are indicated at the top of bars. 111 Conclusions • New equipment for the spraying of large area nanostructured films for supercapacitors electrode applications was built and commissioned. • The spray nozzle and resulting spray mass distribution was characterized as a function of important process variables under independent control. • A numerical model was developed to aid understanding of the interaction of the spray footprint and relative motion of linear and rotary drives, and to understand the ability of the equipment to produce uniform large area films of minimum thickness. • Experimental data was input into the model in order to predict uniformity variations of the sprayed surfaces under real conditions. The next chapter describes the spray deposition of large area films in order to test the capability of the optimized spray deposition procedure to produce thin films of uniform thickness, microstructure and electrochemical properties. 112 Bibliography [1] E. Frackowiak, K. Metenier, V. Bertagna and F. Béguin, Appl. Phys. Lett., 2000, 77, 2421–2423. [2] B. Zhang, J. Liang, C. L. Xu, B. Q. Wei, D. B. Ruan and D. H. Wu, Mater. Lett., 2001, 51, 539–542. [3] M. Toupin, T. Brousse and D. Bélanger, Chem. Mater., 2004, 16, 3184– 3190. [4] J. H. Kim, K.-W. Nam, S. B. Ma and K. B. Kim, Carbon, 2006, 44, 1963–1968. [5] A. A. Green and M. C. Hersam, Nano Lett., 2008, 8, 1417–1422. [6] M. Kaempgen, G. S. Duesberg and S. Roth, Appl. Surf. Sci., 2005, 252, 425–429. [7] M. H. Kim, J. Y. Choi, H. K. Choi, S. M. Yoon, O. O. Park, D. K. Yi, S. J. Choi and H. J. Shin, Adv. Mater., 2008, 20, 457–461. [8] K. Kords, T. Mustonen, G. Tth, H. Jantunen, M. Lajunen, C. Soldano, S. Talapatra, S. Kar, R. Vajtai and P. M. Ajayan, Small, 2006, 2, 1021– 1025. 113 [9] M. H. Andrew Ng, L. Hartadi, H. Tan and C. H. P. Poa, Nanotechnology, 2008, 19, 205703. [10] A. Schindler, J. Brill, N. Fruehauf, J. P. Novak and Z. Yaniv, Physica E, 2007, 37, 119–123. [11] Z. Wu, Z. Chen, X. Du, J. M. Logan, J. Sippel, M. Nikolou, K. Kamaras, J. R. Reynolds, D. B. Tanner, A. F. Hebard and A. G. Rinzler, Science, 2004, 305, 1273–1276. [12] S. W. Lee, B.-S. Kim, S. Chen, Y. Shao-Horn and P. T. Hammond, J. Am. Chem. Soc., 2008, 131, 671–679. [13] C. G. Hu, W. L. Wang, S. X. Wang, W. Zhu and Y. Li, Diamond Relat. Mater., 2003, 12, 1295–1299. [14] B. Gao, G. Z. Yue, Q. Qiu, Y. Cheng, H. Shimoda, L. Fleming and O. Zhou, Adv. Mater., 2001, 13, 1770–1773. [15] C.-y. Liu, A. J. Bard, F. Wudl, I. Weitz and J. R. Heath, Electrochem. Solid-State Lett., 1999, 2, 577–578. [16] X. Zhao, B. T. T. Chu, B. Ballesteros, W. Wang, C. Johnston, J. M. Sykes and P. S. Grant, Nanotechnology, 2009, 20, 065605. [17] J. Mi and P. Grant, Acta Mater., 2008, 56, 1588–1596. [18] V. Helmut, Math. Biosci., 1979, 44, 179–189. 114 115 Chapter 4 A scalable spray deposition route for the manufacture of large area nanostructured supercapacitor electrodes. Part 2: validation of the optimization of the spray deposition procedure with the manufacture and investigation of properties of first large area electrodes. 116 4.1 Introduction In the previous chapter a spray deposition technique was developed for the fabrication of large area 1500 cm2 nanostructured thin film electrodes for supercapacitors where the importance of producing binder-free thin film electrodes was underlined. In this chapter preliminary large area films were manufactured in order to assess the overall process capability and to validate the optimization previously described. As a typical model system, stable aqueous suspensions of carboxyl-functionalized multi-walled carbon nanotubes were formulated and spray-deposited to form thin film mesoporous electrodes onto Al coated flexible polyethylene terephthalate (PET) webs, without the use of binders. The resulting uniformity of thickness, surface morphology, and electrochemical properties were then investigated. 4.2 Experimental details Materials. Chemical vapor deposition (CVD)-grown multi-walled carbon nanotubes (MWNTs) with 4 nm inner diameter, 13 nm outer diameter, and average length of 1 µm were supplied by Bayer Material Science (Germany); nitric acid (69 % ) and hydrochloric acid (37% ) by Sigma Aldrich (UK); Poly (ethyleneimine) (PEI, Mw= 70,000 ) by Alfa-Aesar (UK); and indium tin oxide (ITO) coated glass substrates (7 Ω/sq sheet resistance) by Delta Technologies (USA). Deionized water (10 MΩ · cm) was used for all processing. MWNT processing. MWNTs (1 g) were steam purified as described else- 117 where [1, 2] ]. Subsequently, MWNTs were carboxyl-functionalized (MWNTCOOH) by refluxing in HNO3 (9 M, 18 hours, 100 ◦ C ) followed by filtration and extensive rinsing with deionized water until neutral pH . Aqueous dispersions of MWNTCOOH (1.5 mg ml−1 ) were produced by repeating 5 times alternate ultrasonication (600 W, 20 kHz probe, 2 min) and high shear mixing for (5 min) while maintaining ice cooling. Large area and laboratory scale electrode fabrication. The spray deposition equipment (SDE) was configured to the optimized variables described in Table 1 of Chapter 3 used to manufacture electrodes of 1500 cm2 . A spray deposited single layer is defined as a sprayed deposited pattern completed in 1.64 min (41 revolutions) as described in chapter 3. The large area electrodes manufactured here were comprised a total of 23 layers. Prepared suspensions of MWNTCOOH were spray deposited onto aluminum coated polyethylene thereftalate (PET) flexible films (15 cm x 100 cm) maintained at a temperature of 90 ◦ C. Laboratory scale electrodes were spray deposited onto ITO coated glass substrates (2.5 cm x 5.0 cm total area, 2.5 cm x 2.5 cm electrode area) which were attached onto PET films prior to spray deposition of large area films and used for thickness and electrochemical properties evaluation (easier on stiff rather than directly on the flexible large area electrode, both manufactured simultaneously and under identical conditions). In order to enhance the adhesion and stability of the MWNTCOOH films during electrochemical measurements, prior to spray-deposition of electrochemically active material, substrates were pre-coated with a ≈ 20 nm layer PEI (1.0 % w/w) aqueous solution. Equipment and characterization techniques. Scanning electron microscopy 118 (SEM) images were obtained in a JSM6500F operated at 5 kV and 15 mm working distance. Surface chemical groups were characterized by Xray photoelectron spectroscopy (XPS) in an ion pumped ESCA200 (ScientaGammadata ESCA 200 Upsala Sweden) equipped with a monochromatic Al Kα source, with samples supported on Si/SiO2 substrates. The analyzer operated at a constant pass energy of 500 eV for wide scans and 150 eV for detailed scans. Electrode thickness was determined by step height measurements using a Dektak 6M profilometer (Veeco Instruments, Inc). The weight of deposited films was measured using a Sartorius microbalance with 0.01 mg readability. Electrochemical characterization. Electrochemical measurements were carried out using a Gill potentiostat/galvanostatic/frequency response analyzer in a three electrode cell using the MWNTs film electrodes as the working electrode, a platinum (Pt) plated titanium sheet as the counter electrode, and a Ag/AgCl as the refererence electrode. The electrolyte used was 0.1 M H2 SO4 aqueous solution. Cyclic voltammetry was performed in the potential range of 0.0 to 0.8 V at scan rates of 10, 50, 100 mV s−1 . Electrochemical Impedance Spectroscopy (EIS) measurements were carried out by applying an AC voltage of 20 mV rms in the frequency range 0.01-3000 Hz at room temperature. 119 4.3 4.3.1 Results and discussion XPS characterization of surface functionalized MWNTs XPS elemental analysis of the carboxyl functionalized MWNTs is shown in Figure 7.3 where C1s and O1s photoemission peaks were observed at 284.5 eV and 533 eV respectively. Each peak was deconvoluted into separate GaussianLorentzian shape components to account for the contribution of different functional groups. As shown in Figure 7.3a, the C1s photoemission peak had components attributed to sp2 hybridized graphite-like carbon (284.5 ± 0.1 eV), C-O (285.9 ± 0.1 eV), and -O=C-OH (288.7 ± 0.1 eV) [3–6] . The O1s photoemission peak shown in Figure 7.3b provided evidence for the same functional groups with components attributed to -(O*=C)-OH (530.85 ± 0.1 eV), -C-O ( 532.05 ± 0.1 eV), and -(O=C)- O*-H (534.1 ± 0.1 eV). 4.3.2 Spray deposition of large area uniform thin films The SDE was set up to the optimized variables summarized in Table 1 in chapter 3 for the spray deposition of a MWNTCOOH film comprised of 23 layers (see experimental details) onto aluminum-coated PET, as shown in Figure 4.2. No post-deposition flow was observed by eye during manufacturing suggesting that the drying time at a substrate temperature of 90 ◦ was td < 7.2 s as indicated by the numerical model developed in the previous chapter. As shown later, the film uniformity supports the absence of any significant post-deposition flow. 120 Figure 4.1: X-ray photoelectron spectroscopy (XPS) spectra of functionalized MWNTs: (a) C1s photoemission peak, and (b) O1s photoemission peak. Peaks are deconvoluted in Gaussian-Lorentz components. Figure 4.2: Optical images of a flexible large area electrode (100 cm x 15 cm) manufactured by spray-deposition of an aqueous suspension of carboxylfunctionalized MWNTs onto a PET substrate. 4.3.3 Thickness Analysis In order to analyze the uniformity of thickness of the large area electrode, prior to spraying, 1 cm2 Si substrates were attached to the PET substrate 121 at 24 uniformly distributed locations across the entire substrate area. After spraying, the Si substrates were dettached and thickness measurements of deposited films, shown in Figure 4.3, were done with a profilometer at 3 different points over a scan length of 500 µm. The average thickness across the entire area of the large area electrode was t = 259. 3 ± 11.3 nm being the thickness per layer tlayer = 11.27 ± 0.88 nm. Figure 4.3: Thickness measurements at 24 different locations across the entire area (100 cm x 15 cm) of the MWNTCOOH electrode and average thickness. The error bars are twice the standard deviation of the three point thickness measurements. The SDE proved to be capable of producing electrodes with controlled thickness and mass load. Figure 4.4 shows laboratory scale (2.5 cm x 2.5 cm) MWNTCOOH electrodes spray deposited onto ITO coated glass substrates as explained in chapter 3, the spray deposition of these area films was done by attaching the substrates to flat surfaces of the cylinder of the SDE maintained at a static position while the linear drive reciprocated, and therefore for this case a layer is defined as the film deposited by a single traverse of the linear 122 drive -. The number of layers can be precisely controlled from a few layers up to several hundreds. An almost transparent electrode was produced for 25 spray-deposited layers (28 nm) whereas an electrode of 1 µm thickness was obtained for 900 spray-deposited layers. Figure 4.4: Optical image of spray-deposited MWNTCOOH electrodes (2.5 cm x 2.5 cm) onto ITO coated glass substrates with an increasing number of spray-deposited layers and thickness. Numbers describe the number of layers and the thickness (nm) is specified in brackets. 4.3.4 Surface morphology and electrochemical properties The surface morphology of the large area MWNTCOOH electrode was examined by SEM at 9 different locations uniformly distributed across the entire area. As shown in Figure 4.5, a well-entangled structure with a high degree of mesoporosity and macroporosity that facilitates ion transport repeated across the entire area. The electrochemical properties of the large area electrode were investigated at 7 different locations uniformly distributed across the entire area. ITO coated glass substrates (2.5 cm x 5.0 cm) were attached to desired points of analysis prior to spray-deposition of the large area electrode. Electrochemical experiments were then performed on each of these electrodes 123 Figure 4.5: Scanning electron microscopy images of spray-deposited carboxylfunctionalized MWNTs at 9 different points of the large area electrode in Figure 4.2. (MWNTCOOH study electrodes). As shown in Figure 4.6a, cyclic voltammograms were obtained at three different scan rates: 10, 50 and 100 mV s−1 . The voltammograms showed a characteristic nearly rectangular shape indicative of double layer charge storage but distorted by redox peaks associated with the redox activity of the carboxylic groups. The ITO coated glass substrate was tested on its own under the same conditions showing nule charge storage (not shown). Figure 4.6b shows the specific capacitance calculated from cyclic voltammograms according to, Cs = qa + |qc | 2m∆V (4.1) where qa and qc are the anodic and cathodic charges respectively, m is the electrode mass, and V is the electrochemical window. The specific capaci124 tance was approximately constant across the different analysis points with mean specific capacitances of: 21.8 ± 1.9, 18.6 ± 1.8, and 16.5 ± 1.7 F g−1 for scan rates 10, 50, and 100 mV s−1 respectively. Figure 4.6: Electrochemical characterization at different analysis points on the MWNTCOOH large area electrode in 0.1M H2 SO4 : (a) cyclic voltammograms of a single study electrode at scan rates 10, 50, and 100 mV s−1 , and (b) specific capacitance at the different analysis points. Dashed lines joining data points are for guidance to the eye, the lenght of error bars is twice the data standard deviation. Further investigation of electrochemical properties included Electrochemical Impedance Spectroscopy (EIS). The Nyquist plots in Figure 6.9 show a uniform impedance behaviour across the different analysis points in the thin MWNTCOOH large area electrode, and the absence of a Nyquist ‘knee’ in the high frequency region providing evidence of a low charge transfer resistance of the thin film electrodes. In order to study variation of electrochemical properties with electrode 125 Figure 4.7: Nyquist plots at different analysis points of the MWNTCOOH large area electrode performed at the MWNTCOOH study electrodes in 0.1M H2 SO4 . thickness, MWNTCOOH study electrodes (2.5 cm x 2.5 cm) of an increasing number of spray-deposited layers (from 200 to 1500 layers) onto ITO coated glass substrates were investigated by cyclic voltammetry. As shown in Figure 4.8, the stored charge and geometric capacitance linearly increased as the number of sprayed-deposited layers and thus mass increased. The geometric capacitance may eventually reach a plateau upon increase of thickness at the limit where eletrolyte ions can not further reach electrochemically active sites of the electrode with a given porosity and at a given scan rate. Any additional layers of electrochemically active material will only increase the electrode mass without any additional charge storage increase resulting in a decreasing specific capacitance. The mass load limit for maximum charge storage should be intrinsic to the properties of a specific electroactive material such as surface morphology, electrical conductivity, and specific surface area. However, for the material studied here, this point was not reached and can not be safely predicted a priori. 126 Figure 4.8: Cyclic voltammetry analysis of MWNTCOOH electrodes with an increasing number of spray-deposited layers: plots of (a) integrated charge stored, and (b) geometric capacitance, both at scan rate of 10 mV s−1 . 4.4 Conclusions A simple, room temperature, aqueous based, and scalable processing route has been demonstrated for the manufacture of binder-free, flexible, large area supercapacitor electrodes. The optimization of spray variables, and the spray deposition model previously developed were effective on guiding the manufacture of large area MWNTCOOH electrodes of acceptable uniformity in thickness, surface morphology, and electrochemical properties. Accurate control of thickness and mass load allowed for the investigation of the variation of stored charge and geometric capacitance simply by controlling the 127 number of sprayed layers. 128 Bibliography [1] B. Ballesteros, G. Tobias, L. Shao, E. Pellicer, J. Nogues, E. Mendoza and M. Green, Small, 2008, 4, 1501–1506. [2] G. Tobias, S. Lidong, C. Salzmann, Y. Huh and M. Green, J. Phys. Chem., 2006, 110, 22318–22322. [3] H. Ago, T. Kugler, F. Cacialli, W. R. Salaneck, M. S. P. Shaffer, A. H. Windle and R. Friend, J. Phys. Chem. B, 1999, 103, 8116–8121. [4] S. W. Lee, B.-S. Kim, S. Chen, Y. Shao-Horn and P. T. Hammond, J. Am. Chem. Soc., 2008, 131, 671–679. [5] E. Frackowiak and F. Béguin, Carbon, 2001, 39, 937–950. [6] H. A. Andreas and B. E. Conway, Electrochim. Acta, 2006, 51, 6510–6520. 129 Chapter 5 Investigation of nanostructured thin film α-MoO3 based supercapacitor electrodes in an aqueous electrolyte 130 5.1 Introduction Supercapacitors can store energy by two mechanisms: double layer capacitance and pseudocapacitance. Transition metal oxides can be induced to exhibit pseudocapacitive behaviour and ruthenium oxide is the most widely explored due to its high theoretical and practically achievable pseduocapacitance [1–4] . However, its high cost and toxicity has prevented its widespread use and many alternative transition metal oxides have also been studied [5, 6] including manganese oxide [7–19] , vanadium oxide [20–22] , iron oxides [23–28] , and molybdenum oxides [29–34] . Of these, manganese oxide is the most studied and promising, although its performance is inferior to that of ruthenium oxide. Molybdenum is a metal with a wide range of oxidation states from +2 to + 7 existing in a variety of oxides and other compounds [35, 36] , of which MoO3 and MoO2 have been of interest for supercapacitor applications due to a rich electrochemical activity, low cost, and environmentally friendly nature [29–34] . α-MoO3 possess a layered orthorhombic crystal structure where each layer is linked to an adjacent layer by van der Waals forces along the [010] direction. Each layer is composed of two MoO6 octahedron nets where octahedrons share O-O edges along the [001] direction and corners along the [100] direction [37, 38] . α-MoO3 thus has the attraction not only of multiple oxidation states but also its potential for ion intercalation in between the crystal layers. However, α-MoO3 is a semiconductor, the kinetics of faradaic charge transfer is slow and it has generally shown poor cycling behaviour in Li-ion batttery applications [39–41] . Careful selection of suitable nanostructure, dimensions, and crystal ori- 131 entation, especially if intercalation (and associated strains) is present, can be used to improve the overall energy density and the cyclability of transition metal oxides supercapacitor electrodes [27, 42] . Nanostructuring of α-MoO3 increases its electroactive surface area and shortens diffusion lengths to promote faster ion transport. Nanostructured α-MoO3 and MoO2 have been investigated for their application as supercapacitor electrodes in half cell arrangements in neutral (Na2 SO4 ), basic (NaOH), and acidic electrolytes (H2 SO4 ) [29–34] . Reports to date have described a variety of cyclic voltammetry (CV) responses in different electrochemical windows where attention has been paid mainly to the total capacitance achieved with less emphasis on extended cycling performance and elucidation of the underpinning reaction mechanisms. Reports of MoO3 and MoO2 tested in Na2 SO4 suggested both a double layer capacitive and a Na+ intercalation pseudocapacitive contribution to overall capacitance but with no attention paid to the cause the observed irreversibility [30, 31] . Shakir et al. [33] proposed several charge storage mechanisms for α-MoO3 tested in 1 M H2 SO4 : (1) faradaic-like processes involving proton (H+ ) and coupled electron insertion in a similar fashion to that found in RuO2 [43, 44] ; (2) proton adsorption/desorption; and (3) intercalation of ions into the layered structure of MoO3 . However, there was no compelling experimental evidence providing an insight into one or more of the proposed mechanisms. Similarly, Farsi et al. [29] proposed a combination of a double layer capacitance combined with electron/proton insertion (mechanism 1) for electrodeposited molybdenum oxides tested in a mix of Na2 SO4 and H2 SO4 but in a different electrochemical window than suggested by Shakir et al. [33] and with no justification of 132 their choice of electrolyte and electrochemical window. The multiple oxidation states of Mo provide MoO3 with a rich redox activity in acidic electrolytes [36, 45] . At the electrode surface, Mo (6+) is reduced to Mo (5+) in 1 M H2 SO4 . Pentavalent molybdenum oxide is a unstable compound with a characteristic blue coloration [46] . Stabilization of its crystal structure has been found to occur through the creation of oxygen vacancies followed by crystal relaxation [47, 48] rather than the formation of molybdenum bronzes [49–52] . The reduction of Mo (5+) to Mo (4+) has been described as occurring through an intermediate unknown species by a combination of chemical and electrochemical reactions [47] . Overall the attractive richness of Mo oxidation states creates difficulties in elucidating the underlying mechanisms of Mo based oxide electrochemical activity in supercapacitor applications, and significant uncertainties remain. In this work α-MoO3 nanobelts with a crystalline nanostructure were synthesized using a hydrothermal method, and binder-free, nanometer-thick electrodes with enhanced mechanical stability were fabricated using a spray deposition technique. The α-MoO3 nanobelts electrodes were tested in several electrolytes finding an enhanced charge storage in 1 M H2 SO4 where a series of cathodic and anodic peaks underscored the complex redox activity described above. Insights into the redox activity were gained by using X-ray photoelectron spectroscopy (XPS) combined with electrochemical characterization methods, and led to an improvement of the performance of α-MoO3 as a supercapacitor electrode. 