5. Energy storage

Two elements that can be electrochemically stored in CNTs are hydrogen and lithium. Hydrogen can also be stored in CNTs by gas phase intercalation. Three units are commonly used to describe the hydrogen and lithium contents of storage materials with:

1. [H/C] ([Li/C]) as the ratio of hydrogen (lithium) atoms per atom of storage material, in this

case carbon;

2. [wt%] as the ratio of the mass of hydrogen (lithium) to the mass of storage material (the gravimetric density);

3. [kgH2m-3] as the ratio of the mass of molecular hydrogen to the volume of storage material (volumetric density). 75

5.1 Electrochemical storage of hydrogen

5.1.1 Experimental studies

There are two methods to store hydrogen atoms reversibly in CNTs. One method is called gas phase intercalation and it is explained in section 5.3. The second method described in this section is based on a electrochemical charge-discharge process, in which the hydrogen absorption is controlled by the potential.76

The hydrogen storage capacity of the CNT samples is analysed by means of electrochemical galvanostatic measurement in a 6 M KOH electrolyte. There are commonly three electrodes in the set-up: a work electrode (negative), often made of a mixture of gold or nickel with the nanotube material pressed into a pellet, a reference electrode (Hg/HgO/OH-) and a counter electrode, usually made of nickel. In Figure 5-1, the reference electrode is left out. Instead, a polymer separator separates the working and the counter electrode.

Figure 5-1: A schematic diagram of a charge-discharge cycling apparatus.76

During the charging process, the water in the electrolyte dissociates at the work. The adsorbed atomic hydrogen may be absorbed or intercalate in the electrode or recombine at the surface to molecular hydrogen and diffuse into the electrode or form gas bubbles at the surface of the electrode. During the discharge process, the hydrogen in the electrode recombines with the OH- ions in the electrolyte to form water molecules. The amount of hydrogen desorbed from the electrode can therefore be measured by measuring the electric charge, which is equal to the product of current and time in a galvanostatic setup.77

The following equation gives the reaction in the working electrode:

The reaction at the counter electrode, in the case of a Ni-electrode is given by:

When measuring the electrochemical hydrogen storage capacity of a CNT sample, a double-layer capacitance is measured too. This effect has to be taken into account when looking at the experimental results of the capacity measurements.

Measurement methods are charge-discharge measurements, cyclic voltammetry, constant current measurements, electrochemical impedance measurements and chrono amperometry. The charge/discharge experiments are commonly performed at a constant current. During the discharge, the cell resistance leads to an overpotential. With cyclic voltammetry, different voltages are applied at a certain sweep rate and the current is measured. From the thus obtained oxidation and reduction peaks it can be speculated if the hydrogen adsorption is due to the CNTs or not. With chrono amperometry the current behaviour is studied after applying a voltage step. From impedance spectra, performed by applying sinusoidal voltage over a certain frequency range, a "complex" resistance of the CNT-sample can be calculated. These spectra commonly consist of two semicircles and a slope, similar to a metal hydride.78;79 The smaller semicircle in the high frequency region is probably due to the double-layer capacitance. The semicircle in the low frequency region is attributed to an electrochemical reaction. Gao et al. have studied these spectra of CNTs decorated with metallic nickel as a function of cycling and found that the smaller semicircle hardly changed upon cycling, while the larger semicircle remarkably increased, indicating an increase in surface reaction resistance upon cyling.78

Several experiments have been performed in this research area with often differing conclusions. An overview of most of these results up to 2002 is given in Table 5-1. Note that 1 Ah/g corresponds to 3.54 wt% hydrogen stored in carbon. (References to the articles of the cited results are all in the reference list at the end of the report.)

