Carbon Nanotubes for Energy Applications

2.1.1. Supercapacitor operation and types

There are two types of electrochemical capacitors that are referred to as 1) electric double layer capacitors (EDLC) and 2) pseudocapacitors. The construction of these devices can vary, with electrodes being fabricated from porous carbon materials including activated carbons, graphene, carbon nanotubes, templated carbons, metal oxides and conducting polymers [18, 19]. EDLC or supercapacitors have two electrodes immersed in an electrolyte solution, separated by a semi-permeable dielectric that allows the movement of ions to complete the circuit but prevents a short circuit from being formed. EDLCs are advantageous as they are able to provide relatively large power densities and larger energy densities than conventional capacitors, and long life cycles compared to that of a battery and ordinary capacitor [20]. The performance of supercapacitors is affected by the power density requirements, high electrochemical stability, fast charge/discharge phenomena, and low self-discharging [21]. Table 1 below shows a comparison between the three types of devices.

Energy storage is achieved by the build-up and separation of electrical charge that is accumulated on two oppositely charged electrodes as shown in Figure 2 [12]. As stated previously, no charge transfer takes place across the electrode-electrolyte interface and the current that is measured is due to a rearrangement of charges. The electrons involved in the non-Faradaic electrical double layer charging are the conduction band electrons of the electrode. These electrons leave or enter the conduction band state depending on the energy of the least tightly bound electrons or the Fermi level of the system [2]. Supercapacitors exhibit very high energy storage efficiencies exceeding 95 % and are relatively stable for up to10⁴-10⁵ cycles [4, 5]. The energy given by the equation, E = 0.5CV², means that the operating voltage is the key in determining the energy characteristics of a supercapacitor. The choice of electrolyte when designing and fabricating a supercapacitor device dictates the operating voltage [23]. Operating voltages are approximately 1.2V, 2.7V, and 3.5V respectively for aqueous, organic and ionic liquid with all of them having associated advantages and disadvantages [4, 5].

Like EDLC, a pseudocapacitor consists of two porous electrodes with a separator between them all immersed in an electrolyte solution [24]. However, the difference is that the charge is accumulated during Faradaic reactions near to or at the surface of the electrodes [18], hence non-Faradaic double-layer charging and Faradaic surface processes occur simultaneously [25]. The pseudocapacitance arises from a Faradaic reaction when some of the charge (q) passed in an electrode process is related to the electrode potential (V) via thermodynamical considerations [26]. The two principal cases are adsorption pseudocapacitance arising in underpotential deposition processes [26], and homogeneous redox pseudocapacitance where the reaction is reversible [26, 27]. Pseudocapacitors thus combine features of both capacitors and batteries [18, 28]. A comparison of energy density and power density for various electrical energy storage systems is depicted in Figure 3. Current commercial uses of supercapacitors include personal electronics, mobile telecommunications, back-up power storage, and industrial power and energy management [29, 30]. A recent application is the use of supercapacitors in emergency doors on the Airbus A380, highlighting their safe and reliable performance [30].

2.2.1. Aqueous electrolytes

For aqueous electrolytes, the maximum operating voltage is theoretically limited by the reduction potential of water (1.23 V at room temperature) [32]. Most aqueous electrolyte systems tend to have an electrochemical window of approximately 1 V [32]. Electrolyte conductivity has a significant effect on the equivalent series resistance (esr) of the cell, which determines the power output [4]. Concentrated electrolytes are required to minimise the esr and maximise power capability [32]. In general, strong solutions of acidic electrolytes are much more corrosive than strong basic electrolytes meaning that the electrolyte has to be carefully selected for the particular electrode material.

Aqueous electrolytes tend to have very good kinetic behaviour of the electrolyte ions leading to very efficient charge/discharge rates. This behaviour is due to the relatively high conductivity and low viscosity of the concentrated solutions [4, 33]. For example, the conductivity of 1M H₂SO₄ is 730 mS cm^-1 compared to the much lower value of 10-20 mS cm^-1 for organic solutions of lithium salts [34]. Time constants of symmetrical carbon supercapacitors using H₂SO₄ were reported to be 0.1s [34].

2.2.2. Organic electrolytes

The use of non-aqueous electrolytes in supercapacitors has the main advantage of higher operating voltages compared to aqueous systems. Voltage windows can range up to 2.5 V and since the stored energy increases as V², it is possible to attain large energy and power densities [32, 34]. It must be noted that to operate at these higher voltages, non-aqueous electrolytes must be free of water and oxygen which will ensure no evolution of O₂ and/or H₂O at potential differences above 1.23 V [35]. Salts are added to the system to provide mobile ion movement at the electrode/electrolyte interface. The most common salt used generally consists of lithium ions as these ions move very well under an electric field and the effective ion diameter is very small [17].

The major disadvantages of non-aqueous systems are the lower conductivity and the higher viscosity resulting in higher equivalent series resistance (ESR) and reduced wettability if the electrode is hydrophilic. A decrease in wettability will effectively reduce the surface area used by the electrolyte, reducing the energy and power density. Most commercial systems that use organic electrolytes are manufactured in inert atmospheres and are costly to be produced [4].

2.2.3. Ionic liquids

Ionic liquids are another class of electrolytes that is proving a great area of research for electrolytes in supercapacitors. These electrolytes can be considered as molten salts with melting temperatures usually below room temperature where the ionic conductivity is no more than 20 mS cm^-1 [34].