133 5.2 Experimental Materials. Elicarb single walled carbon nanotubes (SWNTs) were supplied by Thomas Swan and Co. Ltd (UK); Mo powder (99.9 %), hydrogen peroxide (30 %), nitric acid (69 % ) and hydrochloric acid (37 % ) by Sigma Aldrich (UK); Poly (ethyleneimine) (PEI, Mw= 70,000 ) by Alfa-Aesar (UK); and indium tin oxide (ITO) coated glass substrates (7 Ω/sq sheet resistance) by Delta Technologies (USA). Deionized water (10 MΩ · cm) was used for all processing. α-MoO3 nanobelt synthesis. α-MoO3 nanobelts were synthesized by a hydrothermal method previously described by Hu et al. [53] . A colloidal suspension of peroxomolybdic acid was obtained by gradually adding Mo powder (4g, 41.7 mmol) to H2 O2 (30 %, 40 ml) under gentle agitation and ice cooling. The colloidal suspension was then transferred to a teflon lined autoclave and maintained at 180 ◦ C for 48 hours. Subsequently the autoclave was cooled down to room temperature and the solid product was extensively rinsed with deionized water while vacuum filtrating. The solid was then collected and dried at 60 ◦ C in vacuum oven. SWNT processing. SWNTs were used as a conductive additive to improve charge transfer to active sites of the α-MoO3 nanobelts. SWNTs were steam purified as described elsewhere [54, 55] and subsequently carboxylfunctionalized (SWNTCOOH) by refluxing in HNO3 (69 %) for 30 hours at 100 ◦ C followed by filtration and extensive rinsing with deionized water until neutral pH. α-MoO3 nanobelt aqueous suspensions. An aqueous suspension of α- 134 MoO3 (0.1-0.5 mg ml−1 ) nanobelts was produced by ultrasonication (100 W, 30 kHz , small tip) for 5 minutes. α-MoO3 nanobelt/ SWNTCOOH aqueous suspensions. An aqueous suspension of α-MoO3 (1.5 mg ml−1 ) nanobelts was produced by ultrasonication (100 W, 30 kHz , small tip) for 5 minutes. Similarly, an aqueous suspension of SWNTCOOH (0.5 mg ml−1 ) was produced by ultrasonication (600 W, 20 kHz probe) for 12 minutes while maintaining ice cooling. Mixing followed by vigorous vortexing (10 min) of the α-MoO3 nanobelts suspension (30 ml) with the SWNTCOOH (30 ml) suspension yielded a 25 %/75 % w/w SWNTCOOH/α-MoO3 nanobelts composite (MOSC). Electrode manufacture. Electrodes of 1 cm2 area were spray-deposited onto ITO coated glass substrates by a method described elsewhere [56, 57] . Briefly, suspensions of electrochemically active material were fed into a industrial spray head where an atomizing air flow produced a suspension mist that was deposited onto a substrate maintained at a temperature suitable for the immediate volatilization of the fugitive carrier liquid (water). The spray head was moved at a constant spray height and speed in a single direction above the target substrates to produce films of uniform thickness over the entire substrate area. Prior to the spray deposition of electroactive material, substrates were pre-coated with a ≈ 5 nm layer of 0.1 % w/w PEI solution to improve adhesion of the electroactive material to the substrate. Films of α-MoO3 nanobelts were manufactured by spray deposition of 150 ml of a 0.13 mg ml−1 aqueous supension. Films of MOSC were manufactured by spray deposition of 60 ml of 1 mg ml−1 MOSC suspension. The average mass and thickness of α-MoO3 nanobelts electrodes were 0.05-0.15 mg, and 500 135 nm respectively. The average mass and thickness of a MOSC electrode were 0.45 mg and 1.72 µm respectively. Equipment and characterization techniques. Transmission electron microscopy (TEM) images were obtained with a JEOL 2010 operated at 200 kV; scanning electron microscopy (SEM) images were obtained in a field emission FESEM JEOL 840 F operated at 5 kV and a 15 mm working distance; X-ray diffraction (XRD) was performed in a Siemens D5000 powder diffractometer equipped with a monochromatic Cu Kα radiation source (λ = 0.15406nm) and a secondary monochromator. XRD patterns were collected between 5 ◦ < 2θ < 75 ◦ , with a step size of 2θ= 0.05 ◦ and a count time of 12 s/step. X-ray photoelectron spectroscopy (XPS) was performed in a Kratos Axis Ultra spectrometer equipped with a monochromatic Al Kα (1486.6 eV) as the source and with a spot size 0.7 x 0.3 mm2 . The C1s (sp2 ) binding energy (284.9 eV) was used as reference for XPS spectra calibration. The XPS spectra were analyzed and fitted using CasaXPS software. A mixture of Gaussian and Lorentzian functions was used for the least-squares curve fitting procedure. Ultrasonication was performed in a ultrasonicator UP100H, Hielscher (100W, 30 kHz), and a Sonics Vibra cell VC-600-2 probe (600 W, 20 kHz). Electrode thickness was determined by step height measurements using a Dektak 6M profilometer (Veeco Instruments, Inc) and the weight of deposited films was measured using a Sartorius micro balance with 0.01 mg readability. Electrochemical characterization. α-MoO3 based thin film electrodes were tested in a three-electrode electrochemical cell configuration using a Reference 600/EIS300 Gamry potentiostat/galvanostat, Ag/AgCl electrode as 136 reference, a platinum sheet as counter electrode, and 1 M H2 SO4 as the electrolyte (initial tests also in 1 M Na2 SO4 , K2 SO4 , and Li2 SO4 aqueous solutions). Cyclic voltammetry experiments were performed in a potential range from 0 to 1 V vs Ag/AgCl. 5.3 Results and Discussion 5.3.1 Material characterization Figure 5.1a shows the X-ray diffraction spectrum of the as prepared material. The indexing of the distinct and sharp peaks identified highly crystalline and pure α-MoO3 phase with an orthorhombic layered crystal structure (ICDDJCPDS card No. 05-0508). The strong diffraction peaks from planes (0k0) showed a characteristic anisotropic growth of the α-MoO3 phase [38, 58] . Figure 5.1b shows corresponding SEM images of the α-MoO3 indicating largely a 1D elongated morphology with average dimensions of 300-400 nm width and 8-18 µm length. Because the width of the α-MoO3 nanostructures was larger than the thickness (a rectangular cross section rather than a round one), the nanocrystals were designated as nanobelts rather than nanorods [38, 59] . A TEM image and corresponding selected area electron diffraction (SAED) pattern of an individual α-MoO3 nanobelt are shown in Figure 7.1a and Figure 7.1b respectively. The SAED pattern was identified as the [010] zone of α-MoO3 and shows that the growth of the nanobelts was primarily along the [001] direction. The growth of α-MoO3 along the [001] direction is considered to be dictated by its highly anisotropical structure where octahedra 137 share edges along the [001] direction having strong covalent bonds thermodynamically more stable (larger energy expenditure to break those bonds) than the van der Waals forces along the [010] direction, or weaker covalent bonds between corner-sharing octahedrons along the [100] direction [38] . The HRTEM image in Figure 7.1c indicated interplanar distances of 0.4 and 0.36 nm for the (100) and (001) lattice planes respectively. As shown in Figure 5.1b, the combination of sonication followed by spray deposition allowed for the manufacture of electrodes consisting of a highly interconnected open network of α-MoO3 nanobelts. The surface area of the α-MoO3 was determined as 13.7 m2 g−1 by nitrogen adsorption isotherms using the the Brunauer-Emmett-Teller (BET) method. Figure 5.1: (a) XRD spectrum, and (b) SEM image of as made α-MoO3 nanobelts. 138 Figure 5.2: (a) TEM image of a single α-MoO3 nanobelt, and (b) corresponding SAED pattern. (c) HRTEM image of enclosed nanobelt area showing interplanar distances along the [001] and [100] directions. 5.3.2 Cyclic voltammetry Figure 5.3 shows a typical cyclic voltammogram (CV) of a spray-deposited α-MoO3 electrode in 1 M H2 SO4 at 10 mV s−1 , where a capacitance of 12.7 F g−1 was obtained. In the following, all the potentials will be given vs Ag/AgCl in saturated KCl reference electrode. There was a complex electrochemical activity with several anodic (a) and cathodic (c) peaks: a1 (487.3 mV), a2 (391.2 mV), a3 (242 mV), a4 (111.6 mV); and c1 (454.1 mV), c2 (354.7 mV), c3 (195.2 mV). Sulphuric acid was selected here as the electrolyte that provided the larger charge storage after electrodes were tested in other electrolytes including Na2 SO4 , K2 SO4 , and Li2 SO4 in which case, there was only double layer capacitance (< 5 F g−1 ) in the electrochemical window of 0 to 0.8 V as shown in Figure 5.4. A negative electrochemical window rendered either a lower charge storage and/or increased irreversibility (see section 5.3.4). 139 Figure 5.3: Representative cyclic voltammogram of an α-MoO3 electrode in 1 M H2 SO4 at a scan rate of 10 mV s−1 . Labels identify anodic (a) and cathodic (c) redox peaks. Figure 5.4: Representative cyclic voltammograms of an α-MoO3 electrode in neutral aqueous electrolytes at a scan rate of 10 mV s−1 . 5.3.3 XPS surface characterization of the electrodes In order to investigate further the electrochemical activity of α-MoO3 electrodes in 1 M H2 SO4 in the 0 to 1 V electrochemical window (vs Ag/AgCl), XPS studies were performed in conjunction with cyclic voltammetry and chronoamperometry. This method was already successful to elucidate charge storage mechanism in MnO2 thin film electrodes for electrochemical capacitors [16] . Prior to XPS measurements, electrodes were subjected to polar- 140 ization: cyclic voltammetry was performed at 20 mV s−1 from open circuit potential (OCP) up to the polarization potential 0 ≤ E ≤ 1 V (either in reduction or oxidation) at which point the electrode was held at constant potential via chronoamperometry until the current (I )-time(t) response reached a nearly steady state with I ≈ 0. The chosen polarization potentials E (vs Ag/AgCl) are given in Table 5.1. A different sample was used for each polarization potential. The time in between polarizations and XPS measurements was approximately 2-4 hours. Figure 5.5 shows the XPS spectrum in the Mo 3d binding energy region for an as prepared sample of α-MoO3 nanobelts with two well resolved spectral lines which after curve fitting were identified as the characteristic Mo 3d spinorbit doublet peaks of MoO3 : Mo 3d5/2 6+ at 232.6 eV and Mo 3d3/2 6+ at 235.7 eV [60, 61] . This spectrum confirmed the purity of the α-MoO3 phase of as-manufactured electrodes. It also provided a useful baseline against which the effect of various polarizations could be assessed. Figure 5.5: XPS spectrum in the Mo 3d binding energy region of as made α-MoO3 nanobelts. For the α-MoO3 electrodes then polarized at the different potentials E given in Table 5.1, curve fitting of XPS spectrum in the Mo 3d binding energy 141 region suggested the presence of three spin-orbit doublets corresponding to three species. Figure 5.6 shows the Mo 3d XPS spectra for α-MoO3 electrodes polarized at E = 1 V and E = 0 V. For the α-MoO3 electrode polarized at E = 1 V, spin-orbit components were assigned as: 230.0 eV (Mo 3d5/2 4+), 231.2 eV (Mo 3d5/2 5+), 231.8 eV (Mo 3d5/2 6+), 233.0 eV Mo 3d3/2 4+), 234.3 eV (Mo 3d3/2 5+) and 235.1 eV (Mo 3d3/2 6+). For the α-MoO3 electrode polarized at E = 0 V, spin-orbit components were assigned as: 229.7 eV (Mo 3d5/2 4+), 231.1 eV (Mo 3d5/2 5+), 231.5 eV (Mo 3d5/2 6+), 232.8 eV (Mo 3d3/2 4+), 234.3 eV (Mo 3d3/2 5+) and 235.6 eV (Mo 3d3/2 6+). For αMoO3 electrodes polarized at each of the other potentials, the Mo 3d XPS spectrum showed spin-orbit components at similar binding energies and are summarized in Table 5.1. Table 5.1: XPS binding energies (eV) for α-MoO3 electrodes polarized at various potentials E . R and O indicate polarization in reduction and oxidation respectively. E (V vs Ag/AgCl) Mo 3d 5/2 (4+) Mo 3d 5/2 (5+) Mo 3d 5/2 (6+) Mo 3d 3/2 (4+) Mo 3d 3/2 (5+) Mo 3d 3/2 (6+) O1s O1s O1s 0R 229.72 231.11 231.51 232.81 234.30 235.59 528.54 529.70 531.08 0.13R 229.78 231.12 231.66 232.85 234.36 235.92 528.52 529.59 531.12 0.185R 229.84 231.21 232.38 232.96 234.37 235.47 528.67 529.55 531.16 0.26R 229.87 231.06 231.77 233.00 234.18 234.73 528.89 529.62 531.40 0.415R 230.07 231.23 231.94 233.12 234.35 234.85 529.03 529.58 531.10 0.3O 230.18 231.23 231.72 233.18 234.35 234.80 529.10 529.80 531.37 0.46O 230.01 231.15 231.75 233.13 234.30 234.92 529.02 529.60 531.22 The percentage concentration of the different Mo oxidation states were estimated by integration of the area under the Mo 3d5/2 and Mo 3d3/2 peaks separately. There were similar trends for both Mo 3d spin-orbit compo- 142 1O 230.01 231.21 231.87 233.10 234.37 235.13 529.03 529.73 531.43 Figure 5.6: XPS spectra in the Mo 3d binding energy region of α-MoO3 electrodes polarized at (a) E= 1 V, and (b) E=0 V. Spin-orbit components are shown in green for Mo 3d (6+), red for Mo 3d (5+), and blue for Mo 3d (4+). nents. Figure 5.7 shows data for the Mo 3d5/2 spin-orbit component where Mo(4+)/Mo(5+)/Mo(6+) concentration percentages were 11.0 ± 3.0 %/77.2 ± 3.0 %/11.8 ± 3.0 %, and 41.3 ± 3.0 % /53.4 ± 3.0 %/ 5.2 ± 3.0 % for E = 1 V and E = 0 V, respectively. At E = 1 V, the major species was Mo (5+), whereas at E =0 V a decrease of concentration of Mo (5+) was accompanied with an increase of the concentration of Mo (4+). The increase of concentration of Mo (4+) coupled to a reduction of Mo (5+) was a trend 143 observed for polarization potentials below E = 0.26 V and clearly enhanced for potentials E =0.185 V, E = 0.13, and E = 0.0 V (all these potentials vs Ag/AgCl reference electrode). As shown in Figure 5.6, a shift by 0.3 V of the XPS Mo 3d spectra of the electrode held at 0 V provided further supporting evidence of reduction to Mo (4+). Figure 5.7: Plot showing the percentage concentration of Mo (x+) at the surface of polarized α-MoO3 electrodes versus polarization potentials E given in Table 5.1. Dashed lines joining the data points are for visual guide. The percentage concentration of each Mo (x+) species has been calculated from the area integration of the corresponding Mo 3d5/2 x+ XPS spin-orbit component. The electrodes have been scanned from open circuit potential OCP, shown for reference as a dashed black line, up to polarization potential E either in reduction (R, circles) or oxidation (O, squares). Error bars show a ± 3 % error. The limits of a selected electrochemical window (0.26-0.43 V) are indicated by black solid lines. Figure 5.8 shows the O 1s XPS spectra for electrodes held at E = 1 V and E = 0 V. Curve fitting showed three species with binding energy peaks at 529.0, 529.7, and 531.4 eV for the electrode held at E =1 V, and at 528.5, 529.7, and 531.0 eV for the electrode held at E = 0 V. The first two peaks were assigned to Mo-O bonds, and the third peak at 531 ± 0.5 eV was assigned 144 to -OH groups [62] . There were similar binding energy peaks for electrodes polarized at other potentials with the main binding energy peak at 529.6 ± 0.3 eV. It is known that MoO3 and MoO2 have similar O 1s binding energies [60] and therefore curve fitting for qualification of each MoOx species present at the electrode surface, as suggested by the Mo 3d XPS spectra, was particularly difficult. In the case of MoO2 , a shift of 0.5 eV towards lower energies for electrodes polarized at E = 0 V, E = 0.13 V and E = 0.185 V provided further evidence of the presence of this oxide at such polarization potentials [63] . Figure 5.8: XPS spectra in the O 1s binding energy region of α-MoO3 electrodes polarized at (a) E= 1 V, and (b) E=0 V. In summary, spectro-electrochemical experiments showed that during cyclic voltammetry in 1 M H2 SO4 from 0 to 1 V, MoO3 was partially reduced to 145 a mix of oxides of lower oxidation states with relative electrode surface concentrations: Mo (5+) > Mo (4+) > Mo (6+). There was a reduction to Mo (4+) at the cathodic peaks at E = 0.185 V and E =0 V and this is in agreement with data provided by the Pourbaix diagram [45] for the Mo-water system at pH=0 where the prevalent Mo compound at 0.32 V (vs NHE)/0.12 V (vs Ag/AgCl) is MoO2 . Earlier investigations based on cyclic voltammetry at 5 mV s−1 determined a reduction of Mo (6+) to Mo (5+) at -240 mV vs Hg/Hg2 SO4 (237 mV vs Ag/AgCl) with the appearance of a dark blue coloration attributed to an oxygen-deficient oxide [47, 48] : 2M oO3 + 2H + + 2e− → M o2 O5 + H2 0 (5.1) It has also been proposed that direct reduction of Mo(5+) to Mo (4+) (brown) was not possible and that the existence of an intermediate species, likely to be Mo (3+) (also dark blue), occurred at -400 mV vs Hg/Hg2 SO4 (77 mV vs Ag/AgCl) with oxidation to MoO2 at -475 mV vs Hg/Hg2 SO4 (2 mV vs Ag/AgCl) [47, 48] . The detection of MoO2 by XPS along with the observation of a dark blue coloration in electrodes polarized at potentials E = 0 V, E = 0.13 V and E = 0.185 V in the present work, are supportive of these oxido-reduction processes [47, 48] . In the case of Mo (6+), there was a low concentration approximately constant at all polarization potentials which may indicate that there was an underlying layer of Mo (6+) in the electrode, still detectable by XPS, that remained electrochemically inactive. The formation of hydrogen molybdenum bronzes Hx MoO3 - leading to dissolution in 0.5 M H2 SO4 - has been reported within a 80 mV to -555 146 mV (vs Ag/AgCl) electrochemical window with the appearance of four different phases at scan rates below 2 mV s−1 , above which the associated redox peaks disappeared [49–51, 64] . Since the scan rate here used was 20 mV s−1 along with a 0 to 1 V (vs Ag/AgCl) electrochemical window, the formation of even phase II (H1.04 MoO3 ) at 80 mV (vs Ag/AgCl) was unlikely. Further spectro-electrochemical studies will be needed to establish a complete reaction mechanism unambiguously, but the current work provides considerable support for the oxido-reduction processes previously reported [47, 48] . 