Table 5-1: Overview of reported storage capacities of hydrogen in carbon nanotubes up to 2002, investigated in an electrochemical system.80

Nutzenadel et al. demonstrated in 1999 that SWNTs could electrochemically store hydrogen, in which SWNT soot containing a few percent of 0.7 - 1.2 nm diameter SWNTs was mixed with gold as a compacting powder (to stabilise the electrodes) in a 1:4 ratio to form electrodes. Gold was used because it is noble and does not participate in any electrochemical reaction. In the nanotubes-sample, nickel, iron and C60 were present too. Pure samples of both nickel and iron did not hydride during cycling and C60 only showed a low capacity of 14 mAh/g. For the nanotubes-sample they found a maximum capacity of 110 mAh/g, indicating that the nanotubes are responsible for the hydrogen uptake. This result corresponds to ~0.39 wt% hydrogen. Only slight capacity loss was observed after many cycles.81

Daiet al. investigated the electrochemical hydrogen storage behaviour of ropes of aligned SWNTs (up to 100 mm in length and 50 mm in diameter) with a large sample quantity of 200 mg (with a purity in the range 60-70%) at 298 K under normal atmosphere. The work electrode was made without compacting powder by pressing the sample to a sheet of nickel foam. With a constant current density of 25 mA/g, the plateau of the discharge potential was observed around -0.6 V (vs Hg/HgO) and a discharge capacity of 503 mAh/g was obtained, corresponding to a hydrogen storage capacity of 1.84 wt% in SWNTs. They observed the same maximum discharge capacity for three different samples, so the results appeared reproducible. After 100 cycles the electrode still retained 80% of the maximal capacity. The loss is believed to be due to the mechanical instability of the electrode. The investigators explain the higher capacity (in comparison to the 110 mAh/g obtained by Nutzenadel et al.) as due to the purity and larger mean diameter of the SWNTs (1.72 nm, resulting in larger micropores). Furthermore, no compacting powder was used, resulting in a higher percentage of SWNTs in unit volume in the work electrode.82

Fazle Kibria et al. compared the electrochemical hydrogen storage behaviours of CNTs grown by chemical vapour deposition (CVD), arc-discharge (AD) and laser ablation (LA). The latter were purified (90 wt%). The charge capacities of the electrodes remained unchanged with increasing cycle number, but the reversible capacity behaviour decreased with cycling. The samples with the LA grown CNTs stored the highest amount of hydrogen, 1.6 wt% (corresponding to 400 mAh/g), which was 16 times that of CVD and AD grown CNTs. During purification, besides the elimination of metals and amorphous carbon, CNTs may separate partially from their bundles, resulting in an increase in active sites for hydrogen storage for the LA grown CNTs. They also investigated the charge-discharge behaviours for Li-, Na- and K-doped sample electrodes. Each sample obtained a higher hydrogen capacity. The Li-doped CVD electrode showed a unique discharge behaviour during cycling. With increasing cycle number, the sample showed a large capacity reversibility. The investigators concluded from cyclic voltammetric measurements that the high hydrogen storage capacity of the alkali metal-doped CNTs originated from an increase in hydrogen adsorption sites. This is suggested to be due to the introducing of the metals in the CNT bundles and separation of tubes, but not from their chemical effect.83

More recently, Zuttel et al. investigated several nanotubes samples at room temperature, fabricated of a mixture of 10 mg of SWNT material with 90 mg of gold powder to form a mechanically stable and conducting electrode. Figure 5-2 shows the charge-discharge equilibrium curves measured with the SWNT/gold electrode. The shape of the curves is different than the shape of an equilibrium curve for a metal hydride, in which there is a wide plateau due to a phase transition occurring at a specific potential. In nanotubes, there are no well-defined interstitial sites and no phase transition is expected.81

Figure 5-2: Equilibrium charge-discharge curves of an electrode of SWNTs with gold as compacting powder (full circle: charge; open circle: discharge). The shape of the curve is different form that for metal hydrides where a plateau occurs.81

A few SWNT samples showed hydrogen storage capacities in the range of 0.7-0.9 wt%. The fact that this reaction was reversible is a good indication that the measured capacity is due to hydrogen storage. An additional discharge from the oxidation of impurities in the nanotubes would not necessarily be reversible. The electrochemically measured discharge capacity at 293 K of the nanotube samples correlated with the surface area. The linear relationship is shown in Figure 5-3. Note that only the round markers in the graph represent the surface area of CNT-samples, the fitted line corresponds to this data (the data from gas phase experiments, indicated by triangular markers, were found by Nijkamp et al. and are not included in this fit). The line intercepts the axis at the origin and the slope is 1.5 wt%/1000 m2/g. Theoretically, the maximum specific surface area is 2630 m2g-1. From this, they extrapolated a maximum discharge capacity of 2 wt%.77