Common ions include BF₄ ^-, PF₆ ^-, (CF₃SO₂)₂N^-, CF₃SO₃ ^- as well as imidazolium, pyridinium, and quaternary ammonium salts [36]. The physical properties depend on the type of anion and cation and the alkyl chain length [37]. The main advantages are the good solvating properties, relatively high conductivity, non-volatility, low toxicity, large potential window, negligible vapour pressure and good electrochemical stability [37, 38]. Disadvantages include high viscosities and low conductivities compared to that of aqueous electrolytes; while some ionic liquid mixtures yield a potential window that is not much greater than that of aqueous systems. Capacitances approaching 100 F/g for activated carbon (AC) electrodes have been reported by Frackowiak et. al. by using (CF₃SO₂)₂N^- anions and phosphoniumcations. Balducciet et. al. reported capacitance values of 115 F/g for asymmetric poly(3-methylthiophene)/AC electrodes using 1-buytl-3-methyl-imidazolium ionic liquids [36, 37].

2.3.1. CNT synthesis overview

There are a variety of different methods for making SWNTs and MWNTs that have been developed since CNTs were first discovered. These include laser ablation, arc discharge, chemical vapour deposition (CVD), and high pressure carbon monoxide disproportionation (HiPCO). Recently, work by Harris et. al. has successfully scaled-up the synthesis of CNT using a fluidised bed reactor [41]. All growing conditions for synthesising CNTs require a catalyst to achieve high yields, where the size of the catalyst nanoparticles will determine the diameter and chirality of the CNT [42]. The CNTs that are formed are generally in a mixture with other carbonaceous product including amorphous carbon and graphitic nanoparticles.

2.3.2. Main synthesis methods for CNT growth

Both Laser ablation and arc-discharge methods for the growth of CNTs involve the condensation of carbon atoms generated from the evaporation of carbon sources. High temperature is involved, ranging from 3000 ^oC - 4000^o C [43]. In arc discharge, various gases such as Helium or Hydrogen are induced into plasma by large currents generated at a carbon anode and cathode. This process leads to the evaporation of carbon atoms which produces very high quality MWNTs and SWNTs [44, 45]. Laser ablation also produces very high quality CNTs with a high degree of graphitisation by focusing a CO₂ laser (in pulsed or in continuous wave mode) for a period of time onto a rotating carbon target [46]. The HiPCO process utilises clusters of Fe particles as catalysts to create very high quality SWNTs [47]. Catalyst is formed in situ by thermal decomposition of iron pentacarbonyl, which is delivered intact within a cold CO flow and then rapidly mixed with hot CO in the reaction zone. Upon heating, the Fe(CO)₅ decomposes into atoms that condense into larger clusters. SWNTs nucleate and grow on these particles in the gas phase [48, 49].

The CVD method usually consists of a furnace, catalyst material, carbon source, a carrier gas, a conditioning gas, and a collection device (usually a substrate). The carrier gas is responsible for taking the reacting material onto the substrate where CNT growth occurs at catalyst sites [43]. The components mentioned are essential; however, different groups and researchers have alternative experimental conditions which can contain multiple types of furnaces, and a variety of catalyst and carbon sources. The key advantage of this technique is its capability to directly deposit the CNTs onto the substrate, unlike arc discharge and laser ablation that produces a soot / powder. Recent developments by Harris et. al. [41] has led to the development of a large scale batch process for fabricating MWNTs. Here, a furnace like system called a fluidised bed reactor continuously flows a carrier gas over a porous alumina powder that is impregnated with the catalyst material, leading a to continuous creation of MWNTs where tens of grams can be synthesised in one run.

2.3.3. Single walled nanotubes

SWNTs have been studied extensively as a supercapacitor and hybrid energy material [4, 50, 51]. The structure of a SWNTis illustrated in Figure 4 with a cylindrical nature apparent as previously stated. Its advantage is that it has very good thermal and conductive properties where the thermal conductivity can exceed 6000 Wm^-1K^-1 and a potential current carrying capacity of 10⁹ A/cm² [52, 53].

The maximum reported gravimetric capacitance for SWNT fabricated electrodes (PVA / PVC binder; pressed into pellet) is 180 F/g with an energy density of 7 Wh/kg and a power density of 20 kW/kg using KOH electrolyte [54, 55]. Hu et. al. [56] have recently reported a solid state paper based SWNT supercapacitor, which has a specific capacitance of 115 F/g, energy density of 48 Wh/kg and a large operating voltage of 3V. The electrode preparation involved pre-processing where cotton sheets were immersed in the SWNT dispersion, annealed then immersed in an PVA/ H₃PO₄ electrolyte.

2.3.4. Multi-walled nanotubes

Like their SWNT counter parts, MWNTs have also been studied extensively as electrode materials for supercapacitors [4, 50, 51]. The advantages over SWNTs are their ability to be more easily synthesised on much larger scales, making them more suitable for commercial application. The concentric nature of MWNTs can be observed in the SEM image of Figure 5. The maximum gravimetric capacitance attained for electrodes constructed from MWNTs range between 4-140 F/g with the best available commercial result at 130 F/g from Maxwell’s Boost capacitor [59]. Wang et. al. [50] have recently reported partially exfoliated MWNTs on carbon cloth that gave a specific capacitance in the range of 130-165 F/g with a coulombic efficiency of 98 %.

2.3.5. Surface functionalities

The presence of surface functionalities such as oxygen, nitrogen, hydrogen, boron and catalyst nanoparticles (dependent on the synthesis environment and pre-cursor materials) can affect the capacitative behaviour of the electrode through the introduction of Faradaic reactions [60], changes in electric and ionic conductivity [23], and influencing wettability [61]. A schematic representation of an sp² hybridised carbon lattice with various dopants is show in Figure 6.

