Carbon-based materials as supercapacitor electrodes

Transcription

1 / Journal Homepage / Table of Contents for this issue TUTORIAL REVIEW Chemical Society Reviews Carbon-based materials as supercapacitor electrodes Li Li Zhang and X. S. Zhao* Received 9th March 2009 First published as an Advance Article on the web 12th June 2009 DOI: /b813846j This tutorial review provides a brief summary of recent research progress on carbon-based electrode materials for supercapacitors, as well as the importance of electrolytes in the development of supercapacitor technology. The basic principles of supercapacitors, the characteristics and performances of various nanostructured carbon-based electrode materials are discussed. Aqueous and non-aqueous electrolyte solutions used in supercapacitors are compared. The trend on future development of high-power and high-energy supercapacitors is analyzed. 1. Introduction Climate change and the limited availability of fossil fuels have greatly affected the world economy and ecology. With a fast-growing market for portable electronic devices and the development of hybrid electric vehicles, there has been an ever increasing and urgent demand for environmentally friendly high-power energy resources. Supercapacitors, also known as electrochemical capacitors or ultracapacitors, have attracted much attention because of their pulse power supply, long cycle life ( cycles), simple principle, and high dynamic of charge propagation. 1 3 As illustrated in Fig. 1, where various energy conversion and storage devices are compared and presented in the simplified Ragone plot, supercapacitors occupy an important position in terms of the specific energy as well as the specific power. With a high power capability and relatively large energy density compared to conventional capacitors, supercapacitors offer a promising approach to meet the increasing power demands of energy storage systems in the twenty first century. Currently, supercapacitors are Department of Chemical and Biomolecular Engineering, National University of Singapore, 4 Engineering Drive 4, , Singapore. chezxs@nus.edu.sg; Fax: ; Tel: widely used in consumer electronics, memory back-up systems and industrial power and energy management. 4 A more recent application is the use of supercapacitors in emergency doors on the Airbus A380, highlighting their safe and reliable performance. One of the most promising applications is their use in low-emission hybrid electric vehicles, and fuel cell vehicles. In such cases, supercapacitors are coupled with primary high-energy batteries or fuel cells to serve as a temporary energy storage device with a high-power capability to store energies when braking. Supercapacitors are likely to show an equal importance to batteries for future energy storage systems. Generally, on the basis of the energy storage mechanism, supercapacitors can be classified into two categories. 5 One is the electrical double layer capacitor (EDLC), where the capacitance comes from the pure electrostatic charge accumulated at the electrode/electrolyte interface, therefore it is strongly dependent on the surface area of the electrode materials that is accessible to the electrolyte ions. The other category is the pseudo-capacitor, in which fast and reversible faradic processes take place due to electro-active species. These two mechanisms can function simultaneously depending on the nature of the electrode materials. Progress towards supercapacitor technologies can benefit from the continuous Li Li Zhang Li Li Zhang received her Bachelor s degree in Chemical and Biomolecular Engineering (first class honours) from the National University of Singapore in After two years of experience in industry, she took up a PhD scholarship to work under the supervision of Professor X. S. Zhao. Her research project deals with designing novel carbon-based electrode materials for electrochemical energy storage applications. X. S. Zhao X. S. Zhao received his PhD in Chemical Engineering from the University of Queensland in 1999 under the supervision of Professor Max Lu. After two-years of postdoctoral researching training at the same university, he joined the Department of Chemical and Biomolecular Engineering, National University of Singapore, as an Assistant Professor, and then later became an Associate Professor. His main research interest focuses on template synthesis of porous materials and the exploitation of their applications in emerging areas Chem. Soc. Rev., 2009, 38, This journal is c The Royal Society of Chemistry 2009

2 Fig. 1 Ragone plot showing the specific power against specific energy for various electrical energy storage systems (r The Nature Publishing Group, reprinted with permission). 6 development of nanostructured materials. Readers can refer to a number of recent reviews, which discuss the scientific and technological improvements of supercapacitors of various configurations and electrodes. 1,2,5 8 Carbon-based materials ranging from activated carbons (ACs) to carbon nanotubes (CNTs) are the most widely used electrodes because of the often-cited desirable physical and chemical properties. These properties include low cost, variety of form (powders, fibers, aerogels, composites, sheets, monoliths, tubes, etc.), ease of processability, relatively inert electrochemistry, controllable porosity and electrocatalytic active sites for a variety of redox reactions. 9,10 In the development of EDLCs, a proper control over the specific surface area and the pore size adapted to an appropriate type of electrolyte solution are crucial to ensure a good performance of the supercapacitor in terms of both power delivery rate and energy storage capacity. ACs are often considered as EDLC electrode materials because of their high surface area and somewhat controllable pore size, depending on the method of activation (chemical or physical activation). Since Knox and co-workers pioneered the template synthesis of porous carbons, 11 many porous carbon materials with well controlled and uniform pore sizes have been designed and synthesized. 