5.3.4 Optimization of electrochemical window and cycling behaviour On the basis of the XPS studies, optimization of the electrochemical window for improved reversibility was carried out. As shown in Figure 5.7, potentials in the range 0 V < E < 0.26 V lead to irreversible reduction to MoO2 . An electrode tested by cyclic voltammetry in the potential range 0.26 V < E < 0.8 V showed a progressive decrease in charge storage capacity during 150 cycles as shown in Figure 5.9. The optimum reversibility - rather than absolute charge storage capacity on the first cycle - was achieved in the electrochemical window 0.26 V < E < 0.43 V involving redox peaks c2 and a2 . Figure 5.10 shows a cyclic voltammogram of an α-MoO3 electrode swept within this electrochemical window with near constant capacitance and null charge storage capacity degradation after 720 cycles. A maximum capacitance of 8.8 F g−1 was obtained which is equivalent to 64 µF cm−2 when normalized per surface area (as measured by the BET method). A maximum capacitance of 8.8 F 147 g−1 was obtained which is equivalent to 64 µF cm−2 when normalizing per surface area ( 13.7 m2 g−1 as measured by the BET method) or 440 µF cm−2 when normalizing per geometric area of 1 cm2 (mass of the electrode 0.05 mg). Figure 5.9: Cyclic voltammogram of an α-MoO3 electrode in 1 M H2 SO4 at a scan rate of 20 mV s−1 in a 0.26-0.8 V electrochemical window showing cycles 1, 2, 5, 10, 20, 50, 100, and 150. The small arrows show the scanning direction and the large arrows show the peak current decrease upon cycling. Cyclic voltammetry in the optimized electrochemical window at increasing scan rates allowed for further understanding of the electrochemical activity. Figure 5.11a shows a linear dependence of the currents associated with anodic and cathodic peaks with the square root of the scan rate describing a process dominated by linear diffusion. If the electrochemical reactions were under diffusion control, then it was likely that less accessible reaction sites in the interior of the electrode remained electrochemically inactive, and this may further support the earlier XPS findings of persistent non-reduced MoO3 . The reversibility of the system was evaluated further considering the peak- 148 Figure 5.10: (a) Cyclic voltammogram of an α-MoO3 electrode in 1 M H2 SO4 at a scan rate of 20 mV s−1 showing cycles 1, 2, 100, 200, 500 and 720, (b) capacitance versus cycle number. The arrows indicate the scanning direction. to-peak potential difference of anodic and cathodic current peaks, ∆Epp , as function of the logarithm of scan rate. In an ideally reversible system ∆Epp ≈ 57 mV and is independent of scan rate [65] . As shown in Figure 5.11b ∆Epp < 57 mV for all scan rates and was approximately constant (≈ 20 mV) for scan rates ≤ 10 mV s−1 . Since MoO3 is a n-type semiconductor, the rate of electron transport kinetics (measured by the standard electrochemical rate constant k◦ ) is limited and thus the rate of mass transport mtrans ∝ ν 1/2 , where ν is the scan rate, becomes dominant at low scan rates leading to irreversibilities - in a reversible process k◦ /mtrans ≥ 15 [65] . However, a stable 149 charge storage capacity at 20 mV s−1 for 720 cycles, as demonstrated above, was indicative of a quasi-reversible behavior at this scan rate. Figure 5.11: Analysis of cyclic voltammograms in the optimized electrochemical window 0.26 to 0.43 V: (a) anodic (Ia ) and cathodic (Ic ) peak currents as function of the square root of scan rate ν,(b) peak-to-peak potential ∆Epp as function of the logarithm of the scan rate. Previous work reported electrodeposited molybdenum oxide tested by cyclic voltammetry in a 0.005 M H2 SO4 + 0.095 M Na2 SO4 electrolyte in a -0.55 to 0.0 V vs Ag/AgCl electrochemical window where the charge storage mechanism suggested was a combination of redox pseudocapacitance and double layer capacitance [29] . As shown in Figure 5.12, testing of an α-MoO3 electrode in the same electrochemical window and the same electrolyte showed a comparatively high capacitance obtained nevertheless at the expense of an obvious irreversible behaviour and consequently a capacitance degradation after only 60 cycles. 150 Figure 5.12: (a) Cyclic voltammogram of an α-MoO3 electrode in 0.005 M H2 SO4 + 0.095 M Na2 SO4 at a scan rate of 5 mV s−1 showing cycles 2, 5, 10, 15, 20, 30, 40, 50 and 60. The current increases from cycles 2 to 20 (dashed lines), and decreases from cycles 20 to 60 (solid lines), (b) capacitance versus cycle number. The arrows indicate the scanning direction. Finally, in order to improve electron charge transfer to active sites of α-MoO3 nanobelts and to increase capacitance, SWNTs were used as conductive additive. Figure 5.13 shows cyclic voltammograms (at 20 mV s−1 and a 0.26 to 0.43 V electrochemical window) for a SWNTs/α-MoO3 25 %/75 % w/w composite electrode where the capacitance was 31.5 F g−1 (normalized per unit mass of MoO3 ) or 23.6 F g−1 (normalized per unit mass of electrode where the contributions of MoO3 and the SWNTs were 14.8 F g−1 x 0.75= 11.1 F g−1 and 50 F g−1 x 0.25= 12.5 F g−1 respectively), and remained stable for 100 cycles. The redox activity observed is likely to have contributions from both, the redox activity of α-MoO3 and the redox activity of carboxylic groups on the surface of SWNTs. 151 Figure 5.13: (a) Cyclic voltammogram of a SWNTs/α-MoO3 25/75 % w/w composite electrode in 1 M H2 SO4 at a scan rate of 20 mV s−1 showing cycles 5, 10, 20...100, (b) capacitance of the composite electrode per mass of MoO3 versus cycle number. 5.4 Conclusions The suitability of α-MoO3 nanobelts for supercapacitor applications was examined in various aqueous electrolytes, finding the greatest charge storage and a rich redox activity in 1 M H2 SO4 . A combination of XPS with various electrochemical characterization methods revealed a partial reduction of MoO3 to a mix of lower valence oxides with concentrations varying as Mo (5+) > Mo (4+) > Mo (6+) in a 0 to 1 V (vs Ag/AgCl) electrochemical window, and with MoO2 as the species found at potentials below 0.185 V (vs Ag/AgCl). In the light of the redox changes occurring, the electrochemical 152 window was optimized for enhanced reversibility obtaining a stable capacitance of 8.8 F g−1 (64 µF cm−2 ) that was maintained up to 720 cycles. Thus for using α-MoO3 as an electrochemical capacitor device, the potential window has to be drastically decreased at the expense of the absolute capacitance. Similarly, opening the potential window increased the capacitance but drastically lowered the cycling ability of the electrode. Adding SWNTs as conductive additive slightly improved the charge storage of αMoO3 nanobelts from 8.8 F g−1 to 14.8 F g−1 . 153 Bibliography [1] H. Kim and B. N. Popov, J. Power Sources, 2002, 104, 52–61. [2] K.-H. Chang and C.-C. Hu, J. Electrochem. Soc., 2004, 151, A958–A964. [3] P. Soudan, J. Gaudet, D. Guay, D. Bélanger and R. Schulz, Chem. Mater., 2002, 14, 1210–1215. [4] K. M. Lin, K. Chang, C. C. Hu and Y. Y. Li, Electrochim. Acta, 2009, 54, 4574–4581. [5] P. Simon and Y. Gogotsi, Nat. Mater., 2008, 7, 845–854. [6] T. Cottineau, M. Toupin, T. Delahaye, T. Brousse and D. Bélanger, Appl. Phys. A-Mater., 2005, 82, 599–606. [7] H. Y. Lee and J. B. Goodenough, J. Solid State Chem., 1999, 144, 220– 223. [8] S.-C. Pang, M. A. Anderson and T. W. Chapman, J. Electrochem. Soc., 2000, 147, 444–450. [9] H. Y. Lee, S. W. Kim and H. Y. Lee, Electrochem. Solid-State Lett., 2001, 4, A19–A22. 154 [10] C.-C. Hu and T.-W. Tsou, Electrochem. Commun., 2002, 4, 105–109. [11] S.-F. Chin, S.-C. Pang and M. A. Anderson, J. Electrochem. Soc., 2002, 149, A379–A384. [12] J. Jiang and A. Kucernak, Electrochim. Acta, 2002, 47, 2381–2386. [13] C.-C. Hu and T.-W. Tsou, Electrochim. Acta, 2002, 47, 3523–3532. [14] J.-K. Chang and W.-T. Tsai, J. Electrochem. Soc., 2003, 150, A1333. [15] R. N. Reddy and R. G. Reddy, J. Power Sources, 2003, 124, 330–337. [16] M. Toupin, T. Brousse and D. Bélanger, Chem. Mater., 2004, 16, 3184– 3190. [17] T. Brousse, M. Toupin and D. Bélanger, J. Electrochem. Soc, 2004, 151, A614. [18] J.-M. Luo, B. Gao and X.-G. Zhang, Mater. Res. Bull., 2008, 43, 1119– 1125. [19] L. Athouel, F. Moser, R. Dugas, O. Crosnier, D. Bélanger and T. Brousse, J. Phys. Chem. C, 2008, 112, 7270–7277. [20] H. Y. Lee and J. B. Goodenough, J. Solid State Chem., 1999, 148, 81–84. [21] T. Kudo, Y. Ikeda, T. Watanabe, M. Hibino, M. Miyayama, H. Abe and K. Kajita, Solid State Ionics, 2002, 152153, 833–841. 155 [22] C. M. Huang, C. C. Hu, K. H. Chang, J. M. Li and Y. F. Li, J. Electrochem. Soc., 2009, 156, A667–A671. [23] N.-L. Wu, S.-Y. Wang, C.-Y. Han, D.-S. Wu and L.-R. Shiue, J. Power Sources, 2003, 113, 173–178. [24] N.-L. Wu, Mater. Chem. Phys., 2002, 75, 6–11. [25] S. Y. Wang and N. L. Wu, J. Appl. Electrochem., 2003, 33, 345–348. [26] Z. H. Zhou, J. Wang, X. Liu and H. S. O. Chan, J. Mater. Chem., 2001, 11, 1704–1709. [27] X. Zhao, B. Mendoza-Sánchez, P. J. Dobson and P. S. Grant, Nanoscale, 2011, 3, 839–855. [28] X. Zhao, C. Johnston and P. S. Grant, J. Mater. Chem., 2009, 19, 8755–8760. [29] H. Farsi, F. Gobal, H. Raissi and S. Moghiminia, J. Solid State Electr., 2010, 14, 643–650. [30] L. Zheng, Y. Xu, D. Jin and Y. Xie, J. Mater. Chem., 2010, 20, 7135– 7143. [31] F. Gao, L. Zhang and S. Huang, Mater. Lett., 2010, 64, 537–540. [32] W. Sugimoto, T. Ohnuma, Y. Murakami and Y. Takasu, Electrochem. Solid-State Lett., 2001, 4, A145–A147. [33] I. Shakir, M. Shahid, H. W. Yang and D. J. Kang, Electrochim. Acta, 2010, 56, 376–380. 156 [34] R. Janarthanan, K. Pilli Satyananda, V. Balasubramanian and V. Thirukkallam Kanthadai, Electrochem. Commun., 2009, 11, 572–575. [35] A. J. Bard, R. Parsons and J. J., Standard potentials in aqueous solutions, Marcel Dekker, Inc., New York, USA, 1985. [36] A. J. Bard, Encyclopedia of Electrochemistry of the elements, Marcel Dekker, Inc, 1973, vol. 5, pp. 136–221. [37] R. L. Smith and G. S. Rohrer, J. Solid State Chem., 1996, 124, 104–115. [38] S. T. Wang, Y. G. Zhang, X. C. Ma, W. Z. Wang, X. B. Li, Z. D. Zhang and Y. T. Qian, Solid State Commun., 2005, 136, 283–287. [39] J. O. , J. Heydecke and H. P. Fritz, Solid State Ionics, 1982, 6, 215–224. [40] J. O. Besenhard, J. Heydecke, E. Wudy, H. P. Fritz and W. Foag, Solid State Ionics, 1983, 8, 61–71. [41] M. E. Spahr, P. Novak, O. Haas and R. Nesper, J. Power Sources, 1995, 54, 346–351. [42] T. Brezesinski, J. Wang, S. H. Tolbert and B. Dunn, Nat. Mater., 2010, 9, 146–151. [43] S. Hadzi-Jordanov, H. Angerstein-Kozlowska, M. Vukovic and B. E. Conway, J. Electrochem. Soc., 1978, 125, 1471–1480. [44] K. M. Lin, K. Chang, C. C. Hu and Y. Y. Li, Electrochim. Acta, 2009, 54, 4574–4581. 157 [45] M. Pourbaix, Atlas of electrochemical equilibria in aqueous solutions, Section 10, Oxford, New York, Pergamon Press, 1996. [46] N. A. Dhas and A. Gedanken, J. Phys. Chem. B, 1997, 101, 9495–9503. [47] N. Anbananthan, K. Nagaraja Rao and V. K. Venkatesan, J. Electroanal. Chem., 1994, 374, 207–214. [48] N. Anbananthan, K. N. Rao and V. K. Venkatesan, Appl. Surf. Sci., 1993, 72, 189–194. [49] F. Endres and G. Schwitzgebel, J. Electroanal. Chem., 1996, 415, 23–26. [50] F. Endres and G. Schwitzgebel, Mater. Res. Bull., 1996, 31, 1537–1541. [51] F. Endres and G. Schwitzgebel, Electrochim. Acta, 1998, 43, 431–433. [52] M. Greenblatt, Chem. Rev., 1988, 88, 31–53. [53] B. Hu, L. Mai, W. Chen and F. Yang, ACS Nano, 2009, 3, 478–482. [54] B. Ballesteros, G. Tobias, L. Shao, E. Pellicer, J. Nogues, E. Mendoza and M. Green, Small, 2008, 4, 1501–1506. [55] G. Tobias, S. Lidong, C. Salzmann, Y. Huh and M. Green, J. Phys. Chem., 2006, 110, 22318–22322. [56] B. Mendoza-Sánchez, L. Minati and P. S. Grant, Fourth European Symposium on Super Capacitors and Applications, Bordeaux, France, 2010. [57] X. Zhao, B. T. T. Chu, B. Ballesteros, W. L. Wang, C. Johnston, J. M. Sykes and P. Grant, Nanotechnology, 2009, 20, 065605. 158 [58] G. S. Zakharova, C. Taschner, V. L. Volkov, I. Hellmann, R. Klingeler, A. Leonhardt and B. Buchner, Solid State Sci., 2007, 9, 1028–1032. [59] Y. H. Gao, Y. Bando and T. Sato, Applied Physics Letters, 2001, 79, 4565. [60] J. G. Choi and L. T. Thompson, Appl. Surf. Sci., 1996, 93, 143–149. [61] Z. Song, T. Cai, Z. Chang, G. Liu, J. A. Rodriguez and J. Hrbek, J. Am. Chem. Soc., 2003, 125, 8059–8066. [62] J. Torres, J. E. Alfonso and L. D. Lopez-Carreno, Physica Status Solidi C - Conference and Critical Reviews, Vol 2, No 10, Weinheim, 2005, pp. 3726–3729. [63] Y. L. Leung, P. C. Wong, K. A. R. Mitchell and K. J. Smith, Appl. Surf. Sci., 1998, 136, 147–158. [64] D. Guay, G. Tourillon, G. Laperriere and D. Bélanger, J. Phys. Chem., 1992, 96, 7718–7724. [65] R. Compton and C. Banks, Understanding Voltammetry, World Scientific Publishing Co. Pte. Ltd., Oxford, UK, 2007. 159 Chapter 6 An investigation of nanostructured thin film α-MoO3/single-walled carbon nanotube electrodes in LiClO4/propylene carbonate for supercapacitor/battery applications 160 6.1 Introduction As described in the previous chapter, α-MoO3 has a two-dimensional layered orthorhombic crystal structure where each layer is linked to an adjacent layer by van der Waals forces along the [010] direction. Each layer is composed of two MoO6 octahedron nets where octahedrons share O-O edges along the [001] direction and corners along the [100] direction [1, 2] . Because of its layered structure, α-MoO3 has been shown to be suitable for ion intercalation, and it has been largely studied for its potential application as a cathode in lithium-ion batteries [3–6] . α-MoO3 can accommodate a theoretical maximum of 6 mol Li+ / mol MoO3 [7] , and having a high open circuit voltage in the order of 3 V for Li/MoOx due to the high Mo oxidation state [4] , it can potentially provide with a high charge storage capacity. However, among the drawbacks of α-MoO3 are its semi-conducting nature, the low Li mobility in the host matrix and poor mechanical stability due to structural changes and associated strains upon repeated intercalation/de-intercalation. In order to exploit the charge storage capability of α-MoO3 micro- and nano-structuring is necessary, a small particle size: (1) provides an increased surface area available for ion intercalation - even more important in pseudocapacitive ion intercalation processes [8, 9] (see below) -, (2) diminishes pulverization effects during cycling [7, 10] , and (3) along with suitable porosity favours effective liquid phase transport [4, 5] . Pseudocapacitance is a charge storage mechanism that involves kinetically facile electron or ion transfer at the surface of electrochemically active materials with no involvement of bulk phase transformations and diffusion 161 control [11] . Unlike battery-like processes with a single electrode potential independent of the extent of reaction - described by a plateau in a discharge curve -, in a pseudocapacitive process the potential varies continuously with the degree of conversion of materials [12] . Pseudocapacitive charge storage can be of redox type involving electron transfer between oxidized and reduced species, or ion-intercalation type where ions intercalate into a host crystal. Ion intercalation into a host crystal is formally a three-dimensional (3D) event but when occurring in a layered crystal, it resembles a two-dimensional (2D) surface-based sorption processes [13, 14] ; it can then be modelled by a sorption isotherm where the chemical potential is a function of the quasi-two dimensional lattice occupancy and some terms account for guest ion-ion lateral interactions, cohesive forces between layers, and long range atom interactions [14] . Quasi-two dimensional phase separations occur in a pseudoacapacitive ion intercalation process and are described by discontinuities in a discharge curve or pseudocapacitive peaks in a cyclic voltammogram; in contrast, phase separations in a 3D battery ion intercalation process are marked by sharp phase transitions (steps in an otherwise continuous plateau) in a discharge curve. Therefore, consideration of a given ion intercalation process in a host crystal as either a pseudocapacitive or a battery-like process must observe above fundamental thermodynamic principles which ultimately dictate device performance; electrochemical capacitors are primarily high power density devices with a long cycle life whereas batteries are mainly high energy density devices with moderate power performance. In this work, the electrochemical behaviour of α-MoO3 nanobelts/carboxyl162 functionalized single-walled carbon nanotubes (SWNTCOOH) composite electrodes was investigated in a LiClO4 /propylene carbonate electrolyte. αMoO3 nanobelts with a crystalline structure were synthesized using an established hydrothermal method. SWNTCOOH were used as a conductive additive to improve utilization of the electrochemically active material; the SWNTCOOH interwove with the α-MoO3 nanobelts to provide electron conducting paths favouring charge storage. The tangled net of nanobelts produced a mesoporous electrode that favoured enhanced electrolyte access, as well as mechanical stability of the electrode. The particular contribution of the work in this chapter is to consider carefully generic issues of whether the electrochemical behaviour of the composite electrode is better suitable for electrochemical capacitors or battery applications by evaluating: (1) the contributions of capacitive and diffusion controlled mechanisms to charge storage; (2) the total charge stored as a function of scan rate, i.e. energy density and power density; and (3) cycling stability. A combination of chargedischarge experiments at low current density and electrochemical impedance spectroscopy showed some details of the charge and mass transport processes occurring during diffusion controlled 3D Li-ion intercalation. 6.2 Experimental Materials. Li wire (≥ 98 % in mineral oil), n-hexane (anhydrous 95 %), lithium perchlorate (battery grade 99.99 %), and propylene carbonate (anhydrous 99.7 %) were supplied by Sigma Aldrich (UK). For other materials see previous chapter. 163 α-MoO3 nanobelt synthesis, single-walled carbon nanotubes (SWNTs) processing, preparation of α-MoO3 nanobelt/ SWNTCOOH aqueous suspensions, electrode manufacture, and equipment and characterization techniques were described in previous chapter. SWNTCOOH were used as a conductive additive to improve charge transfer to active sites of the α-MoO3 nanobelts and 75 %/25 % w/w α-MoO3 /SWNTCOOH nanobelts composite (MOSC) electrodes were fabricated. The average mass and thickness of a MOSC electrode were 0.45 mg and 1.72 µm respectively. Electrochemical characterization. MOSC electrodes were tested in a threeelectrode electrochemical cell configuration using a Reference 600/EIS300 Gamry potentiostat/galvanostat, Li wires as reference and counter electrodes, and 1 M LiClO4 in propylene carbonate as the electrolyte. The cell was assembled in a glove box, MOSC electrodes were dried in vacuum (100 ◦ C for 1 hour) before all electrochemical tests, and the Li wire was freed of protective mineral oil by extensive rinsing with n-hexane. Cyclic voltammetry experiments and galvanostatic charge-discharge experiments (10 and 500 mA g−1 current density) were performed in a potential range from 1.5 to 3.5 V vs Li/Li+ . Electrochemical impedance spectroscopy (EIS) measurements were performed using an AC voltage of 20 mV rms in the frequency range of 0.01 to 50 kHz with conditioning at various DC voltages in the range 1.875-2.65 V. 164 6.3 Results and discussion 6.3.1 Material characterization The characterization of crystal structure and morphology of synthesized αMoO3 nanobelts was presented in previous chapter. As shown in Figure 6.1, the combination of sonication followed by spray deposition, allowed for the manufacture of a 75 %/25 % w/w α-MoO3 /SWNTCOOH electrode consisting of a highly interconnected open network of α-MoO3 nanobelts with interwoven SWNTCOOH providing efficient electron conducting paths. Figure 6.1: SEM image of a 75 %/25 % w/w α-MoO3 /SWNTCOOH electrode. 