Oxygen

Carbon materials will have functional groups present on their surface as a result of the precursors and preparation conditions [23]. Most of these functional groups are in the form of –COOH, =CO as well as phenol, quinone and lactone groups [4, 23]. Activation procedures such as post treatment with H₂SO₄ and / or HNO₃ also leads to acid oxygen functionalities [4].

Most of these groups are bonded with carbon atoms at the edge of hexagonal carbon layers where Faradaic reactions (via interactions with the electrolyte) lead to pseudocapacitance such as those developed with transition metal oxides RuO₂ and MnO₂ [23, 62]. These functional groups can also be purposely added onto the surface of carbons via oxidation with O₂ or acid treatment with HNO₃ or H₂SO₄ [63]. In aqueous systems, the presence of oxygen containing functional groups can lead to an enhanced wettability as well as pseudocapacitance as mentioned above, which maximises the electroactive surface area leading to larger energy densities [8, 23]. It has been proposed that the pseudocapacitative reactions for oxygen functionalised CNTsinvolve carboxyl groups undergoing electron transfer[64]:

C − OH    − −    C = O  +  H + +  e − C = O  +  e −   − −    C  –  O −

In non-aqueous systems, however, oxygen functionalities are detrimental to device performance. Parasitic redox reactions can lead to a degradation of the electrode, as well as adverse effects relating to voltage proofing and increased leakage current [4, 65]. These redox reactions will reduce the cycle life of a device, as well as lowering the operating voltage.

Shenet. al. [66] reported in 2011 the effects of changing the carboxylic group concentration on SWNTs. The specific capacitance, power density and energy density 0.5 M H₂SO₄ electrolyte increased with carboxylic group density reaching a maximum of 149.1 F/g, 304.8 kW/kg, and 20.71 Wh/kg, respectively. The 10 µm film electrodes were fabricated using vacuum filtration to create “bucky papers” onto a mixed cellulose estate membrane.

Nitrogen

Nitrogen containing carbons have recently attracted interest due to its n-type behaviour that promotes large pseudocapacitance, which can be obtained even if the surface area of the material is decreased [67]. In some instances, up to 3-fold increase in capacitance have been reported [68]. Typical examples of redox reactions involving nitrogen are described below [69]:

C + NH  +  2e − +  2H + ↔  CH 2 − NH 2 CH − NHOH  + 2e − +  2H + ↔  CH 2 − NH 2 +  H 2 O

The chosen precursor material affects the types of functionalities that are attached to the carbon backbone. Nitrogen-containing groups may be added via various methods with compounds containing nitrogen including treatment with urea, melamine, aldehyde resins and polyacylonitrile [4, 70-73]. Surface areas for nitrogen-doped carbon materials are thought to be in excess of 400 m²/g [23]. This is much lower than pure SWNTs and pure MWNTs that have been reported to attain a surface area greater than 1315 m²/g and 830 m²/g respectively; suggesting that pseudocapacitance makes up a substantial portion of the total capacitance [74, 75]. Y. Zhang et. al. [76] have showed that N-doped MWNTs synthesised via CVD growth exhibited a capacitance of 44.3 F/g, which was more than twice the value obtained than that of the un-doped MWNTs in a 6M KOH electrolyte. K. Lee et. al. [77] have shown that the nitrogen content on vertically aligned CNTs increases the capacitance until a certain point due to an increased donation of an electron by the N (N acts as an n-type dopant) and an enhanced wettability in aqueous systems. Excessive N-doping significantly reduced the conductivity and inhibited charge storage and delivery [77]. The doped and un-doped CNTs were directly grown onto a stainless steel substrate using CVD [77].

Boron

Boron is another interesting material for doping CNTs due to its p-type nature which promotes CNT growth and increases the oxidation temperature of the nanotubes [78]. However, the development of boron doped CNTs for the use as electrodes in supercapacitor devices is not well established [23]. Work by Shiraishi et.al. showed that boron doping MWNTs, increased the capacitance per surface area from 6.5 μF/cm² to 6.8 μF/cm² in 0.5M LiBF₄/PC [79]. These electrodes were once again synthesised using CVD [79]. Wang et. al. reported in 2008 that interfacial capacitance was increased by 1.5-1.6 times in boron-doped carbon than that in boron-free carbon with alkaline electrolyte (6 M KOH) and/or acid electrolyte (1 M H2SO4) [80]. The carbon material was made into a slurry using carbon black and PTFE binder and pasted onto a Ni mesh current collector [80].

2.3.6. Advantages, limitations and comparison

It can be seen that CNTs can be tailored different ways in order to tune (to a degree) the performance of the electrode material. This control has been demonstrated by firstly, varying the chiralitiy of the nanotube to produce the single-walled or the multi-walled variety. Both CNT types have associated advantages and disadvantages with SWNTs being able to be synthesised with a high degree of purity; while MWNTs can be synthesised on a larger scale. CNTs can also have functionalities (through addition of oxygen or nitrogen containing groups) added to their structure through treatment in order to change the surface properties and hence wettability of the material. These functionalities enable enhanced compatibility to an electrolyte to maximise electroactive surface area usage and hence performance. Further doping with specific elements such boron and nitrogen can introduce a p-type/n-type behaviour where electrons contribute a Faradaic response to the system and enhance capacitance and energy density. However, it must be noted that when faradaic processes occur at the electrode/electrolyte interface, irreversible processes increase degradation of the electrode over time. Specific capacitance of CNTs (three electrode and device testing) is in the order of 5 – 165 F/g with an increase thereafter as a result of doping (i.e due to Faradaic contribution). It must be pointed out that with electrical energy devices, there is always a trade-off between energy and power density. Therefore the electrode material has to be tailored to meet the requirements of the specific application.