12 Templated porous carbons of microporous, mesoporous and macroporous sizes with a tailorable hierarchical structure hold a great promise as supercapacitor electrode materials. On the other hand, CNTs are of particular interest for the development of supercapacitor electrodes because of their unique tubular porous structures and superior electrical properties, which favor fast ion and electron transportation. More recently, the rise of graphene, which is a new class of two-dimensional carbon nanostructure, having large specific surface area and exceptionally high electronic qualities, is forecasted to have enticing potential in supercapacitor applications if large-scale production of graphene sheets is possible. 13 In developing high-capacitance supercapacitors, the pseudocapacitance due to the presence of foreign electro-active species in the porous carbon framework is coupled with the EDLC to enhance the overall capacitance of the electrode materials. The most commonly studied species having quick and reversible faradic reactions include oxygen- and nitrogencontaining surface functional groups, electrically conducting polymers, and transition metal oxides (e.g., RuO 2, MnO x, etc.). The capacitive performance and some physical properties of various carbon and carbon-based electrode materials are summarized in Table 1, which is based on a literature survey. The use of high-capacitance materials (high surface area or pseudo-active species) is a key factor to ensure high energy density; whereas a high electrical conductivity (thus a fast charge transportation) of the electrode materials as well as the electrolyte solution is necessary for a high rate capability. Hence, one of the most critical aspects in the development of supercapacitors is to optimize the energy density without deteriorating their high power capability as these two parameters determine the ultimate performance of the supercapacitor. In this tutorial review, we provide a brief summary of recent research progress on carbon-based electrode materials and the importance of the electrolyte solutions in the development of supercapacitor technologies. 2. Principle of energy storage in supercapacitors 2.1 The energy storage mechanism Supercapacitors are electrochemical energy storage devices. The way that supercapacitors store energy is based, in principal, on two types of capacitive behavior: the electrical double layer (EDL) capacitance from the pure electrostatic charge accumulation at the electrode interface, and the pseudo-capacitance developed from fast and reversible surface redox processes at characteristic potentials. It is more convenient to discuss the two mechanisms separately, though they usually function together in a supercapacitor The energy storage mechanism of EDLCs. Conventional capacitors store little by way of energy, due to the limited charge storage areas and geometric constrains of the separation distance between the two charged plates. However, supercapacitors based on the EDL mechanism can store much more energy because of the large interfacial area and the atomic range of charge separation distances. As schematically illustrated in Fig. 2a, the concept of the EDL was first described and modeled by von Helmholtz in the 19th century, when he investigated the distribution of opposite charges at the interface of colloidal particles. 14 The Helmholtz double layer model states that two layers of opposite charge form at the electrode/electrolyte interface and are separated by an atomic distance. The model is similar to that of two-plate conventional capacitors. This simple Helmholtz EDL model was further modified by Gouy and Chapman 15,16 on consideration of a continuous distribution of electrolyte ions (both cations and anions) in the electrolyte solution, driven by thermal motion, which is referred to as the diffuse layer This journal is c The Royal Society of Chemistry 2009 Chem.Soc.Rev., 2009, 38,

3 Table 1 Properties and characteristics of various carbon and carbon-based materials as supercapacitors electrode materials Materials Specific surface area/m 2 g 1 Density/g cm 3 Aqueous electrolyte Organic electrolyte /F g 1 /F cm 3 /F g 1 /F cm 3 Carbon materials Commercial activated carbons (ACs) o 200 o 80 o 100 o 50 Particulate carbon from SiC/TiC o o 70 Functionalized porous carbons o o 90 Carbon nanotube (CNT) o 60 o 60 o 30 Templated porous carbons (TC) o o 100 Activated carbon fibers (ACF) o o 120 Carbon cloth Carbon aerogels o 80 o 80 o 40 Carbon-based composite materials TC RuO 2 composite CNT MnO 2 composite AC polyaniline composite Fig. 2 Models of the electrical double layer at a positively charged surface: (a) the Helmholtz model, (b) the Gouy Chapman model, and (c) the Stern model, showing the inner Helmholtz plane (IHP) and outer Helmholtz plane (OHP). The IHP refers to the distance of closest approach of specifically adsorbed ions (generally anions) and OHP refers to that of the non-specifically adsorbed ions. The OHP is also the plane where the diffuse layer begins. d is the double layer distance described by the Helmholtz model. j 0 and j are the potentials at the electrode surface and the electrode/electrolyte interface, respectively. (see Fig. 2b). However, the Gouy Chapman model leads to an overestimation of the EDL capacitance. The capacitance of two separated arrays of charges increases inversely with their separation distance, hence a very large capacitance value would arise in the case of point charge ions close to the electrode surface. Later, Stern 17 combined the Helmholtz model with the Gouy Chapman model to explicitly recognize two regions of ion distribution the inner region called the compact layer or Stern layer and the diffuse layer (see Fig. 2c). In the compact layer, ions (very often hydrated) are strongly adsorbed by the electrode, thus the name of compact layer. In addition, the compact layer consists of specifically adsorbed ions (in most cases they are anions irrespective of the charge nature of the electrode) and non-specifically adsorbed counterions. The inner Helmholtz plane (IHP) and outer Helmholtz plane (OHP) are used to distinguish the two types of adsorbed ions. The diffuse layer region is as what the Gouy Chapman model defines. The capacitance in the EDL (C dl ) can be treated as a combination of the capacitances from two regions, the Stern type of compact double layer capacitance (C H ) and the diffusion region capacitance (C diff ). Thus, C dl can be expressed by the following equation: 1 ¼ 1 þ 1 ð1þ C dl C H C diff The factors that determine the EDL behavior at a planar electrode surface include the electrical field across the electrode, the types of electrolyte ions, the solvent in which the 2522 Chem. Soc. Rev., 2009, 38, This journal is c The Royal Society of Chemistry 2009