6.3.2 Electrochemical characterization: cyclic voltammetry The intercalation of Li+ ions into the crystal matrix of α-MoO3 [3, 5, 6] can be described by the following overall redox intercalation reaction: nLi+ + M oO3 + ne− ⇀ ↽ Lin M oO3 165 (6.1) MOSC electrodes were investigated by cyclic voltammetry (CV). Figure 6.2 shows CVs scanned at a rate of 0.1 mV s−1 from 1.5 to 3.5 V vs Li/Li+ for the first 5 cycles. In the first cycle, three cathodic (c) and three anodic (a) peaks describe the multiple step Li-ion intercalation: 2.77 V (c1 ), 2.48 V (c2 ), and 2.32 V (c3 ); 2.89 V (a1 ), 2.66 V (a2 ), and 2.45 V (a3 ). The cathodic peak c1 has been attributed to an irreversible crystal transformation involving a interlayer crystal expansion [3, 6, 8, 15] . This view was supported by the different intensity of the peaks in the couple (c1 , a1 ) with an apparent higher degree of Li-ion intercalation (c1 ) than de-intercalation (a1 ). During subsequent cycles, the electrochemical activity of both (a1 ) and (c1 ) gradually disappeared. Earlier work suggested that the couples (c2 , a2 ), and (c3 , a3 ) describe an intra-layer and an inter-layer Li-ion intercalation respectively into the orthorhombic α-MoO3 crystal structure [8, 10, 15] . The CVs in Figure 6.2 showed a marked irreversibility of the de-intercalation activity (a2 ) that readily disappeared over the second and subsequent cycles suggesting that some Li ions remained irreversibly trapped in the intralayer spaces of the α-MoO3 crystal. The electrochemical activity of the couple (c3 , a3 ) remained more stable over the first 5 cycles. Electrochemical reversibility holds for processes where k◦ /mtrans ≥ 15 where k◦ is the standard electrochemical rate constant and mtrans is the rate of mass transport [16] ; electrochemical irreversibility is therefore reached at sufficiently high scan rates when mass transport rates become higher than electrode kinetics [16] . For both reversible and irreversible limits, the current varies with the square root of the scan rate (simple one electron transfer with linear diffusion) [16] . Figure 6.3a shows the evolution of the CV shape of the 166 Figure 6.2: Cyclic voltammograms of a MOSC electrode at 0.1 mV s−1 scan rate showing cathodic (c) and anodic (a) redox peaks at cycles (a) 1 and (b) 2 and 5. The current density has been normalized per unit electrode area (1 cm2 ). MOSC electrode with increasing scan rate from 0.1 to 50 mV s−1 . Above 5 mV s−1 , only the couple (c3 , a3 ) remained electrochemically active, and thus the following analysis applies to this couple only. Figure 6.3b shows that the peak currents associated with the couple (c3 , a3 ) were a linear function of the square root of the scan rate νs hence describing one electron transfer activity with linear diffusion [16, 17] . Another criteria to evaluate reversibility is the peak-to-peak potential difference of anodic and cathodic current peaks ∆Epp , which in an ideally reversible system is ∆Epp ≈ 57 mV and is independent of 167 scan rate [16] . Figure 6.3c shows that there was irreversible behaviour where ∆Epp = 130 mV at 0.1 mV s−1 , then for larger scan rates ∆Epp increased in an approximately exponential fashion. This irreversibility of the Li ion intercalation in α−MoO3 has been reported earlier and suggested to be due to structural transformations of the layered host crystal that impose limitations to ion mobility [4, 5] . Despite the manifest irreversibility of the Li/α−MoO3 electrochemical system, a high charge storage capacity was achieved for the MOSC electrode at low scan rates as shown in Figure 6.4a, and Figure 6.4b. A maximum charge storage capacity of 697.7 C g−1 (193.8 mAh g−1 )/348.7 F g−1 (cycle 1), 600.4 C g−1 (166.8 mAh g−1 )/ 300.2 F g−1 (cycle 5) was achieved at 0.1 mV s−1 (20,000 s charging time), decreasing in a exponential fashion to 150 C g−1 (41.6 mAh g−1 ) / 75 F g−1 (cycle 5) at 50 mV s−1 (capacitance and charge have been normalized by total MOSC electrode mass mT ). In order to assess the individual contributions to charge storage, Qx (C g−1 ), of components x = α−MoO3 , SWNTCOOH, electrodes of these materials on their own were tested under identical conditions. In a MOSC electrode then the charge storage contributions of each component was qx (C) = Qx mx where mx is the mass of each component. The quantities qx were then normalized per unit mass of the individual component mx or per unit mass of the MOSC electrode mT as shown in Figure 6.4c. Considering normalization per unit mass of MOSC electrode qx /mT , the charge storage contribution of the SWNTCOOH component was almost negligible accounting for a maximum of 2.7 % of the total charge stored in the electrode. It is worth noting that when an α−MoO3 electrode without any conduc168 Figure 6.3: (a) Cyclic voltammograms of a MOSC electrode at different scan rates: 0.1, 1, 5, 10, 30, 40 and 50 mV s−1 . (b) Cathodic and anodic peak current of the redox couple (c3 , a3 ) as function of the square root of scan rate, and (c) peak-to-peak potential separation between anodic and cathodic peaks (c3 , a3 ) as function of scan rate. tive additive was tested, a poor redox intercalation activity was observed in which only the couple (c3 , a3 ) was electrochemically active (not shown), 169 yielding a charge/capacitance of 174.4 C g−1 /87.2 F g−1 (cycle 5) at 0.1 mV s−1 , i.e. ≈ 25 % of the charge storage capacity achieved by the MOSC electrode. No charge storage was measured at scan rates ≥ 5 mV s−1 . These results underline the key role of the SWNTCOOH in providing electrical conducting paths enabling charge transport towards the active sites of the α−MoO3 nanobelts. Figure 6.4: (a) Capacitance, and (b) charge storage capacity of a MOSC electrode normalized per mass of electrode as function of scan rate: 0.1, 1, 5, 10, 20, 30, 40, 50 mV s−1 corresponding to charging times of 20,000 s, 2,000 s, 400 s, 200 s, 100 s, 66.6 s, 50 s and 40 s respectively for a 1.5-3.5 V vs Li/Li+ working electrochemical window. A detail of (b) is shown in (c) in a reduced time scale (black solid line, solid squares) along with a comparison to data reported in reference [8] (red solid line, solid squares), and charge stored by the individual components of the composite normalized per mass of the component or mass of the composite electrode. Black dashed lines indicate the boundaries between capacitive and diffusion controlled ion intercalation contributions to charge storage. 170 Unlike ion intercalation processes controlled by diffusion where the peak 1/2 current varies linearly with the square root of scan rate νs , as demonstrated above for the couple (c3 , a3 ), capacitive processes vary linearly with νs [17] . The dependence of the cathodic current Ic on scan rate νs at a given potential 1.6 V ≤ E ≤ 2.8 V was analyzed further. Data was best-fitted to the expression Ic ∝ νsp where the power p was the independent variable with possible values 0.5 ≤ p ≤ 1; p = 0.5 indicates a diffusion controlled process and p = 1.0 indicates a capacitive process. Figure 6.5 shows the best-fit p values for scan rates 0.1, 1, 5, 10, 20, 30, 40, and 50 mV s−1 for each considered potential E. There was capacitive behaviour for 1.6 V ≤ E < 2.0 V whereas diffusion controlled behaviour dominated at 2.2 V ≤ E ≤ 2.8 V. The latter potential range included all the anodic and cathodic Li-ion intercalation peaks shown in Figure 6.2. Figure 6.5: Variation of best-fit p obtained from Ic ∝ νsp at potentials 1.6 V ≤ E ≤ 2.8 V within the working electrochemical window of the MOSC electrode. The fit was done for scan rates 0.1, 1, 5, 10, 20, 30, 40, and 50 mV s−1 . The dashed line is used as guide to the eye. The capacitive contributions to charge storage in the MOSC electrode may include double layer capacitance and pseudocapacitance. Redox pseudocapacitive contributions accompanying diffusion controlled Li intercalation 171 in the Li+ /α−MoO3 electrochemical system have been reported to involve surface reduction of Mo6+ to Mo5+ at 2.58 V and to Mo4+ at 1.73 V with only partial re-oxidation back to Mo6+ during subsequent anodic scanning [18] . More recently Brezesinki et al. reported a contribution of Li-ion intercalation pseudocapacitance to charge storage for templated mesoporous α−MoO3 [8] . Ion intercalation processes controlled by diffusion occur over longer time scales than capacitive processes. Brezesinki et al. estimated that for the Li+ /α−MoO3 electrochemical system, the maximum contribution to charge storage of capacitive processes was reached at a charging time of 200 s (time to scan a 1.5 to 3.5 V working electrochemical window at 10 mV s−1 ) [8] ; for times > 200 s, any additional charge storage was considered to come from diffusion controlled process [8] . Using this approximation and considering the contribution to charge storage of solely the α−MoO3 nanobelts in a MOSC electrode (C of MoO3 /g of MoO3 ), the capacitive charge storage was 443 C g−1 (200 s) whereas the contributions by the diffusion controlled ion intercalation mechanism were 463.2 C g−1 (20,000 s), 170.2 C g−1 (2,000 s) and 69.0 C g−1 (400 s). As compared with the results presented by Brezesinki et al., where a capacitive contribution of 450 C g−1 (200 s) and a diffusion controlled ion intercalation contribution of 155 C g−1 (2,000 s) were reported [8] , templated mesoporous α−MoO3 stored only 7 C g−1 more capacitive charge than α−MoO3 nanobelts in the MOSC electrode. On the other hand, the contribution of the diffusion controlled 3D Li-ion intercalation was 15 C g−1 (2,000 s) more in α−MoO3 nanobelts than in the templated mesoporous α−MoO3 . It should be noted that lower cost and possibility for scalability are advantages of the hydrothermal synthesis method combined with spray 172 deposition used in the present study over more costly templating methods. The cycling stability of a MOSC electrode was investigated by cyclic voltammetry at 10 mV s−1 for 110 cycles. The maximum charge/capacitance was 590 C g−1 (164 mAh g−1 )/ 295 F g−1 at cycle 1. Figure 6.6 shows that after 110 cycles, the capacitance retention was of 59 %, with a remanent charge/capacitance of 346 C g−1 (96 mAh g−1 )/ 173 F g−1 which is higher than charge storage provided by other metal oxides with potential application for electrochemical capacitors [19, 20] during first cycles inlcuding iron oxides [21, 22] , and some forms of manganese oxide [23, 24] . The cycling stability was also tested by galvanostatic charge-discharge studies and is discussed further in next section. Figure 6.6: Capacitance retention of a MOSC electrode tested by cyclic voltammetry at a scan rate of 10 mV s−1 . In summary, the Li-ion intercalation activity of the system Li+ /α−MoO3 , described by a series of anodic and cathodic peaks at the slow scan rate of 0.1 mV s−1 , showed to be a slow process (20,000 s) with most of the intercalation events disappearing after the first cycle or for scan rates higher than 5 mV s−1 where only the inter-layer Li-ion intercalation at (c3 , a3 ) remained electrochemically active, was controlled by linear diffusion and had an irre- 173 versible behaviour as evaluated by ∆Epp . Evaluation of the cathodic current as a function of the scan rate up to 50 mV s−1 over the entire working electrochemical window showed that the electrochemical activity of Li+ /α−MoO3 was controlled by diffusion at potentials 2.2 V ≤ E ≤ 2.8 V. Despite the electrochemical irreversible behaviour, a high charge storage contribution of 443 C g−1 was found for charging times from 40 s to 200 s (50 mV s−1 to 10 mV s−1 ) which was comparable to that achieved by templated α−MoO3 , claimed to be mainly a contribution from ion intercalation pseudocapacitance [8] . In the present study, because there was evidence of diffusion control of the electrochemical activity, it was not clear if this contribution to charge storage can be attributed to ion intercalation pseudocapacitance. However, the considerable amount of charge storage achieved at such time scales can potentially be used for electrochemical capacitors of moderate power density provided that the cycling stability could be improved. The maximum contribution to charge storage from slow diffusion controlled processes for charging times < 200 s was 463.2 C g−1 , which was slightly higher than that reported for templated α−MoO3 [8] . In the next sections: (1) the charge storage processes at even longer time scales, hence with a greater contribution by diffusion controlled processes was evaluated by galvanostatic charge-discharge experiments and (2) EIS showed an insight into mass and charge transport processes. 174 6.3.3 Electrochemical characterization: galvanostatic discharge curves Discharge curves of the MOSC electrode at 10 mA g−1 are shown in Figure 6.7. The first cycle detailed in Figure 6.7a shows a plateau at ≈ 2.8 V followed by a sloping region from 2.6 to 2.0 V. The plateau corresponded to the irreversible crystal structure transition [15] c1 in the CVs in Figure 6.2 as shown in Figure 6.7b this plateau was observed only during the first cycle and only for current densities ≤ 10 mA g−1 - and is known to consist of two phases, α-MoO3 and Lin MoO3 with 0 < n < 0.25 (see Table 6.2) where Li intercalation takes place in intralayer lattice spaces causing expansion of two adjacent lattice layers [15, 25] . The second lithiation stage below 2.6 V deviates from the ideal battery behaviour described by a constant discharge potential, and although is considered by some authors to involve almost reversible structural changes [15, 18] , as discussed below this may not be true for n > 1.5 mol Li+ /mol MoO3 . The presence of more than one phase in this second lithiation stage is also plausible as reported by earlier studies indicating that Lin MoO3 with 0 ≤ n ≤ 1.5 may not be regarded as a solid solution with the coexistence of at least two single phases [3, 4, 6] As shown in Figure 6.7b and Figure 6.7c, a capacity of 596.8 mAh g−1 (charge/mass MOSC electrode) was achieved during the first cycle corresponding to n = 4.2 mol Li+ /mol MoO3 1 . From the second and subsequent 1 Here n (mol Li+ /mol MoO3 ) = [(q1 /m1 ) x 3.6 x w1 ]/F with q1 /m1 = (qT /mT = 596.8 mAh g−1 ) x (mT /m1 ) -(q2 /m2 = 29.8 mAh g−1 ) x (m2 /m1 ), qT = q1 + q2 , mT = m1 + m2 where q is charge (mAh), m is mass (g), the subscripts 1 and 2 indicate composite components MoO3 and SWNTCOOH, respectively, F is the Faraday constant and w1 is the molecular weight of MoO3 . The charge q2 = 0 for the second and subsequent cycles. 175 Figure 6.7: Discharge curves of a MOSC electrode at 10 mA g−1 current density for cycles (a)1, (b) 1 to 9. Capacity (left) and utilization fraction of MoO3 (right) shown in (c). Labels in (a) show depth (potential) of discharge chosen for EIS analysis. cycles, as shown in Figure 6.7b, there was a capacity fading; the capacity at cycle 2 and cycle 9 were 422 mAh g−1 (n = 3.0 mol Li+ /mol MoO3 ), and 289.8 mAh g−1 (n = 2.0 mol Li+ /mol MoO3 ) respectively. Notice that a 176 discharge at 10 mA g−1 implies a longer time scale (215.0 x 103 s to discharge first cycle) than cyclic voltammetry at 0.1 mV s−1 (20,000 s), thus the discharge experiment is able to capture slower diffusion charge storage processes than the cyclic voltammetry experiment. Hence the much larger charge storage achieved by the galvanostatic discharge experiment, e.g. 289.8 mAh g−1 (cycle 9) obtained with galvanostatic discharge at 10 mA g−1 > 193.8 mAh g−1 (cycle 1) obtained with cyclic voltammetry at 0.1 mV s−1 Previous studies pointed out that the kinetically accessible stoichiometric range of the reversible lithium insertion/de-insertion was 0 ≤ n ≤ 1.5, and that higher lithium uptake implied major structural rearrangements [3, 4, 26] . Therefore, the observed irreversible behaviour can be explained by: (1) irreversible Li-ion insertion involved in major structural changes at the plateau at 2.8 V and for n > 1.5, and (2) inefficient de-lithiation due to volume expansion/shrinkage effects during the first cycle [27] . The possibility of reaction of α-MoO3 with Li+ to form lower oxides of molybdenum (including MoO2 ), Mo, and Li2 O was discarded as it is known to occur at potentials in the range 0.7-0.5 V [7, 26–28] . On the other hand, a charge storage of 289.8 mAh g−1 (n = 2.0 mol Li+ /mol MoO3 ) for cycle 9 is significant (see Table 6.1) and it was attributed to an enhanced electrochemical utilization due to the presence of SWNTCOOH as conductive additive; earlier studies report that adding a conductive additive improves electrochemical utilization yielding n > 1.5 [5, 7] . In the present study the SWNTCOOH contribution to charge storage was 29.8 mAh g−1 (5 % of the total capacity) only during the first discharge; in subsequent cycles SWNTCOOH had no contribution to charge storage. 177 For a higher discharge current density of 500 mA g−1 , a steady capacity with a maximum of 150.5 mAh g−1 was achieved for the first 13 cycles with a capacity retention of 46.7 % for cycle 117, as shown in Figure 6.8. Figure 6.8: (a) Capacity and (b) capacity retention during galvanostatic discharge of a MOSC electrode at 500 mA g−1 current density for 117 cycles. Table 6.1 provides a summary of previously reported performance of MoO3 electrodes in organic electrolytes. In most cases a direct comparison with our results, also included in the table, is inadequate due to the use of different testing conditions including different discharge current densities, electrolyte and/or slightly different working electrochemical windows. Higher capacities - not to compare with this work due to the different working electrochemical window - have been reported for the use of stoi178 Table 6.1: A summary of previously reported performance of MoO3 electrodes in organic electrolytes and the performance of MOSC electrodes in the present study. Abbreviations indicate ethylene carbonate (EC ), propylene carbonate (PC ), diethyl carbonate (DEC ), dimethyl carbonate (DMC ), and n describes stoichiometry (mol Li+ /mol MoO3 ). Material Micrometersized MoO3 powders [15] Micrometersized MoO3 rods [10] MoO3 nanobelts [29] Electrolyte Current density/ Electrochemical window 10 mA g−1 1.5 to 3.5 V 220-250 mAh g−1 (1) n = 1.5 1 M LiPF6 in 3:1:1 EC/PC/DEC 50 mA g−1 2.0 to 3.5 V 225 mAh g−1 (5) n = 1.2 199 mAh g−1 (100) 1 M LiPF6 in 1:1 EC/DMC 0.1 mA cm2 1.2 to 3.5 V 445 mAh g−1 (1) 375 mAh g−1 (30) ≈ 30 mA g−1 1.5 to 3.75 V 301 mAh g−1 (1) 180 mAh g−1 (15) 300 mA g−1 1.2 to 3.5 V 294 mAh g−1 (1) 150 mAh g−1 (100) 5,000 mA g−1 1.5 to 3.5 V 176 mAh g−1 (1) 115 mAh g−1 (50) 10 mA g−1 1.5 to 3.5 V 596.8 mAh g−1 (1) n = 4.2 289.8 mAh g−1 (9) 500 mA g−1 150.5 mAh g−1 (13) 70 mAh g−1 (117) 1 M LiClO4 in 1:1 EC/DEC MoO3 nanobelts [30] MoO3 nanorods [31] MoO3 nanobelts [32] MoO3 nanobelts (this work) 1 M LiPF6 in 1:1 EC/DMC 1 M LiPF6 in 1:1:1 EC/DMC/DEC 1 M LiClO4 in PC 179 Capacity (cycle), n Extended capacity (cycle) not reported chiometric MoO3 and non-stoichiometric MoO3−x as anode in a 0 to 3.5 V electrochemical window: MoO3 nanoparticles with n = 5.7 Li+ /mol MoO3 , i.e. 1050 mAh g−1 at C/10 current density in a 0.005 to 3.0 V working electrochemical window [7] , and MoO3−x nanowires, n = 4.1 Li+ /mol MoO3 , i.e. 770 mAh g−1 at 25 mA g−1 current density in a 0.1 to 3.5 V working electrochemical window [28] . In summary, the discharge curve of the MOSC electrode shows different intercalation stages with a plateau at 2.8 V corresponding to the irreversible peak at the cathodic peak c1 in the cyclic voltammograms, and a second lithiation stage characterized by a sloping trend and therefore resembling more a behaviour of a pseudocapacitive ion intercalation rather than an ideal battery. The charge storage during first cycles both at 10 mA g−1 and 500 mA g−1 was considerable as compared to previous work and could potentially be used for battery applications provided that the irreversibility and relatively poor cycling stability could be improved. Being a common problem for battery electrode materials, the cycling instability of MoO3 has been attributed to inherent volume expansion during lithiation and subsequent shrinkage during de-lithiation causing a loss of electrical contact among particles which can not be further lithiated during subsequent cycles [27] . Previous reports have improved the cycling stability of MoO3 by: (1) strategic reduction of particle size that increases surface area and reduces diffusion path lengths obtaining a better cycling stability for instance for reduced aspect ratio nanorods rather than nanoparticles [10, 31] , (2) pre-lithiation that enhances electrical conductivity [30] , (3) carbon coating that provides electron conductivity and reduces volume expansion effects [33] , 180 and (4) a reduction of crystallite size combined with the use of amorphous rather crystalline materials [27] . Therefore, our MOSC electrodes could be improved by: (1) reduction of the aspect ratio of alpha-MoO3 nanobelts to improve both charge storage and cycling stability, and (2) although the use of SWNTCOOH enhanced greatly the electrochemical utilization, a more uniform coating of the nanobelt with a conductive additive may be more suitable to prevent volume expansion effects. 