2.5.1. CNT / polymer

Electrode materials comprised of inherently conducting polymers (ICPs) and CNTs is a promising area of research. The conductive polymer matrix, combined with the network like structure of the CNTs provides an enhanced electronic and ionic conductivity that can considerably improve charge storage and delivery [86-88].

Antiohos et. al. reported a SWNT / Pedot-PSS composite electrode material that was fabricated into a device which had a specific capacitance of 120 F/g (1 M NaNO₃ / H₂O), coupled with an excellent stability (~90% capacity retention) over 1000 cycles using galvanostatic charge / discharge [89]. The SWNT / Pedot-PSS composite is depicted in Figure 9 where SWNTs are thoroughly dispersed throughout the Pedot-PSS conducting polymer matrix. Kim et. al. recently fabricated a ternary composite material consisting of MWNTs, graphene, and PANI where a specific capacitance of 1118 F/g was achieved. This electrode was stable with 85% capacity retention after 500 cycles using galvanostatic charge / discharge [90]. Hu et. al. [91] have recently reported a composite electrode materials containing MWNTs coated with polypyrrole that achieved a high capacitance of 587 F/g in a 0.1 M NaClO₄ / acetonitrile electrolyte.

2.5.2. CNT / metal oxide

Metal oxides exhibit pseudocapacitative behaviour over small rages of potentials, through redox processes which contribute electron transfer between the electrode / electrolyte interface (Figure 8). Common materials used in the construction of such devices are oxides of Mn, Ru, Ir, Pt, Rh, Pd, Au, Co and W [22, 26]. By combining metal oxides with CNTs, composites can be formed that combine both Faradaic and non-Faradic effects enabling a larger energy density to be obtained while still holding reasonable power density. Figure 10 shows MnO₂ particles that have been formed (insitu) in the presence of MWNTs.

Very recent work on carbon / metal oxide composites can be found in the review by Wang et. al. [50]. Myoungkiet. al. reported recently in their a RuO₂ / MWNT, electrode material which achieved a specific capacitance of 628 F/g[92]. The electrode was fabricated by dispersing the mixture in ethanol and casting onto carbon paper [92]. Li. et. al. reported that when MWNT were coated with MnO₂, a capacitance of 350 F/g was achieved [93]. More novel materials have been created by incorporating MWNTs and Co₃O_4,which yielded specific capacitances of 200 F/g [94] (acetylene black / PVDF slurry on Ni gauze); while Jayalakshmi et. al. reported in 2007 V₂O₅.xH₂O / CNT film with a specific capacitance of 910 F/gwith the material being ground into a paste with paraffin and spread onto a graphite electrode, and tested in 0.1 M KCl [95].

2.5.3. CNT / carbons

Creating composite materials from CNT and different forms of carbon such as graphene or carbide derived carbons (CDC) can be advantageous due to the fact that the CNTs provide microporosity (large surface area to maximise capacitance and hence energy density); while graphene and CDC can be used to tailor the mesoporosity which improve ions kinetics, enhancing the power density [21]. In Figure 11, a composite of reduced graphene oxide coated with SWNTs is depicted that has been formed into a porous film. The edges of the graphene oxide protrude out with a uniform coating of SWNTs. Recently, Li et. al. fabricated different mass loadings of graphene and CNT composite electrodes by solution casting onto glass, annealing then peeling off [97]. They reported capacitance ranges of 70-110 F/g at a scan rate of 1 mV/s in 1M H₂SO₄ [97]. Luet. al. have reported a CNT / graphene composite which was bound together with polypyrrole (through a filtration process) that achieved a specific capacitance of 361 F/g at a current density of 0.2 A/g in 1 M KCl. The electrode exhibited excellent stability with only a 4% capacity loss over 2000 cycles [97]. Dong et. al. have shown that is it possible to form SWNT/graphene oxide core shell structures and spray coat the subsequent material onto a current collector [98]. The performance of these core structures yielded a material with a specific capacitance of 194 F/g using galvanostatic charge / discharge at a high current density of 0.8 A/g in 1 M KOH [98].

3.1.1. Low grade heat sources and conversion through thermogalvanic cells

Various unharnessed low grade heat sources

The second law of thermodynamics dictates that a heat engine can never have perfect efficiency and will always produce surplus heat (usually around 100 ^oC). This waste heat (or low grade heat) is one of the world’s most ubiquitous sources of untapped energy. (i.e. waste heat is produced by simply turning on an automobile). Roughly 70 % of the energy generated by an automobile motor is wasted; part of it ends up as a hot exhaust pipe and warm brakes. The Wartsila-Sulzer RTA96-C turbocharged two-stroke diesel engine, one of the most efficient engines in the world, is only able to convert around 50 % of the energy in the fuel to useful motion. The rest of that energy gets dissipated as waste heat [102]. Waste heat also exists in factories, particularly in the steel and glass production industry. Pipes that carry hot liquids are also low grade heat sources. Other scenarios wherein heat simply dissipates into the environment are power plants, household appliances, and various electronic gadgets. Research done to convert waste heat into electrical power by the use of ferromagnetic materials, thermocouples and thermionic converters has resulted in low efficiencies [103-105]. Advances in thermoelectric systems have been hampered by its high initial cost and material limitations; as these systems operate in the temperature range much higher than low grade heat [106].