4 electrolyte ions are dissolved, and the chemical affinity between the adsorbed ions and the electrode surface. As the electrode is usually a porous material with a high specific surface area, the EDL behavior at the pore surface of the porous electrode is more complex than that at an infinite planar one as the ion transportation in a confined system can be drastically affected by a number of parameters, such as the tortuous mass transfer path, the space constrain inside the pores, the ohmic resistance associated with the electrolyte, and the wetting behavior of the pore surface by the electrolyte. Fig. 3 schematically illustrates an EDLC made of porous carbon electrode. The capacitance estimation for an EDL-type supercapacitor is generally assumed to follow that of a parallel-plate capacitor: 9 C ¼ e re 0 d A ð2þ where e r is the electrolyte dielectric constant, e 0 is the permittivity of a vacuum, A is the specific surface area of the electrode accessible to the electrolyte ions, and d is the effective thickness of the EDL (the Debye length). Based on eqn (2), there should be a linear relationship between specific capacitance (C) and specific surface area (A). However, some experimental results have shown that this simple linear relationship does not hold. 18,19 Traditionally it was believed that the submicropores of an electrode do not participate in the formation of EDL due to the inaccessibility of the submicropore surfaces to the large solvated ions. However, Raymundo-Pinero and co-workers 20 recently observed the important contributions of micropores to the overall capacitance and suggested that partial desolvation of hydrated ions occurred, leading to an enhanced capacitance. Very recently, Simon and Gogotsi 6 reported an anomalous capacitance increase in carbon electrodes with pore sizes less than 1 nm. The authors observed the maximum EDL capacitance when the electrode pore size was very close to the ion size. These results further confirmed a capacitance contribution from pores with sizes smaller than the solvated ion size. However, these new experimental findings cannot be fully interpreted by the EDL theory because in such confined micropore spaces, there would be insufficient room to accommodate both the compact layer and diffuse layer. Recently, Huang and co-workers 21 have reported a heuristic approach to describing nanoporous carbon-based supercapacitors. In their approach, the pore curvature is taken into account and a different capacitive behavior was proposed depending on the size of the pore. An electric double-cylinder capacitor (EDCC) model is used for describing mesoporous carbon electrodes, while an electric wire-in-cylinder capacitor (EWCC) model is proposed for modelling microporous carbon electrodes, as illustrated in Fig. 4. When the pores are large enough so that the pore curvature is no longer significant, the EDCC model can be reduced natually to the traditional planar EDL model given in eqn (2). The capacitance estimation for the two proposed models are given in eqn (3) for the EDCC model and eqn (4) for the EWCC model, respectively: C ¼ C ¼ e r e 0 b ln½b=ðb dþš A e r e 0 b lnðb=a 0 Þ A in which b is the pore radius, d is the distance of approaching ions to the surface of the carbon electrode, and a 0 is the ð3þ ð4þ Fig. 3 Schematic representation of an EDLC based on porous electrode materials. Fig. 4 Schematic diagrams (top views) of (a) a negatively charged mesopore with solvated cations approaching the pore wall to form an electric double-cylinder capacitor and (b) a negatively charged micropore of radius b with cations lining up along the pore axis to form an electric wire-in-cylinder capacitor (r Wiley InterScience, reprinted with permission). 21 This journal is c The Royal Society of Chemistry 2009 Chem.Soc.Rev., 2009, 38,

5 effective size of the counterions (that is, the extent of electron density around the ions). With these models, the authors were able to make good fits of mathematical results obtained using eqn (3) and (4) to experimental data, regardless of the types of carbon materials and electrolytes employed. Both the anomalous increase in capacitance for pores sizes below 1 nm, and the trend for slightly increased capacitance with an increase in pore size above 2 nm, can be explained by fitting with the proposed EWCC and EDCC models, respectively. In addition, the dielectric constant calculated from the fitting results using eqn (4) is close to the vacuum value, which indicates that desolvation of hydrated ions occurs before entering the micropores. In spite of the recent experimental and theoretical advancements, there still lacks of an in-depth understanding of the EDL charge storage mechanism in the nano-confined spaces. More theoretical and fundamental studies on what happens and how the charge is stored inside the micropores are desired for future EDLC development The energy storage mechanism of pseudo-capacitance. In contrast to EDL capacitance, pseudo-capacitance arises for thermodynamics reasons and is due to charge acceptance (Dq)and a change in potential (DV). 5 The derivative C =d(dq)/(ddv) corresponds to a capacitance, which is referred to as the pseudo-capacitance. The main difference between pseudocapacitance and EDL capacitance lies in the fact that pseudo-capacitance is faradic in origin, involving fast and reversible redox reactions between the electrolyte and electro-active species on the electrode surface. The most commonly known active species are ruthenium oxide, 22 manganese oxide, 23 vanadium nitride, 24 electrically conducting polymers such as polyaniline, 25 and oxygen- or nitrogencontaining surface functional groups. 26 While pseudo-capacitance can be higher than EDL capacitance, it suffers from the drawbacks of a low power density (due to poor electrical conductivity), and lack of stability during cycling. A good example of a material providing a pseudo-capacitive property is ruthenium oxide. Due to its intrinsic reversibility for various surface redox couples 5,27 and high conductivity, the electrochemical behavior of both amorphous and crystalline forms of ruthenium oxide has been widely studied in acidic electrolyte solution over the past few decades. It was shown that amorphous hydrous ruthenium oxide (RuO 2 xh 2 O) exhibited a much higher specific capacitance (720 F g 1 ) than anhydrous ruthenium oxide. 