6.3.4 Electrochemical characterization: EIS analysis EIS studies were performed to provide a further insight into the charge and mass transport processes occurring during the different stages of Li+ intercalation during discharge. As shown in Figure 6.7a, a MOSC electrode was discharged at 10 mA g−1 down to various depths (potentials) of discharge (DOD) at the end of the first plateau and along the second sloping lithiation stage: 2.65 V, 2.4 V, 2.3 V, 2.2 V, and 1.875 V, followed by an EIS test at an AC potential of 20 mV rms, a DC potential corresponding to the DOD, and a frequency range of 50 kHz to 0.001 Hz. DODs are summarized in the second column of Table 6.2. A new electrode and fresh electrolyte were used for tests at each DOD. Table 6.2 summarizes major changes occurring during lithiation by galvanostatic discharge at 10 mA g−1 as reported by Tsumura et al. [15] . The first column describes the potential of discharge (V), called DOD for experimental data in the second colum, the third column describes the crystal interlayer spacing upon Li-ion intercalation (nm), the fourth column describes the de- 181 gree of intercalation n (mol Li+ /mol MoO3 ), and the fifth column describes the phases existing at corresponding potential of discharge. Table 6.2: Major composition and structural changes of MoO3 during Li-ion insertion as documented by Tsumura et al. [15] . Experimental DODs for EIS analysis are included in the second column. Potential (V) OCP 2.8 2.7 2.4 2.3 2.25 2.2 2 DOD (V) 3.145 2.8 2.65 2.4 2.3 2.2 Interlayer spacing (nm) 0.69 0.69,1.175 1.2 n (mol Li+ / mol MoO3 ) 0 0.15 0.25 Phases MoO3 , Lin MoO3 MoO3 , Lin MoO3 MoO3 , Lin MoO3 1.24 1.225 1.175 1.16 1.115 0.55 0.82 1 1.3 1.6 Lin MoO3 Lin MoO3 Lin MoO3 Lin MoO3 Lin MoO3 1.875 The impedance data of a system is generally divided into frequency regions describing events occurring at different time scales or frequencies (f = 1/time). In the following analysis this approach was followed. At high frequencies fast charge transfer processes dominate, at medium frequencies mass transport processes become important and diffusion plays an important role, and at low frequencies the slowest non-homogeneous diffusion processes take place, in some cases accompanied by leakage currents [34] . The transition between the high and medium frequency regions is known as the ‘knee’ or characteristic frequency and it is indicated by an inflection point in a Nyquist plot or in a real (YR ) versus imaginary (YI ) admittance plot [34] . The transition between the medium and low frequency regions is considered to be at the frequency corresponding to a phase angle φ = -45 182 ◦ [34] . Figure 6.9a shows the Nyquist plot at the various DODs. Figure 6.9b is a close up of the Nyquist plot in the high/medium frequency regions that are presented in detail for each DOD in Figure 6.10. Figure 6.9c and Figure 6.9d show the corresponding Bode plot. Figure 6.9a shows that in the low frequency region as DOD increases, i.e. for a progressive degree of lithiation, the real part of the impedance increases while the phase angle, shown in Figure 6.9d as a function of frequency, decreases. The magnitude of the impedance |Z| followed the same trend as its real component ZR at low frequencies as shown in Figure 6.9c. The overall increase of resistance as the Li ion intercalation progressed could be due to electronic resistance and/or ionic resistance, the latter being either in liquid or solid phase [5] . The electronic conductivity of Lin MoO3 phases is known to increase with n up to a maximum followed by a decrease with further intercalation [5] . Because the MOSC electrode has been provided with an electron conducting additive, it is expected that the SWNTCOOH helped providing interparticle contact and electron conducting paths ameliorating electron resistance effects. On the other hand, an increase of total resistance at the end of the discharge of MoO3 electrodes has been associated to a liquid phase ionic resistance of a depleted electrolyte in the micropores of a layer that covers the electrode at the near end of the discharge [5] . Our findings seem to support this view as the MOSC electrode showed an increased resistance at the last stages of intercalation and in the low frequency region controlled by inhomogeneous diffusion. The impedance behaviour at the high/medium frequency is shown in the Nyquist plot in Figure 6.9b and detailed spectra for each DOD are shown in 183 Figure 6.10 where the ‘knee’ or characteristic frequency is indicated with an arrow and black font, and the frequency at phase angle φ = -45 is indicated with font in blue and italics. Figure 6.10 shows that the impedance behavior at the high and medium frequency regions was different for electrodes at the different stages of intercalation. Figure 6.10a and Figure 6.10b show that for early stages of intercalation at 2.65 V and 2.4 V, the characteristic frequency was not uniquely defined with a high/medium frequency region describing what could be two overlapping time constants - in an equivalent electrical circuit with a resistance R and a capacitance C connected in parallel, the time constant of the circuit is defined as the product RC; the Nyquist plot of the RC parallel circuit in the high frequency region is a semicircle; a system with two time constants is characterized by a high frequency Nyquist plot with two contiguous semicircles [12, 17, 35] -. Figure 6.10c, Figure 6.10d, and Figure 6.10e show a characteristic frequency of 199 Hz and a single time constant behaviour at high frequency with an approximate depressed semicircle. As described before and summarized in Table 6.2, at early stages of intercalation of α-MoO3 at least two crystal phases coexist: α-MoO3 , and Lin MoO3 [15, 25] . On the other hand, in the beginning of the intercalation reaction when transport paths are still short the charge transfer resistance controls the rate of intercalation [5] . EIS studies at the high frequency describe processes controlled by charge transfer [34] , and therefore the high frequency Nyquist plot of the electrode with a DOD of 2.65 V in Figure 6.10a might be describing the two time constants associated to the magnitudes of the charge transfer resistances of two coexisting phases. However, nothing can 184 be concluded by EIS alone; complementing EIS with X-ray diffraction for instance would be suitable for a more complete analysis. For DODs of 2.3 V, 2.2 V and 1.875 V shown in Figure 6.10c, Figure 6.10d, and Figure 6.10e respectively, a comparison across the high frequency Nyquist plots shows a similar charge transfer resistance and a diffusion controlled medium frequency region tending to a semi-infinite diffusion behaviour nearly achieved for the DOD of 1.875 V -the Warburg impedance describes the impedance of a diffusion process which in the case of semi-infinite conditions where a phase difference between current and voltage is 45 ◦ and is inde- pendent of frequency, is defined as ZW sem = Aω −1/2 - iAω −1/2 where ω is the √ angular frequency, i = −1, and A is the Warburg prefactor, a constant that contains a concentration independent diffusion coefficient; in a Nyquist plot the semi-infinite Warburg impedance is described by a line with a 45 slope [36] -. 185 ◦ Figure 6.9: EIS spectra of a MOSC electrode subject to various depths of discharge: 2.65 V, 2.4 V, 2.3 V, 2.2 V, and 1.875 V. (a) Nyquist plot, (b) detail at medium and high frequencies of Nyquist plot, (c) and (d) Bode plot. The experimental EIS data was fitted to electrical equivalent circuits shown in Figure 6.11 in an attempt to quantify some parameters involved in charge and mass transport of Li-ion intercalation in αMoO3 . The equivalent circuits were maintained as simple as possible yet including electrical elements accounting for the different charge and mass transport processes involved. The equivalent circuit 1 in Figure 6.11a includes two time constants to account for the overlapping regions at high/medium frequency range of the impedance data at 2.65 V and 2.4 V whereas the equivalent circuit 2 in Figure 6.11b with a single time constant was used to fit impedance data at 2.3 V, 186 Figure 6.10: Nyquist plots at the high/medium frequency regions for MOSC electrodes subject to various depths of discharge: (a) 2.65 V, (b) 2.4 V, (c) 2.3 V, (d) 2.2 V, and (e) 1.875 V. The transition between the low and medium frequency regions, i.e. the frequency at φ = -45 ◦ is indicated with font in blue and italics (Hz). The transition between the high and medium frequency regions, i.e. the characteristic frequency is indicated with an arrow and black font (Hz). 187 2.2 V and 1.875 V. Rs represents the equivalent series resistance (resistance of electrode + resistance of electrolyte + resistance of contacts), Rct represents the charge transfer resistance of the ion intercalation process, ZCP E is the impedance of a constant phase element accounting for the non faradaic processes at a rough electrode [37, 38] , and ZW represents the impedance to mass transport. The impedance of the constant phase element ZCP E is defined as [34] , ZCP E = where i = √ 1 TCP E (iω)αCP E −1, ω is the angular frequency, TCP E is the capacitance when αCP E = 1, and αCP E is the constant phase exponent (0 ≤ αCP E ≤ 1). 188 The impedance of a finite length Warburg element Zw is defined as [34] , ZW = RW coth[(iTW ω)αW ] (iTW ω)αW where RW is the limiting diffusion resistance, TW = L2 /D where L is the effective diffusion length and D is the effective diffusion coefficient, and αW is a exponent between 0 and 1. The case of semi-infinite diffusion holds for αW = 0.5. The data fitting was done using the ZView software that calculates the fitting error using χ2 statistics. In most cases the fitting was best for the high frequency data with less accuracy for the medium and low frequency data; equivalent circuits were kept as simple as possible to the expense of fitting accuracy - a sensible practice fitting experimental EIS data is avoiding complex equivalent electrical circuits with a perfect fit to experimental data but with circuit elements with unknown physical meaning. A summary of the fit parameters are given in Table 6.3 and Table 6.4. Figure 6.11: Equivalent electrical circuits used to fit the EIS experimental data of a MOSC electrode subject to depths of discharge: (a) 2.65 V and 2.4 V, and (b) 2.3 V, 2.2 V, and 1.875 V. Figure 6.12 show the variation of the equivalent series resistance Rs with DOD with a maximum of 81.2 Ω and a minimum of 55.2 Ω. The literature 189 Table 6.3: Fitting parameters of equivalent electrical circuit 1 in Figure 6.11a to EIS experimental data. Equivalent circuit 1 Fit frequency range (Hz) Rs (Ω) CPE1 , TCP E (Ω−1 sα ) CPE1 , αCP E Rct1 (Ω) CPE2 , TCP E (Ω−1 sα ) CPE2 , αCP E Rct2 (Ω) ZW , RW (Ω) ZW , TW (s), ZW , α W χ2 2.65 V 0.001-50,000 81.19 0.0015 0.45 12.04 0.0020 0.64 14.1 139.6 3.48 0.5 2.92 x 10−3 2.4 V 0.001-50,000 65.49 0.0030 0.30 77.48 0.0298 0.91 23.38 1089 8.25 0.5 3.57 x 10−4 Table 6.4: Fitting parameters of equivalent electrical circuit 2 in Figure 6.11b to EIS experimental data. Equivalent circuit 2 Fit frequency range (Hz) Rs (Ω) CPE1 , TCP E (Ω−1 sα ) CPE1 , αCP E Rct1 (Ω) ZW , RW (Ω) ZW , TW (s), ZW , α W χ2 2.3 V 0.001-50,000 81.8 2.01 x 10−4 0.62 19.9 623.2 19.4 0.39 7.04 x 10−4 190 2.2 V 0.0397-50,000 78.3 3.09 x 10−4 0.55 32.6 1029 21.9 0.47 2.92 x 10−3 1.875 V 0.0251-50,000 55.3 3.48 x 10−4 0.55 20.8 1.25 1.25 x 10−4 0.33 1.46 x 10−3 reports bulk measurements of conductivity of electrodes with values of 3.3 x 10−5 S cm−1 , and 1.3 x 10−2 S cm−1 for the non-lithiated phase MoO3 , and the lithiated phase Lin MoO3 respectively [39] . Calculating the electrode resistance with these values as R = ρ x (l/A) where ρ is resistivity, l = 1.72 µm, and A = 1 cm2 are the thickness and area of the electrode respectively, the resistance of the non-lithiated electrode is 5.2 Ω, and the resistance of the lithiated electrode is 1.3 x 10−2 Ω. Thus, the resistance difference due to lithiation is estimated to be 5.1 Ω. Figure 6.12 shows a maximum change in resistance ∆Rs = 81.2- 55.2 = 25.9 Ω which exceeds the resistance variations due to lithiation. The contribution of the electrolyte resistance must be constant across experiments as the concentration was kept constant, therefore variations of Rs are mainly attributed to the contact resistance that may have varied across experiments. Notice that in Figure 6.9 and Figure 6.10, Rs has been set to 82 Ω for the sake of comparison across EIS spectra. Figure 6.12: Equivalent series resistance Rs as a function of DOD. The variation of the charge transfer resistance Rct with DOD is shown in Figure 6.13 where in the case of DODs of 2.65 V and 2.4 V only Rct1 in the equivalent circuit 1 in Figure 6.11 was considered. Since the meaning of apparently two time constants in the high frequency region of EIS spectra 191 corresponding to DODs of 2.65 V and 2.4 V was not clear, no meaning could be ascribed to the absolute values of Rct1 and/or Rct2 . In the case of DODs of 2.3 V, 2.2 V and 1.875 V, the charge transfer resistance Rct1 values were 20.0 Ω, 32.6 Ω, and 20.7 Ω, showing a nearly constant magnitude of charge transfer resistance in the 2.3 V- 1.875 V range. Figure 6.13: Charge transfer of a MOSC electrode as a function of DOD. For DODs of 2.65 and 2.4 V, Rct1 was considered. The Ho et al. method to determine diffusion coefficients [36, 40, 41] requires a semi-infinite diffusion behaviour described above. Since this was not the case for the Li-ion intercalation in the MOSC electrodes -although the αW = 0.5 for DOD of 2.65 V, and 2.4 V, the fitting was not the optimum for the medium frequency region of the EIS spectra- this method could not be used to estimate diffusion coefficients. Values of α 6= 0.5 describe what Bisquert et al. called “anomalous diffusion” [40, 42] . As an approximation for the case of DOD of 1.875 V nearly reaching semi-infinite diffusion behaviour, a diffusion coefficient was estimated using Equation 30 in reference [36] :  Vm (dE/dn)n √ D= F 2AS 2 where Vm is the molar volume of MoO3 (138 cm3 mol−1 ), F is the Faraday 192 constant (96,485.3 C mol−1 ), S is the apparent geometric surface area of the electrode (1 cm2 ), A is the Warburg prefactor read as the slope of the curve ZR versus ω −1/2 in the semi-infinite diffusion region (medium frequency) (224 Ω s−1/2 ), and (dE/dn)n is the slope at a fixed degree of intercalation n of an equilibrium potential composition curve (estimated as 12 V mol Li−1 mol MoO3 ). The diffusion coefficient obtained was 2.9 x 10−9 cm2 s−1 which is comparable to that reported in the literature (3 x 10−9 cm2 s−1 ) for lithiated MoO3 [6] . Variations of the TCP E and αCP E are related to inhomogeneities, structural and volume changes of the electrode [38] . Values for αCP E ranged from 0.30 to 0.91 whereas TCP E values ranged from 0.30 to 29.8 Ω−1 sα , however an explanation for its variation with DOD was not obvious. 6.4 Conclusions The MOSC electrode showed an irreversible and mainly diffusion controlled electrochemical behaviour. Cyclic voltammetry performed at scan rates from 0.1 mV s−1 (20,000 s) to 50 mV s−1 (40 s) showed a total capacity of 443 C g−1 achieved at a charging time up to 200 s, estimated to be the time scale in which capacitive processes contribute to charge storage; a total capacity of 463.2 C g−1 was achieved at 200 s < charging time < 20,000 s where diffusion controlled processes contribute to charge storage. Along with the total capacitive contribution to charge storage cited above, capacitances of 167 F g−1 and 75 F g−1 achieved at 10 mV s−1 and 50 mV s−1 respectively suggested a potential application of the MOSC electrode as a 193 supercapacitor electrode providing a substantial energy density at moderate power capabilities. A larger capacity of 697.7 C g−1 (193.8 mAh g−1 ) achieved at the slowest scan rate of 0.1 mV s−1 suggested a potential application as a battery cathode which is further supported by galvanostatic charge-discharge experiments performed at 10 mA g−1 which showed an even higher capacity of 596.8 mAh g−1 (4.2 mol Li+ /mol MoO3 ) at cycle 1 and 289.8 mAh g−1 (2.0 mol Li+ /mol MoO3 ) at cycle 9. Cyclic voltammetry and galvanostatic charge-discharge experiments showed and irreversible behaviour of the Li+ /α−MoO3 electrochemical system which was reflected on the relatively poor cyclability of the MOSC electrode and attributed to irreversible lithiation during first cycle causing irreversible crystal structure changes, and to inherent volume/shrinkage changes during prolonged cycling. Some suggestions for improvement of cyclability have been given. An insight into charge and mass transport processes involved in Li-ion intercalation in α−MoO3 was given by galvanostatic charge-discharge experiments and EIS studies. Discharge curves showed the different stages of Li-ion intercalation in a 3.5-1.5 V working electrochemical window that included a first plateau at 2.8 V involving diffusion controlled irreversible crystal structural changes and a sloping region at 2.6 to 2.0 V. EIS analysis performed at different Li-ion intercalation stages showed evidence of the changes in electronic and ionic resistance involved in charge and mass transport processes with the calculation of some important variables. However, EIS studies must be complemented with spectroscopic techniques for a thorough study which 194 is beyond the scope of this thesis. 195 Bibliography [1] R. L. Smith and G. S. Rohrer, J. Solid State Chem., 1996, 124, 104–115. [2] S. T. Wang, Y. G. Zhang, X. C. Ma, W. Z. Wang, X. B. Li, Z. D. Zhang and Y. T. Qian, Solid State Commun., 2005, 136, 283–287. [3] M. E. Spahr, P. Novak, O. Haas and R. Nesper, J. Power Sources, 1995, 54, 346–351. [4] J. O. Besenhard, J. Heydecke and H. P. Fritz, Solid State Ionics, 1982, 6, 215–224. [5] J. O. Besenhard, J. Heydecke, E. Wudy, H. P. Fritz and W. Foag, Solid State Ionics, 1983, 8, 61–71. [6] P. G. Dickens and G. J. Reynolds, Solid State Ionics, 1981, 5, 331–334. [7] L. A. Riley, S.-H. Lee, L. Gedvilias and A. C. Dillon, J. Power Sources, 2010, 195, 588–592. [8] T. Brezesinski, J. Wang, S. H. Tolbert and B. Dunn, Nat. Mater., 2010, 9, 146–151. 196 [9] M. Zukalova, M. Kalbac, L. Kavan, I. Exnar and M. Graetzel, Chem. Mater., 2005, 17, 1248–1255. [10] W. Li, F. Cheng, Z. Tao and J. Chen, J. Phys. Chem. B, 2005, 110, 119–124. [11] B. E. Conway, V. Birss and J. Wojtowicz, J. Power Sources, 1997, 66, 1–14. [12] B. E. Conway, Electrochemical Supercapacitors: Scientific Fundamental and Technological Applications, Kluwer Academic/ Plenum Publishers, New York, USA, 1999. [13] B. E. Conway, J. Electrochem. Soc., 1991, 138, 1539–1548. [14] C. B.E, Electrochim. Acta, 1993, 38, 1249–1258. [15] T. Tsumura and M. Inagaki, Solid State Ionics, 1997, 104, 183–189. [16] R. Compton and C. Banks, Understanding Voltammetry, World Scientific Publishing Co. Pte. Ltd., Oxford, UK, 2007. [17] A. J. Bard and F. L., Electrochemical Methods: Fundamentals and Applications, John Wiley & Sons, Inc., Unites States of America, 2nd edn., 2001. [18] J. Swiatowska-Mrowiecka, S. de Diesbach, V. Maurice, S. Zanna, L. Klein, E. Briand, I. Vickridge and P. Marcus, J. Phys. Chem. C, 2008, 112, 11050–11058. [19] K. Naoi and P. Simon, The Electrochemical Society Interface, 2008. 197 [20] X. Zhao, B. M. Sanchez, P. J. Dobson and P. S. Grant, Nanoscale, 2011, 3, 839–855. [21] X. Zhao, C. Johnston, A. Crossley and P. S. Grant, J. Mater. Chem., 2010, 20, 7637–7644. [22] X. Zhao, C. Johnston and P. S. Grant, J. Mater. Chem., 2009, 19, 8755–8760. [23] O. Ghodbane, J. L. Pascal and F. Favier, ACS App. Mater.& Interfaces, 2009, 1, 1130–1139. [24] R. K. Sharma, H. S. Oh, Y. G. Shul and H. Kim, Physica B, 2008, 403, 1763–1769. [25] Y. Iriyama, T. Abe, M. Inaba and Z. Ogumi, Solid State Ionics, 2000, 135, 95–100. [26] J. O. Besenhard and R. Schllhorn, J. Power Sources, 1976, 1, 267–276. [27] Y. S. Jung, S. Lee, D. Ahn, A. C. Dillon and S.-H. Lee, J. Power Sources, 2009, 188, 286–291. [28] P. Meduri, E. Clark, J. H. Kim, E. Dayalan, G. U. Sumanasekera and M. K. Sunkara, Nano Lett., 2012, 12, 1784–1788. [29] B. Gao, H. Fan and X. Zhang, J. Phys. Chem. Solids, 2012, 73, 423–429. [30] L. Q. Mai, B. Hu, W. Chen, Y. Y. Qi, C. S. Lao, R. S. Yang, Y. Dai and Z. L. Wang, Adv. Mater., 2007, 19, 3712–3716. 198 [31] J. S. Chen, Y. L. Cheah, S. Madhavi and X. W. Lou, J. Phys. Chem. C, 2010, 114, 8675–8678. [32] L. Zhou, L. Yang, P. Yuan, J. Zou, Y. Wu and C. Yu, J. Phys. Chem. C, 2010, 114, 21868–21872. [33] M. F. Hassan, Z. P. Guo, Z. Chen and H. K. Liu, J. Power Sources, 2010, 195, 2372–2376. [34] W. Sugimoto, H. Iwata, K. Yokoshima, Y. Murakami and Y. Takasu, J. Phys. Chem. B, 2005, 109, 7330–7338. [35] M. Orazem and B. Tribollet, Electrochemical Impedance Spectroscopy, John Wiley & sons, Inc., United States of America, 2008. [36] C. Ho, I. D. Raistrick and R. A. Huggins, J. Electrochem. Soc., 1980, 127, 343–350. [37] S. R. Narayanan, D. H. Shen, S. Surampudi, A. I. Attia and G. Halpert, J. Electrochem. Soc., 1993, 140, 1854–1861. [38] G. J. Brug, A. L. G. van den Eeden, M. Sluyters-Rehbach and J. H. Sluyters, J. Electroanal. Chem. Interfacial Electrochem., 1984, 176, 275– 295. [39] L. Wang, J. Schindler, C. R. Kannewurf and M. G. Kanatzidis, J. Mater. Chem., 1997, 7, 1277–1283. [40] J. Bisquert, G. Garcia-Belmonte, P. Bueno, E. Longo and L. O. S. Bulhes, J. Electroanal. Chem., 1998, 452, 229–234. 