Description of how a thermogalvanic cell works

A thermogalvanic cell, also known as a thermocell, is a thermal energy converter that utilises electrochemical reactions to attain conversion of low grade heat to electrical power. The two half cells of the system are held at different temperatures causing a difference in redox potentials of the mediator at the anode and cathode [107]. This reaction can drive electrons through an external circuit that allows generation of current and power. A schematic of a thermocell with a ferri/ferrocyanide redox couple is shown in Figure 12.

Ferrocyanide is oxidized at the hot anode, the electron generated then travels through an external circuit and returns to the cell via the cold cathode where it is consumed in the reduction of ferricyanide[109]. The accumulation of reaction products at either half-cell is prevented by the diffusion and convection of the electrolyte that occurs naturally, thus eliminating the need for moving mechanical components.

3.1.2. Desirable material properties for thermal conversion cells

Power conversion efficiency and how it is affected by electrode material properties

The power conversion efficieny (Φ) of a thermocell is defined as follows:

The thermal power flowing through the cell is largely controlled by cell design and electrolyte selection. When a reversible redox couple is used, no net consumption of the electrolyte occurs and the thermal power is given by:

Where K is the thermal conductivity of the electrolyte, A is the electrode cross sectional area, ΔT is the thermal gradient and d is the distance between the two electrodes [110].

Qualitative behaviour of the current and voltage dependencies are shown in Figure 13; it depicts that the maximum electrical output power (P_max) is obtained when the external and internal loads are equal and is given by:

Where V_oc is the open circuit voltage and I_sc is the short circuit current. V_oc is highly dependent on the reaction entropy of the redox couple and the thermal gradient at which the electrodes are exposed to as shown in Equation 4:

where ΔS_B,A is the reaction entropy for a hypothetical redox couple A↔ ne^- B, n is the number of reactions involved in the redox reaction and F is Faradays constant [100].

Combining Equation 2 and Equation 3 allows the power conversion efficiency to be expressed as:

Ohmic, mass transport and activation overpotentials are losses that need to be minimised in order to realise an improvement in thermocell conversion efficiency. At large electrode separations, ohmic overpotential is dictated by the electrolyte resistance; and mass transport overpotential is maximized. By decreasing the inter-electrode separation, an increase in generated power will be observed as both ohmic and mass transport overpotentials will decrease. However, the power conversion efficiency will be lowered as it will be harder to maintain the thermal gradient in the cell [112]. It has been shown that changes in electrolyte concentration affect its thermal conductivity [108]. Optimization of electrolyte concentration coupled with appropriate cell design is necessary to mitigate both overpotentials while maintaining large power conversion efficiency.

Activation overpotential is associated with the activation barrier needed to transfer an electrode to an analyte. For the same activation overpotential, larger current densities are realised when the exchange current density is increased. This increase is attained when the concentration of the redox couple in the electrolyte is maximised, the thermal gradient is increased and the number of possible reaction sites is augmented [113]. Porous electrodes have the advantage of increased electroactive surface area and will directly amplify the short circuit current density [114]. It must be noted that for porous electrodes, short circuit current density does not increase indefinitely with electrode thickness as mass transfer overpotential will become limiting. The reaction products formed within the pores of the anode will not be able to diffuse fast enough to the cathode and vice versa, generating concentration gradients around both electrodes. Another way to decrease the activation overpotential in thermocells is by using catalytic electrodes attained by doping [115].

3.1.3. CNTs vs flat electrodes - why CNTs can improve thermal harvesting

Power conversion efficiency achieved by using flat electrodes and why recent developments in CNTs can augment thermogalvanic cell performance

The chemical stability of platinum led to its extensive investigation in thermogalvanic cells. In fact, a study on the effects of platinum electrode cleaning was performed and it was deduced that this affects the power delivery characteristics of thermogalvanic cells [116].

Comparison of thermoelectric converters operating at different conditions (i.e. thermal gradient, electrolyte, electrode separation, etc.) can be done by measuring their power conversion efficiency relative to a Carnot engine operating at the same temperature (Φ_r).

If power inputs are ignored, such as mechanical stirring, thermocells with platinum electrodes are able to attain power conversion efficiency relative to a Carnot engine of 1.2%. However, if power inputs are strictly excluded, the efficiency drops to 0.5% [100].

As mentioned previously, the discovery of CNTs led to widespread research on this material to investigate its potential uses, one of them being electrochemical applications [39, 117-120]. CNT electrodes are known to exhibit Nernstian behaviour and more importantly, fast electron transfer kinetics with the redox couple ferri/ferrocyanide. Peak potential separation in cyclic voltammograms obtained using micron-sized MWNT electrodes and 5 mM potassium ferrocyanide is 59 mV, which is the expected theoretical value and implies that the highest electron transfer rate was attained [121]. Incidentally, the ferri/ferrocyanide redox couple has been studied intensively in thermocell applications owing to the large voltage that can be induced by a thermal gradient, also known as the Seebeck coefficient. The Seebeck coefficient 1.4 mV/K for the ferri/ferro cyanide redox couple implies that an open circuit potential of 84 mV is attainable at a thermal gradient of 60 ^oC (the usual limit for aqueous systems without significant cooling). The fast electron transfer of CNTs in ferri/ferro cyanide primarily justifies its use as electrodes in thermocells.