28 This has been attributed to the mixed proton electron conductivity within RuO 2 xh 2 O, as the superfacial redox transitions of ruthenium oxide involve proton and electron double injection/expulsion according to the following reaction: 5,22 RuO x (OH) y + dh + + de 3 RuO x d (OH) y+d (5) where RuO x (OH) y and RuO x d (OH) y+d represent the interfacial oxyruthenium species at higher and lower oxidation states. In a proton-rich electrolyte environment (e.g. H 2 SO 4 ), the faradic charges can be reversibly stored and delivered through the redox transitions of the oxyruthenium groups, i.e., Ru(IV)/Ru(III) and Ru(III)/Ru(II). Based on the mean electron transfer numbers, the theoretical specific capacitance of RuO 2 xh 2 O is estimated to range from ca to 2200 F g A very high specific capacitance of 1300 F g 1 was reported for a pure bulk nanostructured RuO 2 xh 2 O electrode. 22 The superior electrochemical performance is related to its tailored nanotubular arrayed porous architecture with metallic conductivity as well as its hydrous nature. As implied by eqn (5), the reversible redox transitions depend on both proton exchange and electron-hopping processes. The tubular arrayed porous structure and metallic conductivity provided an effective pathway for electrolyte and electron transportation. In addition, the hydrous nature of RuO 2 xh 2 O ensures a high rate of proton exchange because the surface of the hydrous oxide has been considered to be a proton liquid. 30 As a result, the designed nanostructure demonstrated a promising electrode material for energy storage. However, the cost of the precious metal oxide and difficulties in large scale production limit its practical applications. 2.2 The performance of supercapacitors There are fundamental differences between batteries and supercapacitors in terms of energy storage mechanism and electrode materials. Therefore, the characteristic performance of supercapacitors that sets them apart from batteries should be maintained for any technology improvement. The performance of supercapacitors is mainly evaluated on the basis of the following criteria: (1) a power density substantially greater than batteries with acceptably high energy densities (4 10 Wh kg 1 ); (2) an excellent cyclability (more than 100 times that of batteries); (3) fast charge/ discharge processes within seconds; (4) low self-discharging; (5) safe operation, and (6) low cost. The basic operation of a single-cell supercapacitor consisting of two electrodes is illustrated in Fig. 5 below in a simple equivalent RC circuit representation. C a and C c are the capacitance of the anode and cathode, respectively (the specific capacitance of a single electrode reported in the literature is based on a threeelectrode cell configuration). R s is the equivalent series resistance (ESR) of the cell. R F is the resistance responsible for the self-discharge of a single electrode and the notation for anode and cathode are R Fa and R Fc, respectively. The total capacitance of the cell (C T ) is therefore calculated according to: 1 ¼ 1 þ 1 ð6þ C T C a C c In an RC circuit, the time constant (t) expressed as resistance (R) capacitance (C) gives important information on the characteristic of a capacitor. For example, the time constant Fig. 5 A simple RC equivalent circuit representation illustrates the basic operation of a single-cell supercapacitor Chem. Soc. Rev., 2009, 38, This journal is c The Royal Society of Chemistry 2009

6 for the self-discharge of the anode is equal to R Fa C a and hence a larger value of R Fa is desirable for smaller leakage of the anode. The maximum energy stored (E) and power delivered (P) for such a single cell supercapacitor are given in eqn (7) and (8): E = 1 2 C TV 2 (7) P ¼ V2 ð8þ 4R s where V is the cell voltage (in volts), C T is the total capacitance of the cell (in farads) and R s is the ESR (in ohms). Each element inside the equivalent circuit is crucial to the final performance of the supercapacitor. The capacitance of the cell depends extensively upon the electrode material. The cell voltage is limited by the thermodynamic stability of the electrolyte solution. ESR comes from various types of resistance associated with the intrinsic electronic properties of the electrode matrix and electrolyte solution, mass transfer resistance of the ions in the matrix, and contact resistance between the current collector and the electrode. Hence, for a supercapacitor to have a good performance, it must simultaneously satisfy the requirements of having a large capacitance value, high operating cell voltage and minimum ESR. It is thus obvious that the development of both electrode materials as well as the electrolyte solutions is required in order to optimize the overall performance of the supercapacitor. With regard to the electrode materials, the development of high surface area carbon electrodes with a defined pore size distribution, and the incorporation of pseudo-active materials to optimize the overall capacitance and conductivity without destroying the stability, are currently hot research areas. High density electrode materials will give a high volumetric energy which is desirable for practical applications, especially for the design of a highly compact energy and power source. On the other hand, electrolyte development, especially non-aqueous electrolytes with lower resistivity are highly desirable for their higher power and energy density owing to higher operating voltages (up to V). In the following sections, an overview of recent developments on carbon-based electrode materials as well as electrolytes will be discussed. 