Figure 5-3: Desorbed amount of hydrogen versus the surface area (round markers, corresponding to the electrochemical experiments) of several CNT samples with a linear fit (corresponding to the data of the round markers). The triangular markers correspond to gas phase experiments, carried out by Nijkamp et al.77

Less encouraging results are from Frackowiak et al. who investigated the accumulation of hydrogen by the electrochemical decomposition of aqueous alkaline medium on a negatively polarised carbon electrode under ambient conditions. For several SWNTs the reversible hydrogen storage capacity never exceeded 0.5 wt%. They compared several published data and found no systematic relationship between the indicated purity and the maximum discharge capacity, suggesting that SWNTs are probably not responsible for the values of sorption observed. Their final conclusion is that hydrogen storage capacity of SWNTs under a high pressure at room temperature is negligible.84

5.1.2 Modelling

In this section, a review of the modelling studies on the chemisorption of hydrogen is given, but it must be mentioned that all these studies involve modelling of storage in the gas phase.

Simultaneously to experimental studies of hydrogen storage in CNTs, model calculations of chemisorption and physisorption by the nanotubes were performed. These are very useful for understanding the elementary steps of the adsorption procedure and give insight in the phenomenon. From 2001, the quantum picture was introduced into the molecular dynamics study of hydrogen either by quantum molecular dynamics algorithms or by minimal ab initio calculations in parts of classically optimised tube geometries. The main reason that ab initio calculations on CNTs have appeared only lately is the system size. In order to take into account a large enough model of a SWNT, you need approximately 200 atoms. The problem is then how to treat a large system using an accurate ab initio method without ending up with a prohibitively large calculation.

There are two different ways to deal with this dilemma. One can use the periodicity of the CNTs and combine an ab initio method with periodic boundary conditions. The advantage is that the total system is treated with ab initio techniques, while the disadvantage is that an external periodicity is forced on the system. The second solution involves a two-level quantum mechanics/molecular mechanics (QM/MM) approach, in which the tube is divided into two different parts treated with different methods. For this model there is no periodic constraint, but its disadvantage is that only a relatively small part of the system is treated quantum mechanically, while the rest is used for constraining the boundaries. Both approaches have although the disadvantage of excluding temperature from the calculations.85

Cracknell has simulated the hydrogen adsorption in MWNTs and SWNTs. The gas-solid interaction was modelled as pure dispersion forces and with a hypothetical model for chemisorption. The method used for modelling dispersion forces is based on integrating the Lennard-Jones interaction between the adsorbent and adsorbate over the surface of the pore walls and ignoring the detailed surface structure. The hydrogen molecule is modelled as a dumbbell molecule with two Lennard-Jones sites. The standard equation for the Lennard-Jones interaction between the potential is86:

where r is the distance between the centres of mass of two hydrogen atoms. The cross parameters for the Lennard-Jones interaction between the hydrogen site and the graphitic surface were calculated using the Lorentz-Berthelot rules86:

For modelling the hypothetical chemisorption within the nanotubes, the chemisorption minimum from a model by Jeloaica et al.86 (called chemisorption2, see Figure 5-4) was combined with the dispersion minimum (10-4-3) calculated from the Lennard-Jones interactions between a hydrogen site and the wall.

Figure 5-4: Comparison of 10-4-3 potential for interaction of a single hydrogen site with a graphitic surface and two hypothetical models for chemisorption (z = distance from wall).87

For each possible nanotube wall-hydrogen distance z, this combined potential was calculated by comparing the energy of the chemisorption potential in a planar system and the potential calculated form the integration of the Lennard-Jones potential over the surface of the pore wall. The minimum of the two values was then taken. Uptake of hydrogen in the internal space of a CNT is predicted to be lower than in the optimal graphitic nanofibre with slitlike pores. An uptake of 3.2 wt% is measured in pores of diameter 1.2 nm (see Figure 5-5).

Figure 5-5: Simulated adsorption isotherms for hydrogen in SWNTs of differing diameters at 298 K. The SWNT adsorption refers only to the internal pore space and ignores any interstitial adsorption.87

The majority of the differences can be attributed to the curvature of the pore. This reduces the uptake of hydrogen in spite of deepening the potential minimum inside the pore associated with dispersion forces. It is concluded that for uptake of hydrogen of 5-10 wt% gas-solid forces other than dispersion forces are required and most of the adsorption must occur in the interstices between the SWNTs.87

A mixed quantum mechanics/molecular mechanics (QM/MM) model is used by Froudakis for investigating the nature of hydrogen adsorption in a 200-atom (4,4) SWNT, treating up to 64 carbon atoms and 32 hydrogen atoms with the higher level of theory. The tube was divided into three cylindrical parts (see Figure 5-6).