199 [41] P. Yu, B. N. Popov, J. A. Ritter and R. E. White, J. Electrochem. Soc., 1999, 146, 8–14. [42] J. Bisquert, G. Garcia-Belmonte, F. Fabregat-Santiago and A. Compte, Electrochem. Commun., 1999, 1, 429–435. 200 Chapter 7 Scaleable ultra-thin and high power density graphene supercapacitor electrodes manufactured by aqueous exfoliation and spray deposition. 201 7.1 Introduction Electrochemical capacitors are high power density and long cycle life electrochemical energy storage devices for applications where a high chargedischarge rate is required for short-term power management and delivery. For example, when combined with a battery, power pulses for short-term electrical demand in the millisecond range from electrochemical capacitors can extend the life and reduce the size of batteries in systems such as mobile phones where circuit RC times < 1 s are needed [1, 2] . A further recently suggested high power application is alternating current line filtering that demands a high frequency capacitive response with RC times < 8.3 ms (120 Hz) [3] . The energy storage mechanism of a electrochemical capacitor consists of the formation of an electrical double layer at an electrode-electrolyte interface where the overall capacitance and energy density is proportional to the surface area of the electroactive material [4] . For this reason, nanostructured high surface area carbon materials have been exploited for electrochemical capacitor applications, firstly activated carbon which is fully commercialized, and more recently, carbon nanotubes and then graphene [5, 6] . Graphene consists of one-atom thick sp2 -bonded carbon sheet forming a honeycomb two dimensional (2D) nanostructure with unique electronic properties: ambipolar electric field effect with high charge carrier (massless Dirac fermions) mobilities independent of temperature, and a room temperature quantum Hall effect [7–11] . These electronic properties and a theoretical surface area of 2,600 m2 g−1 suggest that graphene may be a promising candidate material 202 for electrochemical capacitor applications. However the realization of this potential in practice demands cost-effective and scaleable synthesis methods that preserve the key properties. Major challenges include: (1) obtaining defect-free graphene that preserves its intrinsic high electrical conductivity (if high power density is to be achieved), (2) synthesis of high surface area graphene with a high yield of mono- and bi-layer flakes, and (3) inhibition of graphene’s natural tendency to restack. In order to fully utilize the charge storage capability of nanostructured materials, the fabrication of binder-free thin-film electrodes is of paramount importance. Conventional electrode manufacturing techniques combine electroactive materials with polymeric binders that reduce considerably the surface area available for charge storage, introduce inert electrode materials that reduce gravimetric capacitance and increase electrode resistance. A reduced electrode thickness favours electrochemical utilization for enhanced energy density, and reduces electron and ionic resistance (short diffusion lengths) having an overall effect of enhanced power density. Hence, the concept of “micro-supercapacitor” has emerged making reference to binder-free micrometer thick electrodes manufactured by a variety of techniques such as a combination of chemical vapor deposition (CVD) and etching (carbide derived carbons, 2-120 µm, 80 to 180 F cm−3 depending on thickness [12] ), CVD only (single-walled carbon nanotubes (SWNTs), 100-500 µm, 12 F cm−3 [13] , multiwalled carbon nanotubes (MWNTs) composites [14] ), inkjet printing (carbon nanotubes [15] , carbon, 1-2 µm, 2.1 mF cm−2 [16] ), electrophoretic deposition (carbon onions, 7 µm, 0.9 mF cm−2 , 1.3 F cm−3 [17] ), and micromachining (polymers, 0.03 F cm−2 , [18] ). Fabrication of sub-micrometer binder-free elec203 trodes that maintain high energy density has proven to be more challenging. Particularly binder-free graphene electrodes of nanometer-scale thickness have been fabricated by plasma-enhanced chemical vapour deposition (PECVD) (600 nm, 195 µF cm−2 [3] ), CVD (few layers, 80 µF cm−2 [19] ), and chemical reduction of graphene oxide followed by a dip-dry deposition in a layer-by-layer assembly (10 nm, 394 µF cm−2 [19] ). High power density has been achieved by non-defective (on basal planes) graphene produced by PECVD and vertically oriented respect to the current collector [3] , whereas energy density has been shown to be dependent not only on electrode thickness but also on graphene morphology [20] , degree and type of porosity [21–23] , and the presence of surface chemical functionalities [24] . A popular method to produce graphene has been the exfoliation of graphene oxide followed by chemical/thermal/microwave reduction combined with the manufacture of several micrometers-thick electrodes by the conventional method of adding binders [25–30] . This method produces “graphene” that contains a high fraction of residual oxides and structural defects located on the basal plane that provide an added pseudocapacitive effect [24] but undermine electronic conductivity [27, 31, 32] which in turn reduces power density. Graphene oxide-derived graphene has also been combined with functionalized MWNTs in thin-film electrodes (up to 400 nm) manufactured by a layer-by-layer assembly technique with considerable gravimetric (157 F g−1 ) and volumetric capacitance ( 144 F cm−3 ), due to the presence of pseudocapacitive chemical groups and increased packing density of the electroactive material, but with comparative low power density [24, 33] . Most of the cited electrode fabrication methods poorly control deposited 204 mass, film thickness, morphology and uniformity. Other problems include poor reproducibility, long processing times, high cost and low possibility for scalability. In this work we exfoliate graphite to produce graphene suspended in water-surfactant solutions. This method allows us to produce un-oxidised graphene in solution with high yield and a low degree of defects [34, 35] . The combination of this low cost and environmentally benign method with a scaleable spray-deposition technique allows the manufacture of nanometerthick electrodes, both at laboratory and semi-industrial scale [36, 37] . The performance of graphene electrodes is compared with that of carboxylated single walled carbon nanotubes of similar thickness, and shows unusual and potentially attractive high power density performance. 7.2 Experimental Details Materials. Elicarb SWNTs were supplied by Thomas Swan and Co. Ltd (UK); graphite, sodium cholate (NaC)(> 99 %), nitric acid (69 %) and hydrochloric acid (37% ) by Sigma Aldrich (UK); poly (ethyleneimine) (PEI, Mw= 70,000 ) by Alfa-Aesar (UK); and indium tin oxide (ITO, 7 Ω/sq sheet resistance) from Delta Technologies. Deionized water (10 MΩ · cm) was used for all processing. Graphene synthesis. Graphene was produced as previously reported by Lotya et al. [34] . Graphite (1.5 g) was added to a NaC aqueous solution (0.1 mg ml−1 , 300 ml). Ultrasonication was carried out in a low power (Ultrawave U1250D, 200 W, 30-40 kHz) sonic bath (51 hours). The resulting dispersion was then centrifuged (3000 rpm, 90 minutes), and the supernatant contain205 ing the graphene was collected. The graphene concentration C determined using the Lambert-Beer law for absorbance at λ = 660 nm and a extinction coefficient of αG = 6,600 L g−1 m−1 , was C = 0.072 mg ml−1 . SWNT processing. SWNTs were steam purified as described elsewhere [38, 39] . Subsequently, SWNTs were carboxyl-functionalized by refluxing in HNO3 (69 %, 30 hours, 100 ◦ C ) followed by filtration and extensive rinsing with deionized water until neutral pH . Aqueous dispersions of carboxylated SWNTs (0.5 mg ml−1 ) were produced by ultrasonication (600 W, 20 kHz probe) for 12 minutes while maintaining ice cooling. Electrode manufacture. Electrodes of 1 cm2 area were spray-deposited onto ITO coated glass substrates by a method described in chapter 5 [36, 37] . Prior to spray-deposition of electroactive material, substrates were pre-coated with a 5 nm layer of PEI (0.1 % w/w) solution in order to improve adhesion to the substrate. Electrodes were manufactured by spray-deposition of either a graphene or a carboxylated-SWNT aqueous suspension. The average thickness of graphene, and SWNT electrodes was 350 nm and 550 nm respectively. The average mass of spray deposited graphene electrodes was approximately read in a microbalance as 0.01-0.03 mg (see next section). The same applied for SWNT electrodes. A large area electrode, A = 1 m x 0.15 m, was spray-deposited onto an aluminium coated thin polymer web following the method previously described [36] . Equipment and characterization techniques. Absorption measurements were made using a Varian Cary 5000 UV-visible-NIR spectrometer ; transmission electron microscopy (TEM) images were obtained with a JEOL 2010 operated at 200 kV; and Raman spectroscopy was performed with a JY 206 Horiba Labram Aramis imaging confocal Raman microscope with a solid state laser of 532 nm wavelength (frequency doubled YAG) as excitation source. Surface chemical groups were characterized by X-ray photoelectron spectroscopy (XPS) in an ion pumped ESCA200 (Scienta-Gammadata ESCA 200 Upsala Sweden) equipped with a monochromatic Al Kα source, with samples supported on ITO coated glass substrates. The analyzer operated at constant pass energy of 500 eV for wide scans and 150 eV for detailed scans. Electrode thickness was determined by step height measurements in a Dektak 6M profilometer (Veeco Instruments, Inc). The weight of deposited films was measured using a Sartorius microbalance with 0.01 mg readability. Sheet resistance measurements of a 150 nm thick spray-deposited graphene film on glass substrates were performed using the Van der Pauw method and a Keithley 2400 source meter [40] . Electrochemical characterization. Graphene and SWNT electrodes were tested in a three-electrode electrochemical cell configuration using a Reference 600/EIS300 Gamry potentiostat/galvanostat, Ag/AgCl electrode as reference, a platinum sheet as counter electrode, and 1 M H2 SO4 as the electrolyte. Cyclic voltammetry and galvanostatic charge-discharge experiments were performed in a potential range from 0 to 1 V vs Ag/AgCl. Electrochemical impedance spectroscopy (EIS) measurements were performed using a AC voltage of 5 mV rms in the frequency range of 0.01 to 200 kHz with a DC voltage of 0.396 V. 207 7.3 Results and Discussion 7.3.1 Graphene characterization As-exfoliated graphene was sprayed for few seconds onto holey-carbon TEM grids for characterization. Figure 7.1a shows a low magnification image of exfoliated graphene monolayer and multilayer flakes. Figure 7.1b shows the zoomed-in image (circled area) of a monolayer in Figure 7.1a as confirmed by the corresponding electron diffraction pattern showed in Figure 7.1c where the intensity profile I along the (1210) − (0110) − (1010) − (2110) axis shows a I{1100} /I{2110} > 1, known as a signature for monolayer graphene [41–44] . According to Lotya et al. [34] and based on edge-counting of exfoliated graphene flakes, all thicker flakes consisted of less than 10 layers. In order to characterize the nanostructure of graphene electrodes themselves, TEM grids were also sprayed in a continuous way for 15 min simultaneously with the spray deposition of electrodes onto current collector substrates. A full and uniform coverage of the grid was obtained and is shown in Figure 7.1d and Figure 7.1e. Notice that the non-covered areas are the holes in the holey-carbon TEM grid where graphene flakes passed through upon spraying. Figure 7.1f shows a higher magnification view of the circled area in Figure 7.1d indicating a small degree of restacking of graphene layers which may inhibit the ultimate electrochemical performance. Spray-deposited electrodes were characterized by Raman spectroscopy and a representative spectrum is shown in Figure 7.2. Three bands were present: a D band (1340.9 cm−1 ), a G band (1571.4 cm−1 ), and a 2D band (2683.5 cm−1 ) [45] . The degree of structural defects in graphene can be eas208 Figure 7.1: TEM images of graphene flakes. (a) A wide-field image showing several graphene flakes, (b) a higher magnification view of circled area in (a) of a monolayer graphene;(c) its corresponding diffraction pattern with a intensity profile along (12110) − (0110) − (1010) − (2110) axis showing I{1100} /I{2110} > 1; (d) a wide-field image of a 15 min spray-deposited graphene; (e) higher magnification view of (d); (f) higher magnification view of circled area in (d). ily characterized by monitoring the D to G band intensity ratios ID /IG . Graphene prepared and sprayed according to our method showed ID /IG = 0.34, which was much lower than the ID /IG = 0.57 (mean value for graphene obtained by the same exfoliation procedure but with sonication times varying 209 from 2 to 430 hours) reported by Lotya et al. [34] , ID /IG = 0.67 for graphene produced by PECVD [3] , and ID /IG ≥ 1 of reduced graphene oxide [27, 46, 47] , indicating that our processing methods did not result in the formation of significant quantities of defects [41, 48] . Furthermore, the lack of broadening of the G band, typical of graphene oxide or reduced graphene oxide Raman spectra (where the D band is also typically broad due to the numerous oxygen functionalities/defects on the basal plane of graphene oxide/ reduced graphene oxide [49] ) indicated that the D band intensity arose from edge defects rather than basal plane defects [27, 41, 46] . The shape of the 2D band was typical of multilayer graphene and consistent with the progressive deposition of a large number of graphene flakes on one another until full coverage was achieved. Figure 7.2: A Raman spectrum of a spray-deposited graphene electrode. XPS C1s and O1s photoemission peaks for the graphene electrodes are shown in Figure 7.3a and Figure 7.3b respectively, where each peak was deconvoluted into separate Gaussian-Lorentzian shape components to account for the contribution of different functional groups. The components of the C1s core line were assigned to sp2 (284.5 eV) and sp3 (285.1 eV) hybridized 210 carbons, C-O, C-OH bonds (286.0 eV), and -COOH bonds (288.3 eV) [45, 50] . The same functional groups were found by analysis of the O1s core line: (C=O*)-OH (531.0 eV), C-O, C-OH (532.3 eV), -(C=O)-O*H (533.2 eV), and adsorbed water (536.4 eV) [51–53] . As supported by cyclic voltammetry studies (vide infra), the majority of detected oxygen containing functional groups, and the sp3 hybridized carbon contribution were due to the presence of NaC molecules that remained attached to graphene sheets prior, during and after spray deposition, with only a minor contribution of graphene edge defects [54, 55] . Figure 7.3: X-ray photoelectron spectroscopy of a spray-deposited graphene electrode: (a) C1s , and (b) O1s spectra. 211 7.3.2 Electrochemical characterization of graphene electrodes Graphene electrodes of 350 nm thickness were characterized by cyclic voltammetry in 1 M H2 SO4 in a 1 V voltage window at a range of scan rates. Figure 7.4a shows cyclic voltammograms at scan rates of 10, 50, 100, 200, 300, 400 and 500 mV s−1 with a quasi-rectangular shape and redox-type peaks at approximately 0.4 V. There was a combination of pseudocapacitance and double layer capacitance that was maintained at unusually high scan rates of 10,000 mV s−1 as shown in Figure 7.4b. This high scan rate behavior is in marked contrast to most of the graphene oxide-based electrochemical capacitor electrodes reported to date where significantly resistive behavior is developed at scan rates as low as 100 mV s−1 and not higher than 1,000 mV s−1 (in both aqueous and organic electrolytes) [25–29, 56] . The high power capability of the graphene electrodes can be attributed to: (1) the pristine nature of the graphene produced by exfoliation that preserves the integrity of the sp2 bonding configuration and therefore intrinsic electrical conductivity; (2) minimized resistance because of nanometer-scale thickness electrodes, as shown later by electrical conductivity and EIS measurements; (3) and the use of an electrolyte with high ionic conductivity. Geometric capacitance Cs of the electrode films is used in this work to avoid inaccuracies related to the mass determination of the ultra-thin R graphene electrodes, and is given by: Cs (F cm−2 ) = 21 Idt/A∆V where I is current (A), t is time (s), A is the electrode geometric area (1 cm2 ), and ∆V is the voltage window (1 V). A plot of geometric capacitance versus scan 212 Figure 7.4: Cyclic voltammograms of graphene electrodes before annealing (DC conductivity of 859 S m−1 ) at increasing voltage scan rates of (a) 10, 50, 100, 200, 300, 400, and 500 mV s−1 , and (b) 1,000, 2,000, 3,000, 4,000, 5,000, 6,000, 7,000, 8,000, 9,000, 10,000 mV s−1 . rate of the graphene electrodes is shown in Figure 7.5a. The highest capacitance at the lowest scan rate of 10 mV s−1 was 543 µF cm−2 , decreasing by 43 % to 309 µF cm−2 at 10,000 mV s−1 , as shown in Figure 7.5b. The capacitance retention was superior to other graphene-oxide-derived “graphene” electrodes where capacitance faded by 35 % at a comparatively slow scan rate of 800 mV s−1 [56] . The performance here was comparable with high performance pseudocapacitive RuO2 that had 53 % capacitance retention at 10,000 mV s−1 [57] . For comparison, Figure 7.5b shows a similar plot for carboxylated-SWNT spray-deposited 550 nm thick electrodes that had an almost complete capacitance fading (85 %) by 4,000 mV s−1 , suggesting that the high power of the graphene electrodes cannot be attributed solely to their thin film character. 213 The energy density and power density of the graphene electrodes are shown in the ragone plot in Figure 7.5c, where energy and power densities were calculated from cyclic voltammetry data as follows: E(Wh cm−2 ) = 1 C ∆V 2 2 s and P (W cm−2 ) = E t where Cs is the geometric capacitance (F cm−2 ), t is the time for the cathodic sweep in a voltammogram (h), and ∆V is defined as before. The data points corresponded to voltage scan rates from 10 to 20,000 mV s−1 . The energy density of the graphene electrode followed a gradually decreasing trend with a maximum of 75.4 nWh cm−2 at power density of 2.7 µW cm−2 , and a maximum power density of 2.6 mW cm−2 at 36.1 nWh cm−2 . Corresponding peak capacitance (CM ), energy density (EM ) and power density (PM ) per unit mass were estimated as CM ≈ 18 F g−1 (from cyclic voltammetry at a scan rate of 10 mV s−1 ), EM ≈ 2.5 Wh kg−1 , and PM ≈ 86.8 kW kg−1 . We are cautious in cross-comparing these per unit mass values with those in the literature because of inherent problems in determining accurately the mass of all thin film electrodes, as recognized in previous work where suitable metrics are used to characterize ultra-thin electrochemical capacitors electrodes [3, 19, 58] . In any case, irrespective of mass, the peak geometric capacitance reported here of 543 µF cm−2 (15.6 F cm−3 ) is the highest so far reported for ultra-thin graphene electrodes: Yoo et al. and Wang et al. reported geometric capacitances of 394 µF cm−2 and 279 µF cm−2 respectively [19, 59] . Markedly, EIS studies showed that the capacitance of our spray deposited graphene exfoliated flakes lying predominantly in the plane of the electrode was 221.9 µF cm−2 at 0.01 Hz which is 1.1 times greater than 195.0 µF cm−2 (capacitance of a single electrode of 2 cm2 in a symmetric 214 Figure 7.5: Plots of (a) geometric capacitance as a function of scan rate for a graphene electrode, and (b) capacitance retention as a function of scan rate for a graphene and a carboxylated SWNT electrodes, and (c) ragone plot normalized by geometric electrode area for a graphene electrode. The solid dots represent previously reported data for thin graphene electrodes [19] . full cell of ≈ 390 µF capacitance, i.e. 97.5 µF cm−2 when normalized per a 4 cm2 area of 2 electrodes) at 0.01 Hz reported for vertically aligned PECVD grown graphene [3] . In both cases, no capacitance saturation was observed at 0.01 Hz. As shown in Figure 7.5c, both the energy and power density of our graphene electrodes were higher than those reported in the literature of 14 nWh cm−2 , and 9 µW cm−2 for ultra-thin graphene electrochemical capacitors [19] . 