3. Carbon-based electrode materials 3.1 Activated carbons (ACs) ACs are the mostly widely used electrode materials due to their large surface area, relatively good electrical properties and moderate cost. ACs are generally produced from physical (thermal) and/or chemical activation of various types of carbonaceous materials (e.g. wood, coal, nutshell, etc.). Physical activation usually refers to the treatment of carbon precursors at high temperature (from 700 to C) in the presence of oxidizing gases such as steam, CO 2 and air. Chemical activation is usually carried out at lower temperatures (from 400 to 700 1C) with activating agents like phosphoric acid, potassium hydroxide, sodium hydroxide and zinc chloride. Depending on the activation methods as well as the carbon precursors used, ACs possessing various physicochemical properties with well developed surface areas as high as 3000 m 2 g 1 have been produced and their electrochemical properties have been studied ,31 34 It is well known that the porous structure of ACs produced by activation processes have a broad pore size distribution consisting of micropores (o2 nm), mesopores (2 50 nm) and macropores (450 nm). Several researchers have pointed out the discrepancy between the capacitance of the ACs and their specific surface area. With a high surface area up to 3000 m 2 g 1, only a relatively small specific capacitance o 10 mf cm 2 was obtained, much smaller than the theoretical EDL capacitance (15 25 mf cm 2 ), 5 indicating that not all pores are effective in charge accumulation. 32 Therefore, although the specific surface area is an important parameter for the performance of EDLC, some other aspects of the carbon materials such as pore size distribution, pore shape and structure, electrical conductivity and surface functionality can also influence their electrochemical performance to a great extent. Furthermore, excessive activation will lead to large pore volume, which results in the drawbacks of low material density and conductivity. These would in turn cause a low volumetric energy density and loss of power capability. In addition, high active surface areas may increase the risk of decomposition of the electrolyte at the dangling bond positions. 34 The presence of the acidic functionalities and moisture on the surface of ACs is responsible for the aging of the supercapacitor electrodes in organic electrolytes. 35 Efforts have been made to search for the relationship between the nanoporous structure of ACs and their capacitance performance in different electrolytes. In general, the capacitance of ACs is higher in aqueous electrolytes (ranging from 100 F g 1 to 300 F g 1 ) than in organic electrolytes (less than 150 F g 1 ). One reason for this is believed to be the larger effective size of the electrolyte ions in organic solutions when compared with those in water. Organic electrolytes increase the number of pores that are smaller than the ions, and therefore increase the number not contributing to the charge storage. The wettability of the carbon surface by the organic electrolyte, which is determined by the chemical affinity between the two may be another reason. Salitra and co-workers have concluded that a pore size above 0.4 nm can be active in EDL charging in aqueous solution. 33 The studies by the Beguin group also concluded that the optimal pore size for EDL capacitance is 0.7 nm in aqueous media and 0.8 nm in organic electrolytes. 20 All these findings have shown the essential role of micropores that are electrochemically accessible by the electrolyte ions on the capacitance performance. Most recent studies by Largeot et al. have clearly demonstrated the relationship between ion size and pore size for an EDLC using carbide-derived-carbons (CDCs) and concluded that the maximum EDL capacitance is achieved when the pore size matches with the ion size. 36 The pore size-dependent capacitive behavior is shown in Fig. 6. Besides the porous structure of ACs, the surface functionalities also play important roles on the carbon electrode performance, as they will affect the wettability of the carbon surface by the electrolyte ions and give additional pseudo-capacitance. 10,26,34 An activated carbon containing a high concentration of oxygen-functional groups with low This journal is c The Royal Society of Chemistry 2009 Chem.Soc.Rev., 2009, 38,

7 Fig. 6 Normalized capacitance change as a function of pore size for CDC samples prepared at different temperatures. Ionic liquid EMI TFSI was used as the electrolyte and the maximum capacitance was obtained at a pore size that matched the maximum ion dimension (r The American Chemical Society, reprinted with permission). 36 porosity (BET specific surface area = 270 m 2 g 1 ) has been prepared by one-step carbonization of an oxygen-rich carbon precursor. 34 The material showed a good electrochemical performance with a high energy density of 10 Wh kg 1 even at high power densities close to 10 kw kg 1 in acidic electrolyte. Moreover, good capacitance values were still retained after cycles, indicating few catalytically active sites on the carbon surface, probably due to the comparatively low surface area. However, most commercial supercapacitors use organic electrolytes because of the higher operating voltages, which offer a higher energy density. Pandolfo and Hollenkamp have given an overview of ACs with various types of functional groups and pointed out that the presence of some active surface oxides and a trace amount of water result in instability of the electrode, an increase of series resistance and the decomposition of the organic electrolyte. 10 Some other studies also showed the aging of AC electrodes in organic electrolytes due to the active surface of carbon. 35,37 Therefore, surface functionality and porosity of ACs should be optimized for improved long-term performance. Pure ACs without heteroatoms and water might be better for use in an organic electrolyte. In short, ACs have been commercially used as supercapacitor electrode materials. However, their applications are still restricted to only certain niche markets due to the limited energy storage and rate capability. Although activated carbons provide a high surface area, the control of pore size distribution and pore structure is still challenging. Therefore, designing ACs to have narrow pore size distribution (accessible to the electrolyte ions) with an interconnected pore structure and short pore length together with controlled surface chemistry would be beneficial for enhancing the energy density of supercapacitors, without deteriorating their high power density and cycle life. 3.2 Carbon nanotubes (CNTs) The discovery of CNTs has significantly advanced the science and engineering of carbon materials. Fig. 7 shows the atomic structures of different types of carbon nanotubes. As discussed previously, the major factor that determines the power density Fig. 7 Atomic structure of carbon nanotubes, (a) schematic diagram showing how a graphene sheet is rolled according to a pair of chiral vectors to form different atomic structures of carbon nanotubes, (b) zig-zag (n,0) (c) chiral (n,m) and(d) armchair(n,n) carbon nanotubes. is the overall resistance of the components in a supercapacitor. Carbon nanotubes, due to their unique pore structure, superior electrical properties, and good mechanical and thermal stability, have attracted a great deal of attention for supercapacitor electrode applications. 