Figure 5-6: A QM/MM model simulating a (4,4) SWNT. The tube is separated into three cylindrical parts. The inner part is treated with DFT (40 carbon atoms) while the two outer parts are treated with MM. The dangling bonds at the ends of the tube were saturated with hydrogen atoms.88

The inner part was treated with DFT, while the two outer parts were treated with molecular mechanics. The three-parameter hybrid functional of Becke was employed for the higher theoretical level. The two outer cylindrical parts were treated with the universal force field (UFF), while the dangling bonds at the ends of the tube were saturated with hydrogen atoms. It was investigated what potential the hydrogen atom feels as it approaches the CNT. A combined interpretation of the potential curves during the hydrogen approach to a CNT in two different pathways (towards a carbon atom and towards the centre of a C-hexagon) shows that despite the approaching direction, the H will feel an attraction from the C atoms of the tube and finally will bind to the wall of the tube and will not enter inside. Two forces compete in the approaching procedure. The more hydrogen atoms are in the C-hexagon, the larger the hexagon is and the easier a hydrogen atom enters the tube. On the other hand, the hydrogen atoms in the hexagon are screening the attraction of the carbon atom to the external hydrogen. As a result, the energetically favourable H-approach is when the tube wall is half-filled with hydrogen atoms. Furthermore, it was shown that the binding will take place in zigzag rings around the tube walls and not in lines toward the tube axis, changing the tube shape and causing an enlargement of the tube volume by 15 %.88

By Yang et al.ab initio molecular orbital (MO) calculations are performed to study the adsorption of H atoms on three faces of graphite: basal plane, zigzag edge and armchair edge. It is thought that the adsorption on nanotubes is comparable to that on graphite. Clear indications of chemisorption are the slow uptake of hydrogen and the necessity to heat up the sample to desorb the adsorbed hydrogen completely. The edge sites of graphite may play a significant role in hydrogen storage, as they are abundant in GNFs as well as in nanotubes after acid and heat treatments. Hence it is important to study edge sites where hydrogen dissociation can take place.

For the calculations the unrestricted Hartree-Fock method was used for geometric optimisation and the self-consistent field energy. B3LYP/6-31G (method and used basis set) was used for the single point energy and bond population calculations for the more accurate results. The geometry-optimised structures were used to calculate the energy of chemisorption (Eads): Eads = Egraphite-hydrogen - Egraphite - Ehydrogen. A higher negative value of Eads in kcal/mole corresponds to a stronger adsorption. The different used models are given in Figure 5-7 and Figure 5-8.






Figure 5-7(A-F): Model A is the armchair graphite model. Model B is the zigzag graphite model. Models C and D represent the structure of one H atom chemisorbed per edge carbon on these graphite planes. The models E and F represent H atoms adsorbed on the basal plane sites.89


Figure 5-8 (G-H): Model G represents the structure of two H atoms chemisorbed per edge carbon of the armchair face of graphite. Model H represents the same, but for the zigzag face of graphite.89

Model C represents the structure of one hydrogen atom chemisorbed per edge carbon on the armchair plane of graphite. Model D represents the same, but for the zigzag edge of graphite. For chemisorption on the edge sites, no significant changes in bond length or bond population are seen. Only the C(15)-C(16) bond is weakened by H adsorption.

The models E and F represent hydrogen atoms adsorbed on the basal plane sites. In model E, two hydrogen atoms are bonded to two adjacent carbons as in model F they are further apart. There are, in both cases, significant increases in C-C and C-H bond lengths where hydrogen is directly involved. Also, the bond population decreases, which indicates weaker C-C and C-H bonds. The ab initio results from this work show that the adsorption of an H atom on the basal plane is exothermic and stable.

Another possibility is the chemisorption of two hydrogen atoms per edge carbon. For the armchair face of graphite this is shown in model G en for the zigzag face this is represented by model H. In this case, the C-C bonds are weakened where H atoms are directly involved. This bond weakening is significantly higher with two hydrogen atoms bonded on each carbon. The same holds for the C-H bond.