215 A geometric capacitance of 543 µF cm−2 is however in the low range when compared with mesoporous carbon materials such as networks of carbon nanotubes and activated carbon with higher energy density inherent to their porous nature (high surface area), although these carbon forms have poorer power density performance (higher ionic resistance) than graphene [3] . The particularly high power density of our electrodes derives from the nonporous and highly conductive nature of the non-defective exfoliated graphene - it is too pristine to provide a high energy density, but ideal to provide high power density. Further supporting this view, it is known that the reactive sites of several forms of sp2 hybridized carbon, including highly ordered pyrolytic carbon and carbon nanotubes, reside on the edge planes populated with surface defects rather than the basal planes free of defects [60] , therefore suggesting that the relatively low capacitance observed in our graphene electrodes is correlated to the absence of basal plane defects having a charge storage contribution only from edge defects. On the other hand, adding a degree of non-subnanometer scale porosity to graphene would substantially increase capacitance while keeping a high power density as indicated by the behaviour of porous carbon-onions containing less than 5 % subnanometer micropores [21, 22] . Sodium cholate (NaC) is the corresponding salt of cholic acid used in the graphene exfoliation and consists of a steroid nucleus (hydrophobic face) with three hydroxyl groups (hydrophilic face) and an aliphatic chain terminated with a carboxylic group [54] . It is thus considered an amphiphilic molecule where the hydrophobic face of NaC interacts with graphene during exfoliation whereas the hydrophilic face provides compatibility with the aqueous envi216 ronment having an overall encapsulation-like interaction with graphene that produces a stabilizing effect [54, 55] . Therefore, the NaC-graphene interaction likely leads to NaC entrapped between graphene layers after spraying giving rise to the redox peaks in the cyclic voltammograms which coincided with typical redox activity of oxygen functionalities on carbon surfaces in acidic media (0.4 V vs Ag/AgCl) [61, 62] . In order to confirm the contribution of residual NaC to capacitive behavior, a spray-deposited graphene electrode was annealed at 500 ◦ C for 4 hours in a hydrogen/argon (10/90) atmosphere [32] and Figure 7.6 shows a cyclic voltammogram of the annealed electrode. The redox peaks previously attributed to residual NaC almost completely disappeared suggesting that redox activity was associated to residual NaC functional groups. Figure 7.6: Cyclic voltammograms of graphene/NaC electrodes before and after annealing at 500 ◦ C in a H2 /Ar (10%/ 90 %) atmosphere for 4 hours. The voltage scan rate is 50 mV s−1 . Investigation of the electrical properties of the electrodes showed a direct current (DC) conductivity of 859 S m−1 , and 4905 S m−1 before and after annealing respectively. These DC conductivity values are much higher than those previously reported for graphene oxide derived electrodes of 100 217 S m−1 , and 500 S m−1 [23, 28] , and supports further the maintenance of a large fraction of the sp2 bonding configuration of graphene. It is suggested that the high quality of the graphene produced by exfoliation underpins the high power performance of the electrodes. Furthermore, conductivity measurements suggest that higher power performances than shown in Figure 7.4 can be expected upon annealing-driven removal of NaC, albeit with an unavoidable loss of the pseudocapacitance contribution to charge storage. Although a contribution to electrochemical behavior from NaC should be expected - and residual NaC is required to inhibit graphene restacking - it was unexpected that the pseudocapacitive activity could be maintained without significant resistive behavior up to 10,000 mV s−1 , as shown in Figure 7.4. Except for nanostructured ruthenium oxide [57] such high power density is not widely reported, for either metal oxides or other pseudocapacitive groups on carbon-based materials. One implication of this result may be that when pristine graphene is used as substrate for pseudocapacitive materials or other compounds with electrochemically active groups, the underlying graphene can help delay the onset of resistive behaviour at very high scan rates. This approach could help increase the otherwise comparatively poor energy density of graphene electrodes, and facilitate the development of systems that offer a better balance of energy and power densities. Galvanosatic charge/discharge studies for the graphene electrodes were performed at current densities of 25 to 400 µA cm−2 , and charge/discharge curves for the the first cycles at 25 µA cm−2 are shown in Figure 7.7a. There were two regimes of behavior, from 1 V to 0.4 V and from 0.4 V to 0 V. The inflection point at 0.4 V describes the redox activity due to NaC previously 218 observed by cyclic voltammetry. Discharge curves of the 10th cycle at current densities of 25, 50, 100, 200, and 400 µA cm−2 are shown in Figure 7.7b. Underscoring their high power density performance, the discharging times of the graphene electrodes (31.1, 14, 6.6, 3.1, and 1.4 s for current densities of 25, 50, 100, 200, and 400 µA cm−2 , respectively) were at least one order of magnitude shorter than those reported in the literature to date [19, 20, 25, 28] . Figure 7.7: Galvanostatic charge-discharge curves of (a) a graphene electrode during first cycles at 50 µA cm−2 ; (b) a graphene electrode for the 10th cycle at increasing current densities of 25 (a), 50 (b), 100 (c), 200 (d), and 400 (e) µA cm−2 . Impedance spectroscopy studies of graphene electrodes were performed from 0.01 Hz to 200 kHz at an AC voltage of 5 mV rms. In order to investigate both the double layer and pseudocapacitance energy storage mechanisms, a DC voltage of 0.396 V was applied. We applied the methodology proposed by Sugimoto et al. for analysis of the impedance data dividing it 219 into different frequency regions (time scales) where different mass and charge transport processes dominate [63] . Figure 7.8 shows the different frequency regions in the Nyquist plot for a graphene electrode. The knee-frequency and characteristic frequency were 12.7 kHz and 57.8 Hz respectively. The high frequency region (200 kHz < f < 12.7 kHz) in Figure 7.8b showed a depressed semicircle, better described as a low gradient plot running almost parallel to the real axis Z’, revealing the low impedance at the electrode-electrolyte interface. Best fitting of the data at high frequency was performed using the equivalent circuit Rs (CPEf (Rf ))(CPEdl (Rct ZW )), where Rs represents the resistive contribution from the electrolyte, CPEf and Rf are a constant phase element and a resistance, respectively, representing the electrical equivalent of the film as a whole, CPEdl is the double layer capacitance, Rct is the charge transfer resistance of pseudocapacitive processes, and ZW is the impedance of a finite length Warburg element representing diffusion processes of the electrolyte ions into the film [64] (see Supporting Information). A low charge transfer resistance was estimated from the fit, Rct = 1.0 Ω, providing evidence of the fast kinetics of the pseudocapacitive activity. As compared with other carbon materials, Rct for the graphene electrode was one order of magnitude lower than that reported for multi-walled carbon nanotubes and carbon black functionalized with oxygen-containing chemical groups (23-29 Ω) [64] . The medium frequency region (12.7 kHz< f < 57.8 Hz) shown in Figure 7.8c followed a nearly vertical trend indicating capacitive behavior. Notice that the EIS behavior of the graphene electrode is due to the electroactive material and not to the ITO coated glass substrate as evidenced by EIS studies carried out under identical conditions on a bare ITO coated glass 220 substrate and shown in Figure 7.8. The Nyquist plot in Figure 7.8c shows a nearly pure capacitive behavior at high frequencies for the ITO coated glass susbstrate. The capacitor response, defined as the inverse of the characteristic frequency, was determined as 17.4 ms and was comparable with that of anhydrous pseudocapacitive RuO2 (10-20 ms) [63] . Further optimization of the electrode performance included the improvement of the electrical contact between the current collector and the electroactive material by excluding the PEI precoating, annealing for NaC removal, and decreasing of the electrode thickness to 40 nm. These changes resulted into an even faster capacitor response of 2.3 ms. Recently, Miller et al. suggested the application of a graphene double layer capacitor for filtering voltage ripple for which a pure capacitive behavior (phase angle near -90 ◦ ) is needed at high frequency [3] . A phase angle of -82 ◦ at 120 Hz was reported for vertically plasma enhanced CVD grown graphene directly on current collectors whereas in this study a slightly lower phase angle of -72 ◦ at 120 Hz was obtained for exfoliated and sprayed graphene electrodes (see Appendix A). The cyclability of the graphene electrodes was examined by cyclic voltammetry and galvanostatic charge/discharge experiments. There was approximately 100 % capacitance retention after 5,000 cycles at a comparatively high scan rate of 10,000 mV s−1 , as shown in Figure 7.9a. Similarly, there was an approximately 100 % capacitance retention following charge-discharge experiments at 50 and 100 µA cm−2 for 1,000 cycles (see Appendix A). 221 Figure 7.8: (a) Nyquist plot of a graphene electrode and a bare ITO coated glass substrate performed at 5 mV rms AC voltage in the frequency range of 0.01 Hz to 200 kHz, close ups of (b) high, and (c) medium frequency ranges. Numbers indicate (a) maximum frequency, (b) knee-frequency, (c) characteristic frequency, and (d) lowest frequency. 7.3.3 Scaleability Finally, the scaleability of the spray deposition manufacturing method was demonstrated for a 0.15 m x 1 m graphene electrode spray-deposited onto an aluminum coated polyethylene thereftalate (PET) flexible substrate as shown in Figure 7.9b and Figure 7.9c. 222 Figure 7.9: (a) Capacitance retention of a graphene electrode upon cyclic voltammetry at 10,000 mV s−1 for 5,000 cycles; the inset shows optical images of the graphene electrode before and after cycling, (b) A 0.15 m x 1 m graphene electrode deposited onto an aluminum coated PET web. 7.4 Conclusions Typically 350 nm thick graphene supercapacitor electrodes with an outstanding power performance and excellent cyclability demonstrated by cyclic voltammetry and galvanostatic charge-discharge experiments have been manufactured by the combination of aqueous exfoliation followed by spray deposition. Unlike supercapacitor devices manufactured with graphene-oxide derived graphene, double layer capacitance and pseudocapacitance were maintained up to unusually high scan rates of 10,000 mV s−1 , giving a peak power density of 2.6 mW cm−2 and significantly out-performing similar electrodes using functionalized single-walled carbon nanotubes. The use of comparatively pristine graphene underpinned the high scan rate pseudocapacitance 223 of residual exfoliant species, which made a useful contribution to overall capacitance at all scan rates. Electrochemical impedance spectroscopy analysis showed a charge transfer resistance of 1 Ω and a fast capacitor response of 17.4 ms which was further improved to 2.3 ms for the thinnest 40 nm electrodes after annealing. Near 100 % capacitance retention at 10,000 mV s−1 was demonstrated. The aqueous exfoliation and spray deposition manufacturing route was shown to be easily scaleable, and offers the potential for low cost processing of graphene electrodes with ultrahigh power density for niche applications. 224 Bibliography [1] R. Kötz and M. Carlen, Electrochim. Acta, 2000, 45, 2483–2498. [2] J. R. Miller and P. Simon, Science, 2008, 321, 651–652. [3] J. R. Miller, R. A. Outlaw and B. C. Holloway, Science, 2010, 329, 1637– 1639. [4] B. E. Conway, Electrochemical Supercapacitors: Scientific Fundamental and Technological Applications, Kluwer Academic/ Plenum Publishers, New York, USA, 1999. [5] P. Simon and Y. Gogotsi, Nat. Mater., 2008, 7, 845–854. [6] X. Zhao, B. Mendoza-Sánchez, P. J. Dobson and P. S. Grant, Nanoscale, 2011, 3, 839–855. [7] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, M. I. Katsnelson, I. V. Grigorieva, S. V. Dubonos and A. A. Firsov, Nature, 2005, 438, 197–200. [8] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva and A. A. Firsov, Science, 2004, 306, 666–669. 225 [9] A. K. Geim and K. S. Novoselov, Nat. Mater., 2007, 6, 183–191. [10] Y. Zhang, Y.-W. Tan, H. L. Stormer and P. Kim, Nature, 2005, 438, 201–204. [11] S. Pisana, M. Lazzeri, C. Casiraghi, K. S. Novoselov, A. K. Geim, A. C. Ferrari and F. Mauri, Nat. Mater., 2007, 6, 198–201. [12] J. Chmiola, C. Largeot, P.-L. Taberna, P. Simon and Y. Gogotsi, Science, 2010, 328, 480–483. [13] D. N. Futaba, K. Hata, T. Yamada, T. Hiraoka, Y. Hayamizu, Y. Kakudate, O. Tanaike, H. Hatori, M. Yumura and S. Iijima, Nat Mater, 2006, 5, 987–994. [14] V. L. Pushparaj, M. M. Shaijumon, A. Kumar, S. Murugesan, L. Ci, R. Vajtai, R. J. Linhardt, O. Nalamasu and P. M. Ajayan, P Natl Acad Sci USA, 2007, 104, 13574–13577. [15] K. Kords, T. Mustonen, G. Tth, H. Jantunen, M. Lajunen, C. Soldano, S. Talapatra, S. Kar, R. Vajtai and P. M. Ajayan, Small, 2006, 2, 1021– 1025. [16] D. Pech, M. Brunet, P.-L. Taberna, P. Simon, N. Fabre, F. Mesnilgrente, V. Conedera and H. Durou, J. Power Sources, 2010, 195, 1266–1269. [17] D. Pech, M. Brunet, H. Durou, P. Huang, V. Mochalin, Y. Gogotsi, P.-L. Taberna and P. Simon, Nat Nano, 2010, 5, 651–654. [18] W. Sun and X. Chen, Microelectron Eng, 2009, 86, 1307–1310. 226 [19] J. J. Yoo, K. Balakrishnan, J. Huang, V. Meunier, B. G. Sumpter, A. Srivastava, M. Conway, A. L. Mohana Reddy, J. Yu, R. Vajtai and P. M. Ajayan, Nano Lett., 2011, 11, 1423–1427. [20] C. Liu, Z. Yu, D. Neff, A. Zhamu and B. Z. Jang, Nano Lett., 2010, 10, 4863–4868. [21] C. Portet, G. Yushin and Y. Gogotsi, Carbon, 2007, 45, 2511–2518. [22] C. Portet, J. Chmiola, Y. Gogotsi, S. Park and K. Lian, Electrochim. Acta, 2008, 53, 7675–7680. [23] Y. Zhu, S. Murali, M. D. Stoller, K. J. Ganesh, W. Cai, P. J. Ferreira, A. Pirkle, R. M. Wallace, K. A. Cychosz, M. Thommes, D. Su, E. A. Stach and R. S. Ruoff, Science, 2011, 332, 1537–1541. [24] H. R. Byon, S. W. Lee, S. Chen, P. T. Hammond and Y. Shao-Horn, Carbon, 2011, 49, 457–467. [25] S. Vivekchand, C. Rout, K. S. Subrahmanyam, A. Govindaraj and C. N. R. Rao, J. Chem. Sci., 2008, 120, 9–13. [26] K. S. Subrahmanyam, S. R. C. Vivekchand, A. Govindaraj and C. N. R. Rao, J. Mater. Chem., 2008, 18, 1517–1523. [27] S. Stankovich, D. A. Dikin, R. D. Piner, K. A. Kohlhaas, A. Kleinhammes, Y. Jia, Y. Wu, S. T. Nguyen and R. S. Ruoff, Carbon, 2007, 45, 1558–1565. [28] Y. Wang, Z. Shi, Y. Huang, Y. Ma, C. Wang, M. Chen and Y. Chen, J. Phys. Chem. C, 2009, 113, 13103–13107. 227 [29] M. D. Stoller, S. J. Park, Y. W. Zhu, J. H. An and R. S. Ruoff, Nano Lett., 2008, 8, 3498–3502. [30] Y. Zhu, S. Murali, M. D. Stoller, A. Velamakanni, R. D. Piner and R. S. Ruoff, Carbon, 2010, 48, 2118–2122. [31] X. Li, G. Zhang, X. Bai, X. Sun, X. Wang, E. Wang and H. Dai, Nat. Nanotechnol., 2008, 3, 538–542. [32] S. De, P. J. King, M. Lotya, A. O’Neill, E. M. Doherty, Y. Hernandez, G. S. Duesberg and J. N. Coleman, Small, 2010, 6, 458–464. [33] D. Yu and L. Dai, J. Phys. Chem. Lett., 2009, 1, 467–470. [34] M. Lotya, P. J. King, U. Khan, S. De and J. N. Coleman, ACS Nano, 2010, 4, 3155–3162. [35] M. Lotya, Y. Hernandez, P. J. King, R. J. Smith, V. Nicolosi, L. S. Karlsson, F. M. Blighe, S. De, Z. Wang, I. T. McGovern, G. S. Duesberg and J. N. Coleman, J. Am. Chem. Soc., 2009, 131, 3611–3620. [36] B. Mendoza-Sánchez, L. Minati and P. S. Grant, Fourth European Symposium on Super Capacitors and Applications, Bordeaux, France, 2010. [37] X. Zhao, B. T. T. Chu, B. Ballesteros, W. L. Wang, C. Johnston, J. M. Sykes and P. Grant, Nanotechnology, 2009, 20, 065605. [38] B. Ballesteros, G. Tobias, L. Shao, E. Pellicer, J. Nogues, E. Mendoza and M. Green, Small, 2008, 4, 1501–1506. 228 [39] G. Tobias, S. Lidong, C. Salzmann, Y. Huh and M. Green, J. Phys. Chem., 2006, 110, 22318–22322. [40] V. d. L. Pauw, Philips Res. Rep., 1958, 13, 1–9. [41] Y. Hernandez, V. Nicolosi, M. Lotya, F. M. Blighe, Z. Sun, S. De, I. T. McGovern, B. Holland, M. Byrne, Y. K. Gun’Ko, J. J. Boland, P. Niraj, G. Duesberg, S. Krishnamurthy, R. Goodhue, J. Hutchison, V. Scardaci, A. C. Ferrari and J. N. Coleman, Nat. Nanotechnol., 2008, 3, 563–568. [42] J. C. Meyer, A. K. Geim, M. I. Katsnelson, K. S. Novoselov, D. Obergfell, S. Roth, C. Girit and A. Zettl, Solid State Commun., 2007, 143, 101–109. [43] J. C. Meyer, A. K. Geim, M. I. Katsnelson, K. S. Novoselov, T. J. Booth and S. Roth, Nature, 2007, 446, 60–63. [44] S. Horiuchi, T. Gotou, M. Fujiwara, R. Sotoaka, M. Hirata, K. Kimoto, T. Asaka, T. Yokosawa, Y. Matsui, K. Watanabe and M. Sekita, Jpn. J. Appl. Phys., 2003, 42, L1073–L1076. [45] H. Murphy, P. Papakonstantinou and T. I. T. Okpalugo, J. Vac. Sci. Technol. B, 2006, 24, 715–720. [46] D. Yang, A. Velamakanni, G. Bozoklu, S. Park, M. Stoller, R. D. Piner, S. Stankovich, I. Jung, D. A. Field, C. A. Ventrice Jr and R. S. Ruoff, Carbon, 2009, 47, 145–152. [47] K. Haubner, J. Murawski, P. Olk, L. M. Eng, C. Ziegler, B. Adolphi and E. Jaehne, ChemPhysChem, 2010, 11, 2131–2139. 229 [48] A. C. Ferrari, J. C. Meyer, V. Scardaci, C. Casiraghi, M. Lazzeri, F. Mauri, S. Piscanec, D. Jiang, K. S. Novoselov, S. Roth and A. K. Geim, Phys. Rev. Lett., 2006, 97, 187401. [49] K. N. Kudin, B. Ozbas, H. C. Schniepp, R. K. Prud’homme, I. A. Aksay and R. Car, Nano Lett., 2007, 8, 36–41. [50] H. Ago, T. Kugler, F. Cacialli, W. R. Salaneck, M. S. P. Shaffer, A. H. Windle and R. H. Friend, J. Phys. Chem. B, 1999, 103, 8116–8121. [51] L. Weng, C. Poleunis, P. Bertrand, V. Carlier, M. Sclavons, P. Franquinet and R. Legras, J. Adhes. Sci. Technol., 1995, 9, 859–871. [52] A. M. Diez-Pascual, G. Martinez, J. M. Gonzalez-Dominguez, A. Anson, M. T. Martinez and M. A. Gomez, J. Mater. Chem., 2010, 20, 8285–8296. [53] G. P. López, D. G. Castner and B. D. Ratner, Surf. Interface Anal., 1991, 17, 267–272. [54] S. Mukhopadhyay and U. Maitra, Curr. Sci. India, 2004, 87, 1666–1683. [55] A. A. Green and M. C. Hersam, Nano Lett., 2009, 9, 4031–4036. [56] W. Lv, D.-M. Tang, Y.-B. He, C.-H. You, Z.-Q. Shi, X.-C. Chen, C.-M. Chen, P.-X. Hou, C. Liu and Q.-H. Yang, ACS Nano, 2009, 3, 3730–3736. [57] K. M. Lin, K. Chang, C. C. Hu and Y. Y. Li, Electrochim. Acta, 2009, 54, 4574–4581. [58] S. W. Lee, B.-S. Kim, S. Chen, Y. Shao-Horn and P. T. Hammond, J. Am. Chem. Soc., 2008, 131, 671–679. 230 [59] D.-W. Wang, F. Li, Z.-S. Wu, W. Ren and H.-M. Cheng, Electrochem. Commun., 2009, 11, 1729–1732. [60] C. E. Banks, T. J. Davies, G. G. Wildgoose and R. G. Compton, Chem. Comm., 2005, 829–841. [61] E. Frackowiak and F. Béguin, Carbon, 2001, 39, 937–950. [62] H. A. Andreas and B. E. Conway, Electrochim. Acta, 2006, 51, 6510– 6520. [63] W. Sugimoto, H. Iwata, K. Yokoshima, Y. Murakami and Y. Takasu, J. Phys. Chem. B, 2005, 109, 7330–7338. [64] M. D. Obradovic, G. D. Vukovic, S. I. Stevanovic, V. V. Panic, P. S. Uskokovic, A. Kowal and S. L. Gojkovic, J. Electroanal. Chem., 2009, 634, 22–30. 