23,38 41 CNTs can be categorized as single-walled carbon nanotubes (SWNTs) or multi-walled carbon nanotubes (MWNTs), both of which have been widely explored as energy storage electrode materials. CNTs are usually regarded as the choice of a high-power electrode material because of their good electrical conductivity and readily accessible surface area. Moreover, their high mechanical resilience and open tubular network make them a good support for active materials. The energy density is, however, a concern due to their relatively small specific surface area (generally o 500 m 2 g 1 ) as compared to ACs. Of greater importance is the difficulty in retaining the intrinsic properties of individual CNTs on a macroscopic scale 39 and the high purity and electrolyte-dependent capacitance performance. 40 Niu et al. have reported a MWNT-based supercapacitor electrode showing a high specific capacitance of 102 F g 1 with surface area of 430 m 2 g 1 and a power density of 8 kw kg 1 in an acidic electrolyte. 38 It should be taken into account that recent studies have shown that entangled CNTs are less efficient in facilitating fast ionic transportation when compared to aligned CNTs, due to the irregular pore structures and high entanglement of the CNT structure in the former. 42 Hence, the use of aligned CNT seems to be more advantageous in terms of power performance. Futaba and co-workers have presented a method to fabricate a densely packed aligned SWNT solid by using the zipping effect of liquids, which allowed the bulk 2526 Chem. Soc. Rev., 2009, 38, This journal is c The Royal Society of Chemistry 2009

8 materials to retain the intrinsic properties of the SWNTs. 39 The energy density of the obtained SWNT solid was about 35 Wh kg 1 (normalized to a cell consisting of two identical electrode materials) in organic electrolyte, and the rate capability was better than that of ACs. These studies showed the importance of the aligned tubular structures and preserved intrinsic CNT properties on the electrochemical performance of the electrode materials. Research has been undertaken to improve the energy density of CNTs by increasing their specific surface area via chemical activation (KOH activation). 10 However, a proper balance between the porosity and the conductivity must be achieved in order to have both high capacitance and good rate performance. Very recently, an interesting CNT aerogel composite material was synthesized by uniformly dispersing a carbon aerogel throughout the CNT host matrix without destroying the integrity or reducing the aspect ratio of the CNT. 41 A high specific surface area of 1059 m 2 g 1 and extremely high specific capacitance of 524 F g 1 were obtained, but at the expense of a tedious preparation pathway. Another way to enhance the specific capacitance is by modification of CNTs with active materials to realize pseudo-capacitance through faradic processes, which will be described later. In spite of having excellent properties, the limited surface area of CNTs restricted their use as high energy performance EDLCs. In addition, the present difficulty in purification and high cost of production still hinder their practical applications. 3.3 Templated carbons A templating method offers another effective way to produce nanostructured carbons with well controlled narrow pore size distributions, ordered pore structures, large specific surface areas and an interconnected pore network, making them promising candidates for supercapacitor electrode materials. There has been remarkable progress in the synthesis of ordered nanostructured carbons through templating techniques. 43 In general, the preparation procedure of templated carbons is infiltration of a carbon precursor into the pores of the template, followed by a carbonization treatment and finally the removal of the template to leave behind a porous carbon structure. The schematic representation in Fig. 8 shows the general concept of the templating technique and the porous carbons obtained from different templates. Various carbon structures with well controlled micropores, mesopores and/or macropores produced from different types of template and carbon precursors have been studied for supercapacitor applications ,9 A functionalized microporous carbon material was obtained by using zeolite Y as a template and the resultant carbon material possessed a high gravimetric capacitance of about 340 F g 1 in aqueous electrolyte with good cyclability (over cycles). 44 Compared to ACs, whose micropores are essentially disordered and broad in pore-size distribution, the templated microporous carbons with a narrow pore size distribution, well adapted pore size to the electrolyte ions and the ordered straight pore channels are better for use as high-energy-density electrode materials. Moreover, the well controlled porous Fig. 8 (a) Macroscopic representation showing the general concept of the templating technique; microscopic synthesis of (b) macroporous carbons using silica spheres as template, (c) mesoporous carbons using SBA-15 as template and (d) microporous carbons using zeolite Y as template. structure facilitated an efficient use of pseudo-capacitance from the nitrogenated and oxygenated functionalities of the carbon materials. It has been proposed that the presence of mesopores (usually 2 8 nm) can accelerate the kinetic process of the ion diffusion in the electrodes and improve the power performance at high current densities, whereas micropores that are accessible to the electrolyte ions are essential for high energy storage. 