The nanometre diameter of CNTs gives rise to large gravimetric and volumetric specific surface areas (SSA). Their unique aspect ratios allow porous electrodes to be fabricated by a variety of methods. Theoretically the SSA of CNTs can range from 50-1315 m²/g, the value dictated by the number of walls [74]. Theoretical predictions are in good agreement with experimental values obtained by the measurement of amount of gas (usually N₂) adsorbed at 77 K and calculations using the Brunauer-Emmett-Teller (BET) isotherm. Kaneko et al. [122] have reported that MWNTs are mesoporous while Rao et al. [123] have shown that SWNTs are microporous. MWNT buckypapers of the same geometric area compared with platinum foil are known to have three times larger charging current density during cyclic voltammetry in ferri/ferrocyanide aqueous electrolyte [114], evidence of the large accessible SSA of CNT electrodes.The large SSA of CNTs allows for a greater number of electroactive sites. When the CNT electrode porosity is controlled and the tortuousity is minimised in thermogalvanic cells (so that mass transfer is not limited within the electrode), the short circuit current generated can be significantly augmented.

3.2.1. SWNT and MWNT

CNTs were first used as thermocell electrodes in 2009 [114]. Baughman et al. tested 0.5 cm² MWNT buckypaper electrodes (with less than 1 % catalyst and with MWNT diameter of around 10 nm) in a U-Cell with electrode separation of 5 cm, a temperature gradient of 60 ^oC wherein T_cold= 5 ^oC. A schematic of the cell they used is shown in Figure 14. A specific power density of 1.36 W/m² was obtained [114]. Platinum electrodes tested under the same conditions generated a specific power density of 1.02 W/m², proving that CNTs are viable materials for thermocell electrodes.

SWNT powders produced by arc discharge (ASA-100F, Hanwha Nanotech) with an average diameter of 1.3 nm, and composition of 20-30 wt. % CNTs, 40 wt. % carbon nanoparticles, 20 wt. % catalyst material, 10 wt% amorphous carbon and graphite, was tested by Kang et al. in thermal harvesting [113]. A vertical test cell with a “hot above cold” orientation (Figure 15), glass frit separator and electrode separation of 4 cm was employed with T_hot= 46.4 ^oC and T_cold = 26.4 ^oC. SWNT electrodes with an area of 0.25 cm² were immersed in 0.2M K₃Fe(CN)₆/K₄Fe(CN)₆ electrolyte. The specific power density obtained was around 5.15 W/kg. Commercially available purified SWNT powders (P-SWNT) sourced from Hanwha Nanotech (ASP-100F), refined by thermal and acid treatment (60-70 wt. % nanotubes, 10 wt. % catalyst material, 20 wt. % graphite impurities) was tested by the same group. The specific power density improved by 32 %, generating 6.8 W/kg. Using the same test conditions, commercially available purified MWNT having 3-6 walls with a median diameter of 6.6 nm (SMW100, Southwest Nanotechnologies, Inc) and approximately 98 wt. % carbon yielded a specific power density of 6.13 W/kg. It must be noted that these tests were not done to maximise the power generation capability of the thermocell but to gain further insight into CNT electrodes for thermal harvesting. Hence the small electrode separation, low electrolyte concentration and small thermal gradient.

Electrical impedance spectroscopy (EIS) of the various carbon nanomaterials tested by Kang et al revealed that the P-SWNT electrode has a marginally lesser ohmicoverpotential (21 Ω) than the pristine SWNT (22 Ω). This finding explains the increased specific power density generated when the P-SWNT electrodes are used.

It has been proven that the catalytic nature of MWNTs is due to the edges or sites where the tube terminates, regions that are more numerous in MWNTs than in SWNTs [124]. Due to this and the fact that P-MWNTs had lower ohmic resistance (18 Ω) compared to P-SWNTs, it was expected that P-MWNTs would perform better in thermal harvesting. The authors attributed the enhanced performance of P-SWNT to the larger specific surface area, which compensated for the decreased electroactive sites and higher ohmic resistance.

3.2.2. Functionalized CNTs

Functionalising or doping (using Nitrogen or Boron atoms) may be used to fine tune the physical and chemical properties of CNTs [125-127]. With advances in technology, CNT functionalization is a reasonably simple process [128]. Dai et. al have shown that nitrogen-doped carbon nanotubes (NCNTs) have high electrocatalytic activity in oxygen reduction reactions as compared to undoped CNTs [129]. The increase in performance was brought about by a four electron pathway for oxygen reduction reactions that was attained by aligning the nanotubes and integrating nitrogen into the carbon lattice. The additional electrons contributed by nitrogen atoms can enhance electronic conductivity by providing electron carriers for the conduction band [130]. The many active defects and hydrophilic properties of NCNTs allow for enhanced electrolyte interaction in aqueous solutions [131]. Boron doped CNTs (BCNTs) are also attractive for electrochemical applications owing to the increased edge plane sites on the CNT surface; proven to be the predominant region for electron transfer [132]. Examples of the electrocatalytic performance of BCNTs are the improved detection of L-cysteine, enhanced electroanalysis of NADH and enhancement of field emission [126, 127,133].

The possibility of using nitrogen-doped CNT and boron-doped CNT electrodes in thermocells was investigated by Cola et.al [115]. Doping was attained by using a plasma-enhanced chemical vapour deposition process. Tests were run using a U-cell configuration (Figure 14), T_cold = 20 ^oC, thermal gradient of up to 40 ^oC, and electrolyte concentration of 0.1 M potassium ferri/ferrocyanide. The electrodes were sized to 0.178 cm² and were set up in a symmetric and asymmetric (N(hot)-B(cold) and B(hot)-N(cold)) fashion.