231 Chapter 8 Engineering hybrid electrodes combining the properties of graphene and SWNTs using a layer-by-layer manufacturing method 8.1 Introduction The previous chapter demonstrated that graphene thin film electrodes can provide a high power density but suffer from low energy density. In order to enhance energy density, a ‘hybrid’ electrode that combines the high power density properties of graphene with high energy density of a second material providing with double layer capacitance and/or pseudocapacitance can 232 be conceived. In this chapter this idea is explored using a layer-by-layer (LBL) film manufacturing approach that consists on iteratively laying down alternate layers of material of different nature and/or frequently of opposite charges to form self-assembled bilayers. LBL films have been manufactured mainly by two techniques: (1) dipping and (2) spray deposition. Spray deposition allows for faster processing times and a uniform film growth at the early stage (few layers) of film deposition [1–4] . The versatility of the technique suggest that a wide variety of materials combinations may be possible for the achievement of key properties of the bi- or multi-layer construct. The LBL assembly of bilayers of multi-walled carbon nanotubes functionalized with chemical groups of opposite charges have achieved nanometer-thick layers of controllable thickness, enhanced density, tunable morphology, and a controlled linear increase of charge storage with increased thickness [1, 5, 6] . The morphology of the films is controlled by tuning the degree of ionization of functional groups which is in turn controlled by pH regulation of suspensions - a principle that is generally applied to polyelectrolytes [2] - resulting in networks of tunable porosity and increased mechanical strength. Hybrid films combining carbon nanotubes with nanostructured metal oxides have been manufactured following a LBL approach to improve electrochemical utilization of the metal oxide for charge storage [7, 8] . The LBL technique has also been applied for the coating of textiles of high surface area with colloidal particles for photocatalytic applications, an approach that can be extended to the manufacture of flexible supercapacitor electrodes [9] . Here, LBL electrodes comprised of alternate layers of graphene (high 233 power density material) and carboxyl-functionalized single-walled carbon nanotubes (high energy density material) were manufactured by spray deposition. The electrochemical performance of the resulting ‘hybrid’ electrode was investigated and compared with the performance of electrodes of the individual components as well as with a typical composite electrode manufactured by spray deposition of a suspension of a mix of the components. 8.2 Experimental details Graphene and single-walled carbon nanotubes. Graphene and carboxyl-functionalized single walled carbon nanotubes (SWNTCOOH) were synthesized/ processed as described in the previous chapter. Graphene exfoliated in presence of SWNTCOOH. An aqueous dispersion of SWNTCOOH (0.07 mg ml−1 ) was produced by ultrasonication (600 W, 20 kHz probe) for 12 minutes while maintaining ice cooling. NaC (0.1 mg ml−1 ) and then graphite (5 mg ml−1 ) were added to the SWNTCOOH dispersion. The mix was ultrasonicated in a low power (Ultrawave U1250D, 200 W, 3040 kHz) sonic bath (41 hours). The resulting dispersion was then centrifuged (3000 rpm, 90 minutes), and the supernatant containing the graphene and previously dispersed SWNTCOOH was collected. Due to the presence of 2 solute species, the concentration of the final mix could not be determined using the Lambert-Beer law. As described in the previous chapter, a graphene suspension of 0.07 mg ml−1 was obtained following the same procedure as here but without the presence of SWNTCOOH. Therefore it was considered that an approximate concentration in the mix of graphene alone was 0.07 mg 234 ml−1 . As the initial concentration for SWNTCOOH was also 0.07 mg ml−1 , the mix was considered to be 50 %/ 50 % w/w graphene/SWNTCOOH. Electrode manufacture. Electrodes were manufactured by the standard spray deposition technique described in previous chapters. LBL electrodes were obtained by spray deposition of successive layers of first graphene (0.072 mg ml−1 , 6 min, 2 ml min−1 flow rate) and then SWNTCOOH (0.5 mg ml−1 , 5 min, 2 ml min−1 flow rate) aqueous suspensions. The thickness of a seven-layer electrode (LBL7) was 292 nm with an approximate mass of 0.04 mg. A composite electrode of 50 %/ 50 % w/w graphene/SWNTCOOH was obtained by spray deposition of the respective suspension (30 min, 2 ml min−1 flow rate) resulting in an electrode with a thickness of 123 nm and an approximate mass of 0.02 mg. The manufacture of the graphene and SWNTCOOH electrodes was as reported in the previous chapter. Equipment, characterization techniques, and the set up for electrochemical characterization were as described in the previous chapter. 8.3 Results and discussion As shown in Scheme 8.1, LBL electrodes were manufactured by iterative spray deposition of first graphene films and then SWNTCOOH films until completion of a total of 7 layers (LBL7). The first step for the manufacture of LBL electrodes was the achievement of uniform films of minimum thickness of each of the materials conforming the blocks of the LBL assembly. As shown in Figure 8.1, an aqueous suspension of graphene was sprayed onto TEM grids at increasing times of 0.5, 1, 2, 4, 235 Scheme 8.1: Layer by layer spray deposition of an electrode consisting of alternated layers of graphene (layer 1) followed by SWNTCOOH (layer 2) and so on until completion of a total of 7 layers. and 6 min until full coverage was achieved at the minimum thickness. The same procedure was followed for an aqueous suspension of SWNTCOOH for increasing times of 0.5, 1, 2, 4, and 6 min as shown in Figure 8.2. Next, the LBL spray deposition depicted in Scheme 8.1 was performed. Figure 8.1: TEM images of spray deposited graphene at increasing times: (a) 0.5 min, (b) 1 min, (c) 2 min, (d) 4 min, (e) and (f) 6 min. The LBL electrodes were then tested by cyclic voltammetry and their performance was compared to graphene and SWNTCOOH electrodes pre236 Figure 8.2: TEM images of spray deposited graphene at increasing times: (a) 1 min, (b) 2 min, (c) 4 min, and (d) 5 min. Insets in (b) and (d) show higher magnifications views of respective images. sented in the previous chapter. Representative cyclic voltammograms of the LBL7 electrode are shown in Figure 8.3a with a insignificant resistive behaviour for scan rates up to 1,000 mV s−1 . As expected, due to the carboxylfunctionalization of the SWNTCOOH layers and carboxylic groups present in the sodium cholate (surfactant for graphene exfoliation), there was a pseudocapacitive contribution to the overall capacitance evidenced by the presence of typical redox peaks of carboxylic groups in 1 M H2 SO4 at 0.4 V (vs Ag/AgCl). Capacitance, energy density, and power density were normalized per unit volume to account for differences in thickness of the electrodes. As shown in Figure 8.3b, the total capacitance achieved by the LBL7 electrode was 113 F cm−3 at 10 mV s−1 which was intermediate between the capacitance achieved by the SWNTCOOH electrode (130 F cm−3 , 10 mV s−1 ), and the 237 graphene electrode (20 F cm−3 , 10 mV s−1 ). As shown in Figure 8.3c, the capacitance retention of the SWNTCOOH electrode was improved by the intervening graphene films in the LBL7 electrode; whereas the capacitance of a SWNTCOOH electrode almost completely faded at 4,000 mV s−1 (15 % capacitance retention), the LBL electrode maintained 48 % of its maximum capacitance. Similar capacitive performance (132 F cm−3 ) was reported for LBL electrodes made of multi-walled carbon nanotubes (MWNTs) functionalized with amine groups and carboxylic groups [1] . However, scan rates only up to 50 mV s−1 were reported. The Ragone plot in Figure 8.3d shows that the LBL7 electrode had an improved energy density performance when compared with the graphene electrode. The LBL7 electrode achieved an energy density of 16 Wh L−1 that remained comparatively stable (12 Wh L−1 ) up to a power of 35.3 kW L−1 (800 mV s−1 ). In contrast, the SWNTCOOH electrode had an energy density of 18 Wh L−1 which was stable only up to a power of 6.7 kW L−1 (100 mV s−1 ), after which the energy density dropped dramatically. The advantages of the LBL manufacturing were evident by comparing the performance of the LBL7 electrode versus that of a composite electrode. Exfoliation of graphene in the presence of previously suspended SWNTCOOH in water was carried out to give a pre-mixed suspension then spray-deposited to manufacture a graphene/SWNTCOOH 50 %/ 50 % w/w composite electrode with a thickness of 123 nm. Figure 8.4 shows that the LBL7 electrode provided with a greater volumetric capacitance than the composite electrode for all scan rates, e.g. 113 F cm−3 > 45 F cm−3 at 5 mV s−1 . It is possible that an improved utilization of electrochemically active materials is favoured 238 Figure 8.3: (a) Cyclic voltammograms of a LBL electrode and its (b) capacitance, (c) capacitance retention , and (d) Ragone plot compared with that of SWNTCOOH and graphene electrodes. Dashed lines in (b) and (c) show performance at 4 V s−1 of the different electrodes. by the LBL manufacturing approach - ensuring the laying down of uniform layers of minimum thickness maximizing density (mass per unit volume) and as a result capacitance per unit volume - as compared to a composite electrode where electroactive materials are randomly distributed and where density and contributions to charge storage depend on the weight combination ratio, and final electrode micro/nanostructure and porosity. The densiy of the LBL7 electrode was esimated to be 1.3 g cm−3 which was greater than 1.0 g cm−3 estimated for the graphene electrode, 0.7 g cm−3 estimated for the SWNTCOOH electrode, and 0.83 g cm−3 reported for LBL MWNT electrodes [5] . 239 Figure 8.4: Capacitance versus scan rate for a composite electrode graphene/SWNTCOOH (50 %/ 50 % w/w) and a LBL7 electrode. 8.4 Conclusions and future work A LBL electrode combining SWNTCOOH and graphene showed a capacitance of 113 F cm−3 that was superior to that of a discrete layer graphene only electrode with a capacitance retention of 48 % at 4,000 mV s−1 . Similarly, the LBL electrode showed an improved energy density of 16 Wh L−1 that was comparatively stable up to a power density of 35.3 kW L−1 (800 mV s−1 ) whereas the SWNTCOOH electrode showed a energy density of 18 Wh L−1 stable only up to a power of 6.7 kW L−1 (100 mV s−1 ). A LBL electrode showed a higher volumetric capacitance as compared to that of a composite electrode which was attributed to an enhanced density of uniform and minimum thickness layers of electrochemically active materials. This work demonstrates that the combination of graphene with other materials providing a double layer capacitance or pseudocapacitance using a LBL manufacturing approach can successfully produce a hybrid electrode of: (1) increased power density provided by the graphene layers, and (2) and 240 increased energy density provided by intervening layers of material of higher capacitance but otherwise poorer power capability. The LBL manufacturing approach offers the possibility to produce hybrid supercapacitor electrodes of improved performance in a wide range of combinations with a variety of pseudocapacitive materials. 241 Bibliography [1] S. W. Lee, B.-S. Kim, S. Chen, Y. Shao-Horn and P. T. Hammond, J. Am. Chem. Soc., 2008, 131, 671–679. [2] K. C. Krogman, N. S. Zacharia, S. Schroeder and P. T. Hammond, Langmuir, 2007, 23, 3137–3141. [3] N. Fukao, K.-H. Kyung, K. Fujimoto and S. Shiratori, Macromolecules, 2011, 44, 2964–2969. [4] G. M. Nogueira, D. Banerjee, R. E. Cohen and M. F. Rubner, Langmuir, 2011, 27, 7860–7867. [5] S. W. Lee, N. Yabuuchi, B. M. Gallant, S. Chen, B.-S. Kim, P. T. Hammond and Y. Shao-Horn, Nat. Nanotechnol., 2010, 5, 531–537. [6] D. Yu and L. Dai, J. Phys. Chem. Lett., 2009, 1, 467–470. [7] H. Zheng, F. Tang, Y. Jia, L. Wang, Y. Chen, M. Lim, L. Zhang and G. Lu, Carbon, 2009, 47, 1534–1542. [8] S. W. Lee, J. Kim, S. Chen, P. T. Hammond and Y. Shao-Horn, ACS Nano, 2010, 4, 3889–3896. 242 [9] K. C. Krogman, J. L. Lowery, N. S. Zacharia, G. C. Rutledge and P. T. Hammond, Nat. Mater., 2009, 8, 512–518. 243 Chapter 9 Conclusions and future work A scalable and environmentally friendly spray deposition technology has been demonstrated for the manufacture of up to 1,500 cm2 binder-free, flexible, thin film nanostructured electrodes. Large area electrodes in a variety of materials including carbon nanotubes and graphene with uniform mass load, uniform nanometric thickness and uniform and attractive electrochemical properties have been successfully manufactured. The spray technology was also versatile in producing laboratory scale electrodes in a variety of electrochemically active materials here tested with a throughput of 12 to 20 identical samples manufactured at the same time. Furthermore, in order to produce electrodes combining two or more materials in a layer-by-layer design, the equipment was modified to incorporate two spray nozzles each handling a different suspension and working in an alternate fashion. Future improvement of the spray deposition equipment includes upgrading to enable safe handling of suspensions in organic solvents, which will increase the range of electrochemically active materials that can be spray 244 deposited, full automation of the layer-by-layer spray procedure, and further scale-up to a roll-to-roll near industrial scale. Future work should also include the characterization in full of 1,500 cm2 area electrodes in a two-electrode cell configuration - rather than half-cell tests used here for laboratory scale electrodes - assembled and tested in collaboration with industrial partners. Due to its low cost, and rich redox and intercalative chemistry, nanostructured α-MoO3 was synthesized using a hydrothermal method and spraydeposited electrodes were investigated for its potential application in supercapacitors. In the aqueous electrolyte 1 M H2 SO4 , α-MoO3 nanobelts showed a complex redox activity yielding a moderate capacitance but poor cycling stability. Study of the redox activity using electrochemical and surface characterization techniques led to the reduction of the electrochemical window, and therefore capacitance, to a potential range of most reversible redox activity improving the overall cycling stability up to 720 cycles with 100 % capacitance retention. The capacitance was improved by a combination of the semiconducting α-MoO3 with conductive single walled carbon nanotubes (SWNTs) in a nanocomposite. Future work could consider the improvement of the capacitance of α-MoO3 by increasing the effective surface area of the electrochemically active material either by further reducing the size of the α-MoO3 nanobelts or by trying alternative nanostructures that could strategically incorporate a conductive additive enabling an efficient utilization of otherwise poorly utilized electroactive sites of the semiconducting α-MoO3 . Turning the attention to the layered crystal structure of α-MoO3 nanobelts, its suitability as a Li-ion intercalation material was also investigated in a half cell electrode configuration in LiClO4 in propylene carbonate. There was 245 Li-ion intercalation activity only when combining α-MoO3 with SWNTs. Electrochemical characterization of the composite α-MoO3 /SWNT showed a combination of capacitive and diffusion controlled mechanisms contributing to charge storage. The electrochemical activity had a degree of irreversibility nonetheless with a high charge storage capacity of 697.7 C g−1 /348.7 F g−1 at low scan rates of 0.1 mV s−1 reduced to 75 F g−1 at 50 mV s−1 . Therefore, the composite could potentially be used as a supercapacitor electrode in a device with moderate power density. The ion intercalation activity controlled by diffusion contributed a charge storage of 463.2 C g−1 , that might indicate use as battery cathode. However, the stability upon cycling, tested by cyclic voltammetry and charge-discharge experiments, needs to be improved; the capacitance retention after 117 charge-discharge cycles at 500 mA g−1 was 46.7 %. EIS studies provided an insight into changes of ionic and electronic resistance of the MoO3 electrode as the lithiation progressed, however the use of complementary techniques such as XRD will be suitable to track at the same time the emerging existing phases and correlation among lithiation, phases, and changes in resistance. Future work could consider the assembly of a full cell device in a hybrid configuration where the cathode α-MoO3 /SWNT will be combined with a suitable anode that could be a form of carbon with high surface area, e.g. activated carbon or carbon aerogels, that needs to be stable in a complementary electrochemical window enabling an enhancement of the total energy density. Graphene electrodes of high power density were manufactured by a surfactantwater based exfoliation method followed by a scaleable spray-deposition pro246 cess. Cyclic voltammetry and galvanostatic charge-discharge experiments revealed a combination of electric double layer and pseudocapacitive behavior that, unlike the many graphene-oxide derived electrodes reported to date, was maintained to unusually high scan rates of 10,000 mV s−1 , reaching a maximum capacitance of 543 µF cm−2 and with a capacitive retention of 57 % at 10,000 mV s−1 . The performance of graphene electrodes was contrasted with carboxylated single walled carbon nanotubes that showed a sharp decrease in capacitance above 200 mV s−1 . Electrochemical impedance spectroscopy analysis showed a fast capacitor response of 17.4 ms for as manufactured electrodes which was further improved to 2.3 ms for surfactant-free 40 nm thick electrodes. A maximum energy density of 75.4 nWh cm−2 gradually decreased as power density increased up to 2.6 mW cm−2 . Graphene electrodes showed 100 % capacitance retention for 5,000 cycles at the high power scan rate of 10,000 mV s−1 . The combination of the aqueous exfoliation method to produce graphene and the spray deposition to produce flexible large area electrodes demonstrated the possibility for scalability at low cost. Hybrid electrodes were manufactured following a LBL manufacturing approach combining the high power density properties of graphene and the relatively high energy density properties of carboxyl-functionalized single walled carbon nanotubes. The LBL electrode showed an improved capacitance, energy density and power density performance when compared with similar electrodes of the individual components. Graphene can therefore serve as a component providing a high power density in hybrid electrodes that could incorporate a variety of double layer capacitive and pseudocapacitive materials. Future work should be focused on the manufacture and characterization 247 of hybrid electrodes with this approach which will offer a wide range of materials combinations. 248 Appendix A Complementary data to Chapter 7 Equivalent circuit used to fit high frequency electrochemical impedance spectroscopy (EIS) data Figure A1: Equivalent electrical circuit used to fit the high frequency EIS data of a graphene electrode. Elements are described as follows [1] : • Rs is the electrolyte resistance. • CPEf is a constant phase element representing the film capacitance. • Rf is the film resistance. • CPEdl is a constant phase element representing the double layer capacitance. • Rct is the charge transfer resistance of pseudocapacitive processes. • ZW is the impedance of a finite length Warburg element representing diffusion processes of the electrolyte ions into the film. 249 The impedance of a constant phase element ZCP E is defined as [2] , ZCP E = where i = √ 1 TCP E (iω)αCP E −1, ω is the angular frequency, TCP E is the capacitance when αCP E = 1, and αCP E is the constant phase exponent (0 ≤ αCP E ≤ 1). 250 The impedance of a finite length Warburg element Zw is defined as [2] , ZW = RW coth[(iTW ω)αW ] (iTW ω)αW where RW is the limiting diffusion resistance, TW = L2 /D where L is the effective diffusion length and D is the effective diffusion coefficient, and αW is a exponent between 0 and 1. Table A1: Parameters obtained by fitting the high frequency impedance data to the electrical equivalent circuit in Figure A1. Parameter DC potential Fit frequency range Rs CPEf Rf CPEdl Rct ZW χ2 Value 0.396 V (vs. Ag/AgCl) 200 kHz - 398 Hz 7.86 Ω TCP E = 5.12 x 10−9 F cm−2 , αCP E = 1 9.45 Ω TCP E = 5.87 x 10−6 Ω−1 sα , αCP E = 0.98 1.0 Ω RW = 45.84 Ω , TW = 7.7 x 10−3 s, αW = 0.81 3.76 x 10−5 251 Investigation of cyclability via galvanostatic charge-discharge studies. Figure A2: Galvanostatic charge-discharge curves of a graphene electrode recorded for 1000 cycles at (a) 50 µA cm−2 , and (b) 100 µA cm−2 ; and (c) capacitance retention. 252 Fast capacitor response of a 40 nm thick graphene electrode described by EIS Figure A3: EIS of a graphene electrode of 40 nm thickness after removal of surfactant by annealing: (a) Phase angle versus frequency showing a -72 ◦ impedance phase angle at 120 Hz, (b) Nyquist plot showing a nearly pure capacitive behaviour. 253 Bibliography [1] M. D. Obradovic, G. D. Vukovic, S. I. Stevanovic, V. V. Panic, P. S. Uskokovic, A. Kowal and S. L. Gojkovic, J. Electroanal. Chem., 2009, 634, 22–30. [2] W. Sugimoto, H. Iwata, K. Yokoshima, Y. Murakami and Y. Takasu, J. Phys. Chem. B, 2005, 109, 7330–7338. 254