9,21,46 Wang and co-workers have reported a 3-dimensional (3D) hierarchical porous graphitic carbon (HPGC) material with macroporous cores, mesoporous walls and micropores for high rate supercapacitor applications. 46 In their designed hierarchical structure, macropores serve as ionbuffering reservoirs; the graphitic mesopore walls provide excellent electrical conductivity, capable of overcoming the primary kinetic limits of electrochemical process in porous electrodes. In addition, the presence of micropores can enhance the charge storage. The schematic representation of the 3D structure and the performance of the carbon materials are shown in Fig. 9. Yamada s group have also synthesized ordered porous carbons containing meso/macro/micropores with large surface areas by a colloidal-crystal templating technique. 47 A high EDL capacitance of F g 1 was achieved in an acidic electrolyte solution. Their investigation on pore-dependent capacitance properties have revealed that the micropores This journal is c The Royal Society of Chemistry 2009 Chem.Soc.Rev., 2009, 38,

9 Fig. 9 (a) Schematic representation of the 3D hierarchical porous structure (HPGC) and (b) Ragone plot comparing the performance of HPGC material with those of CMK-3, CMK-5, activated carbon (Maxsob, Japan), ALG-C, PVA porous carbon, and small-pore EC. The dotted lines show the current drain time (r Wiley InterScience, reprinted with permission). 46 adjacent to the open mouths of pores are effective in charge storage, and larger pores that are interconnected are important for smooth electrolyte transportation. Hence, carbon materials having well controlled hierarchical porous structures would be favorable for designing high performance EDLC electrodes. The templating method has been shown to be one of the most suitable techniques to control the preparation of porous materials, especially ordered porous solids. 43 Through careful selection of the template materials as well as the carbon precursors, and with good control over the carbonization process, nanoporous templated carbons with desirable physical and chemical properties can be obtained. In view of their relatively high cost of production, development of a simple, economical and environmentally benign template route would be advantageous for their future application. Despite the cost, templated carbons are good materials to study, which provide valuable information about the effect of pore size, pore shape, channel structures and other parameters on the ion diffusion and charge storage in the nanoconfined system. 3.4 Other carbon structures Other carbon structures such as activated carbon fibers (ACFs), carbon aerogels (CAGs) and carbon onions have also been studied for supercapacitor applications. The general rules for the selection of supercapacitor electrode materials are a high and accessible specific surface area with good electrical conductivity. ACFs typically have high specific surface areas, up to 3000 m 2 g 1, and a more or less controllable pore size distribution. 48 They are usually produced from the carbonization of pre-formed fibrous carbon precursors followed by activation processes. A high capacitance value of 371 F g 1 in KOH electrolyte solution was reported recently. However, such a high specific surface area together with the presence of surface functional groups may cause a long term stability problem. 6 In addition, the cost of ACF production is generally higher than AC. CAGs are another interesting material suitable for use as a supercapacitor electrode. They are ultralight, highly porous materials, predominantly with mesopores, and have the possibility of usage without binding substances. CAGs are typically fabricated through a sol gel process with subsequent pyrolysis of the organic aerogels. The special porosity of a CAG is based on the interconnection of colloidal-like carbon nanoparticles. Due to the dominance of mesopores (42 nm), the EDL capacitance is not so effective in CAG-based electrode materials. Hence, an additional activation process was employed to further increase the specific surface area by introducing microporosity. Despite a large increase in the total surface area (from 592 to 2371 m 2 g 1 ), the improvement on the charge storage capacity was relatively small, especially at high discharge rates, 49 which is primarily due to the inaccessibility of the micropores produced during the activation and the relatively high internal resistance of the CAG matrix. Nevertheless, a new type of carbon nanotube aerogel electrode material gave promising capacitive properties, despite the difficulty in preparation Carbon-based composites Materials with pseudo-capacitive properties are of great interest in the development of supercapacitors with enhanced energy storage capacity. However, most of the pseudo-capacitance comes from surface redox reactions and hence only a very thin surface layer is active in the faradic processes. In addition, the poor electrical conductivity of pure transition metal oxides such as MnO 2, and the mechanical degradation of the conducting polymers during faradic charging processes, hinder their applications as electrode materials. Therefore, fabrication of composites by using carbon as a support not only increases the effective utilization of the active materials, but also improve the electrical conductivity and mechanical strength of the composite materials. Several studies have reported the improved capacitive properties of the composite materials using various types of carbons, such as carbon monolith, 25 CNT 23,50 and templated carbon 51, as a support. Hierarchically porous carbon monoliths were used as a support for the electrodeposition of electroactive conducting polymers, i.e. polyaniline (PANI) to prepare carbon PANI composite electrode materials. 25 A high capacitance value about 300 F g 1 was obtained for the composite material and a very high capacitance F g 1 was calculated based on the weight of PANI in the carbon PANI composite. A more interesting and important finding is that the hierarchically porous carbon monolith served as an effective support for electrodeposition of pseudo-active material, leading not only to an enhanced capacitance value with a high rate capability, but also good cyclability. The improved stability and the high pseudo-capacitance value of the PANI in the composite materials are of great interest for supercapacitor electrode applications Chem. Soc. Rev., 2009, 38, This journal is c The Royal Society of Chemistry 2009