Results (Figure 16a) indicate inferior thermocell performance for both NCNT and BCNT as compared to Pt and pristine CNTs. The poor performance of the doped CNTs was brought about by the sluggish kinetics, evidenced by the large peak separations in the cyclic voltammograms taken at a scan rate of 100 mV/s (Figure 16b). It was theorized that the slow-moving kinetics for the doped electrodes was caused by the electrostatic effects at the electrode-electrolyte interface [134]. The positively charged BCNTs repulsed the similarly charged potassium (K⁺) counter ion which decreased the electrolyte concentration in the vicinity of the electrode. The negatively charged NCNTs led to a strong electrostatic attraction with K⁺, an effect which at low concentrations can improve electron transfer kinetics. However, the large bulk concentration (needed to achieve significant short circuit currents in thermocells) led to a high density of the K⁺ ions in the vicinity of the electrode. This effectively acted as a barrier to the redox reactions that were supposed to occur at the electrode.

Using BCNT and NCNT in an asymmetric configurations resulted in increased currents at small thermal gradients as compared to the symmetric arrangements. This current then decreased non-linearly as the temperature difference was increased. At small thermal gradients, with the BCNT at the cold side of the cell, the slower kinetics induced an accumulation of reactants at its surface. The faster kinetics at the NCNT, brought about by the increased temperature, kept the electrolyte concentration in its vicinity low. Both factors allow the redox reactions to occur rapidly until a threshold thermal gradient is reached. At this point, the ion concentration in the vicinity of the NCNT is sufficiently large to slow the kinetics and reduce the currents generated. The threshold temperature gradient is attained at lower temperatures when the NCNT is kept at the cold side of the cell because the slower kinetics at this temperature promotes accumulation of K⁺ ions on the NCNT surface and leads to the “blocking” effect discussed previously.

3.2.3. Composites

The recent discovery of graphene through micromechanical exfoliation has sparked a flurry of research into its possible applications [135]. Graphene consists of a single layer of carbon atoms bonded in a hexagonal lattice. Like CNTs, its remarkable properties (charge carrier mobility of 200000 cm²/V-s and specific surface area of 2630 m²/g) make it an ideal candidate for electrochemical applications [136, 137]. In order to scale up graphene production, graphite is normally exfoliated in the liquid state through surfactant/solvent stablilization [138] or chemical conversion resulting in a graphene like structure known as reduced graphene oxide (RGO), shown in Figure 17 [139]. Being of the same composition as CNTs, investigation of the possibility of synthesizing composites of these two carbon materials and exploring their performance as electrode materials has been done by several research groups [83, 91, 140].

Kang et al. have shown that when composites composed of 1:1 weight RGO and P-SWNT are used in thermocells, the specific power generated (5.3 W/kg) is comparable to that of SWNTs (the experimental conditions used were discussed in section 3.2.1) [113]. It must be noted that when RGO alone is used, the specific power generated is 3.87 W/kg. However, the composite electrode produces a specific power that is only 78 % of the P-SWNT. This decrease in performance can be attributed to the large ohmic resistance observed in the RGO electrode (35.6Ω) that is 55 % higher than the P-SWNT electrode. The chemical conversion of graphite to RGO involves oxidising graphite, exfoliation and then a subsequent reduction. It is surmised that the incomplete removal of the oxygen containing functional groups is the cause of the pronounced ohmic resistance. Another reason for the poor performance of the composite electrodes is the restacking of the RGO sheets during electrode preparation, which impedes electrolyte diffusion and results in sluggish kinetics.

Optimisation of the RGO-SWNT composition for thermocell electrodes was done by Chen et. al [141]. The amount of RGO added ranged from 1 to 20 % by weight. A U-cell was used with 0.75 cm² electrodes separated by 10 cm, a thermal gradient of 60 ^oC with T_cold= 20 ^oC and 0.4 M ferri/ferrocyanide electrolyte. The optimised composite 99 % SWNT-1 % RGO generated a specific current density of 26.78 W/kg. By using large amounts of SWNTs the RGO sheets were prevented from restacking, which resulted in the appropriate nanoporosity that promoted redox mediator diffusion. The sheet like structure of the RGO provided increased pathways for electrons in the composite thus contributing to its enhanced performance. The interaction between SWNTs and RGO is clearly seen in Figure 18.

3.3.1. Current processing techniques for CNTs relevant to thermocells

Solvent/surfactant exfoliation

One of the major obstacles to research on CNT characterisation and application is their spontaneous aggregation brought about by attractive van der Waals interactions in both aqueous and organic solutions. The resulting aggregates or “bundles” can reach lengths of several microns and diameters of tens of nanometers. Debundling of these aggregates is essential as they have mediocre properties as compared to individual tubes. Reproducibility of results also becomes an issue when CNTs are not dispersed adequately.

Liquid phase separation is one of the simplest methods wherein stable CNT dispersions are attainable. Stable well-exfoliated CNT dispersions is achieved by appropriate selection of the solvent as forced dispersion via ultrasonication will result in agglomeration of the CNTs in a very short span of time. Selection of solvents can be based on the enthalpy of mixing per solvent volume (Δ H - _mix ) [138]:

Where R b u n is the radius of the dispersed nanotube bundles, δ N T and δ s o l are square roots of the nanotube and solvent surface energies and ∅ is the nanotube volume fraction. The solubility theory states that a negative free energy of mixing ( ∆ G - m i x ) is indicative of a stable dispersion.