10 Fig. 10 (a) Schematic representation of manganese oxide CNTA composite material and (b) specific capacitance of manganese oxide CNTA, manganese oxide entangled CNT (ECNT), manganese oxide AC and CNTA versus discharge current density (r The American Chemical Society, reprinted with permission). 23 It is known that CNTs serve as an effective support for transition metal oxides and conducting polymers due to their open mesoporous tubular structure and superior electrical and mechanical properties. Zhang and co-workers have reported the use of a carbon nanotube array (CNTA), which is directly connected to the current collector (Ta foil) as the support to make composite electrodes with hierarchical porous structures. Manganese oxide 23 and PANI 50 were selected as pseudo-active materials to prepare manganese oxide CNTA and PANI CNTA composite materials, respectively. The electrochemical studies have shown that a very high specific capacitance of about 1000 F g 1 was obtained for the PANI CNTA composite material with a high rate capability (95% capacity retention at 118 A g 1 ) and good stability (up to 5000 cycles test). The manganese oxide CNTA composite also possessed moderate specific capacitance of 101 F g 1 at high current density of 77 A g 1 and with good cyclability. In addition, the high density of the manganese oxide CNTA composite material leads to a high volumetric capacitance which is suitable for compact electrode design. Fig. 10 shows a schematic representation of the manganese oxide CNTA composite materials and their electrochemical performance at different rates compared with other carbon substrates. The high rate and energy performance of the composite materials are due to the superior electrical properties of the substrate, i.e. CNTA grown directly on the current collector, as well as the pseudo-capacitive behavior of the active materials. Hence, fine-tuning the structure of manganese oxide CNTA and/or investigation on the stability and capacitive properties of PANI CNTA in non-aqueous electrolytes may further improve the performance of the composite electrode materials. However, the high cost of making such sophisticated nanostructures is of concern. Templated ordered mesoporous carbon (OMC) was also 51 frequently used as a support for metal oxides such as RuO 2 52 and MnO 2 to prepare carbon metal oxide composite materials for supercapacitor applications. The OMC-based composite materials not only improve the electrical conductivity of the metal oxide (MnO 2 ), but also increase the number of active sites for pseudo-active species to achieve a high level of utilization of pseudo-capacitance. In addition, the presence of interconnected mesopores, ordered channel structures and a relatively high specific surface area, simultaneously provide an efficient EDL capacitance, leading to an improved energy performance. Dong and co-workers have prepared a MnO 2 OMC composite and they have shown that a high capacitance value over 200 F g 1 was obtained for the composite material and 600 F g 1 was calculated based on MnO 2 content in the composite. 52 Li s group have investigated the electrochemical performance of RuO 2 OMC composite materials. 51 They found out that by increasing the amount of RuO 2 (up to 30 wt%), the overall capacitance of the composite material reached a value as high as 633 F g 1. However, the utilization rate of the precious metal oxide was decreased due to the limited reaction of the large RuO 2 particles. This also highlights an important point that although metal oxides possess high pseudo-capacitance values, their real contribution to the total charge storage depends strongly on the surface utilization of the active materials, which could be achieved by dispersing nano-sized particles well onto highly ordered highsurface carbon substrates. Simon and Gogotsi in their recent review article proposed that, one possible strategy to improve both energy and power densities for supercapacitors is to achieve conformal deposition of pseudo-capacitive materials onto highly ordered high-surface-area CNTs or ACs. 6 Thus, future research could be directed to enhance the surface activity/interaction between the high-surface-area carbon substrate with the thin or nano-sized active materials to improve the faradic processes across the interface, so as to achieve efficient pseudo-capacitance usage on top of the EDL capacitance. 4. The electrolytes The performance of a supercapacitor is not only dependent on the electrode materials, but is also strongly affected by the electrolytes employed. A high cell operating voltage provides both high energy density and power density, but is limited by the stability of the electrolyte in the applied potential. The current trend in supercapacitor development involves the switch from an aqueous electrolyte, which has a low operating voltage of 1 V (due to the thermodynamic decomposition of water) to a non-aqueous medium which allows a much higher voltage window of about 2.5 V. The most recent supercapacitors available in the market use electrolytes based on aprotic solvents, typically acetonitrile or carbonate-based solvents (i.e. propylene carbonate). 3 Fig. 11 clearly shows the advantages of a higher energy density when using aprotic electrolyte rather than aqueous electrolyte. 53 However, there are other considerations on the use of non-aqueous electrolytes such as high cost, low conductivity compared to aqueous electrolyte leading to power deterioration, low dielectric constant resulting in smaller capacitance, complex purification procedure as well as safety concerns due to the flammability and toxicity of the organic solvent. Ionic liquids (ILs), which are known as room temperature molten salts, are under considerable research for use as next generation electrolytes. 54,55 They are attractive candidates for electrolytes used in energy storage because of their good chemical and physical properties, such as high thermal stability, high electrochemical stability over a wide potential window (43 V), non-toxicity, non-flammability, variety of combination This journal is c The Royal Society of Chemistry 2009 Chem.Soc.Rev., 2009, 38,