The entropy of mixing per unit volume ( ∆ S - m i x ) of nanotubes is generally small owing to their size and rigidity [142]. In order to realize a minimisation of ∆ G - m i x then solvents which result in small values of ∆ H m i x are necessary. Based on Equation 7, the most effective solvents at dispersing CNTs would be those that have a surface energy close to the nanotube surface energy (~70 mJ/m²); i.e. solvents with surface tension around 40 mJ/m² [143].

Another method to attain stable dispersions of CNTs is through the use of surfactants. Its inherent advantage over solvent dispersion was the fact that it was carried out in aqueous media, lessening its hazards and environmental impact. It relies on the principle wherein colloids are stabilized by surface charges [144]; i.e. Coulomb repulsion. Adsorption of the amphiphilic surfactant molecules onto CNTs is attained through their hydrophobic tails. This introduces a removable surface charge that creates an electric double layer around the nanotube; of which the magnitude and sign is proportional to its zeta potential [145]. This double layer provides repulsive forces that counteract the attractive van der Waals forces [146]. Selection of surfactants for CNTs dispersion depends on the size of their molecules. Low molecular weight surfactants will be able to pack tightly around the nanotube surface resulting in better stabilization [147].

Filtration of CNT dispersions results in a planar mat of randomly arranged tubes [148] that can be up to several hundred microns thick. These mats or “bucky papers” (Figure 19) have been instrumental in CNT evaluation (owing to their simple processing) as electrode materials not only for thermocells but for other electrochemical applications as well. Post-treatment of buckypapers via annealing or acid wash is essential to ensure complete removal of the solvent or surfactant used to attain the CNT dispersion [149].

Chemical vapour deposition

Chemical vapour deposition (CVD), a process that involves deposition of solids from a gas phase, has proven to be a viable method for attaining highly oriented CNTs on planar substrates (Figure 20) [150]. One of the theories behind the large degree of alignment is the reduction in free energy brought about by the van der waals interactions along the tube length inducing coordinated growth by holding the tubes together [151]. The ability to tailor the growth of CNTs in three dimensional configurations is highly advantageous in thermocell applications. This configuration promotes enhanced ion accessibility with the CNT matrix allowing larger current to be generated. The alignment of the tubes also minimises the tortuosity of the CNT electrodes which decreases the probability of forming concentration gradients within the electrode itself, leading to a decrease in the mass transfer overpotential. The wide range of substrates that can be used for CNT synthesis (metallic, carbon, etc) via CVD allows the fabrication of electrodes with a high degree of flexibility; materials that are highly desirable in thermocells [152].

3.3.2. Cell design breakthroughs attained using CNTs

The development of flexible electrodes for electrochemical applications has paved the way for innovative cell designs for thermal energy conversion. CNT forests grown directly on thermocell casings, scroll electrodes, and thermocells that can be wrapped around cooling/heating pipes have been attained through flexible CNT electrodes [114].

Mark II thermocell

When inter electrode separation is decreased, larger specific power is generated as mass transfer is enhanced over shorter distances. However, this results in decreased power conversion efficiency as larger thermal energy is required to maintain a similar thermal gradient [112]. Scroll electrodes (Figure 21) can be employed to mitigate this problem. Using scrolled MWNT buck papers, each with a diameter of 0.3 cm and mass of 0.5 mg, aligned along their rolling axis inside a glass tube containing 0.4 M ferri/ferrocyanide, an electrode separation of 5 cm, T_cold of 5 ^oC and thermal gradient of 60 ^oC a specific power density of 1.8 W/m² was obtained. The power conversions efficiency of 0.24 % is an order of magnitude higher than thermocells using Pt electrodes tested under similar conditions [114]. The relative efficiency of the Mark II thermocell is 17 % higher than that obtained when using platinum electrodes, giving Φ_r of 1.4 %.

Coin cell

Thin coin type thermocells which could be powered by extremely low thermal gradients were developed using MWNTs and 0.4M ferri/ferrocyanide as the electrolyte (Figure 22). Coin cells fabricated using MWNT bucky paper electrodes and exposed to a thermal gradient of 45^oC generated a specific power of 0.389 W/m² (equivalent to a normalised power density P_max/ΔT² of 1.92 x 10^-4 W/m²K). Coin cells with electrodes made of MWNT forests around 100 μm tall, grown directly on the internal stainless steel surface of the packaging substrate using a trilayer catalyst (30 nm Ti, 10 nm Al, 2 nm Fe) plasma enhanced CVD method was also developed. The specific power generated at a thermal gradient of 60^oC was 0.980 W/m²,giving P_max/ΔT² =2.72 x 10^-4 W/m²K. The larger normalised power density of the coin cell with MWNT forest electrodes is due to its nanotube alignment, which promotes electrolyte diffusion, and the lower thermal (0.01 cm²K/W) and electrical resistance at the MWNT forest/substrate junction [153]. The thermal resistance for bucky papers is around 0.05 cm² K/W which leads to larger loss of thermal energy at the electrode/substrate junction and 30% less power conversion efficiency [154].

Flexible thermocell

One of the main applications of thermocells is to harvest thermal energy from automobile exhaust pipes and cooling or heating lines in industrial facilities. Flexible thermocells can be wrapped around these pipes and convert them to sources of electrical power. A flexible thermocells consisting of two MWNTbucky paper electrodes kept apart by 2 layers of NomexHT 4848 impregnated with 0.4M ferri/ferrocyanide and wrapped in a stainless steel sheetis shown in Figure 23. The cell was wrapped around a cooling pipe and a thermal gradient of 15^oC was applied using a resistive heater. A specific power 0.39 W/m² was generated proving that flexible thermocells are now a possibility [114].