In this study with ab initio MO calculations, it is found that both edge planes and basal planes in graphite can chemisorb hydrogen. The strength of this chemisorption is higher on the edge planes than on the basal planes. The order of this chemisorption strength is as follows:

zigzag edge > armchair edge > basal plane

For adsorption on both edge planes, the bond energy decreases when two H atoms are adsorbed on each carbon site as compared with one H per site. For adsorption on the basal plane sites, the energy of adsorption is substantially lower when the H atoms are occupying adjacent sites.

Bauschlicher studied the bonding of H atoms on the exterior wall of SWNTs. He calculated the bond energies for the tube with 1 H, 2 H, 24 %, 50 % and 100 % coverages. The average C-H bond for the first H was 21.6 kcal/mole and 40.6 kcal/mole for the first two H atoms. For 50 % coverage this average bond energy was 57.3 kcal/mole, decreasing to 38.6 kcal/mole for 100 % coverage.90

The C-H bond energies on the tube are in general agreement with that on the basal plane of graphite. The experimental results are thus in agreement with the results from Yang et al. So, it is unlikely that the adsorption of H atoms would be significantly different on the basal plane of graphite than on the exterior wall of the nanotube.89


5.2 Electrochemical storage of lithium

5.2.1 Experimental studies

Lithium intercalated graphite and other carbonaceous materials are commercially used in Li-ion batteries. In this case, the specific energy capacity is partially limited by the thermodynamically determined equilibrium saturation composition of LiC6. During intercalation (discharging) the following adsorption takes place:

Or written in a different way:

During de-intercalation (charging) the process is reversed. In graphite this adsorption could be visualised as in Figure 5-9.91

Carbon nanotubes are interesting intercalation hosts because of their structure and chemical bonding. Nanotubes might have a higher saturation composition than graphite as guest species can intercalate in the interstitial sites and between the nanotubes. Therefore, carbon nanotubes are expected to be suitable high energy density anode materials for rechargeable Li-ion batteries.92

Figure 5-9: Structure of graphite intercalated with lithium. 91

SWNTs spontaneously form bundles that are called nanoropes. These bundles are kept together by Van der Waals forces. Reversible, electrochemical intercalation of SWNT bundles with lithium has been demonstrated in the past few years. If SWNTs have been purified, they show a reversible saturation composition of Li1.7C6 (632 mAh/g). In any case, this is higher than LiC6, which is the ideal value for graphite this corresponds to a capacity of 372 mAh/g. This LiC6 value also holds for MWNTs. However, ball-milling can further increase the reversible saturation Li composition of SWNTs. This process induces disorder within the SWNT bundles and fractures the individual nanotubes. After ball-milling the saturation composition can be as high as Li2.7C6 which corresponds to a capacity of 1000 mAh/g. 70;93;94

B. Gao et al.70 use SWNT bundles synthesised by laser ablation. The crude materials were purified by filtering off the impurities over a micro-pore membrane while keeping the nanotubes in suspension. The purified material existed for 80% of SWNT bundles with a bundle diameter varying between 10 and 40 nm. The individual nanotube diameter in the bundles was between 1.3 and 1.6 nm. The purified SWNTs from the same batch were divided into several groups. Every group was treated differently by impact ball-milling in air for up to 20 minutes. The charge-discharge data of each group are now being observed.

A composition of Li5.4C6 was obtained after the first discharge with the untreated purified SWNTs. The reversible capacity, Crev, is Li1.6C6. A large voltage hysteresis between discharge and charge cycle was observed in all samples measured. This hysteresis is related to the kinetics of the occurring electrochemical reaction. Preliminary experiments have shown that cutting nanotubes to smaller segments can reduce hysteresis.

The ball-milled samples showed a better performance. Crev increased with increasing processing time showing a maximum after 10 minutes of ball-milling. In this case, Crev equals Li2.7C6. Further processing of the nanotubes led to a lower Crev. The for 10 minutes ball-milled samples showed very little capacity decay after 5 discharge-charge cycles at a current of 500 mA/g.

It was suggested that SWNTs were fractured and shortened by ball-milling. With Raman spectroscopy, different samples (before and after ball-milling) were analysed (Figure 5-10). The peak around 1600 cm-1 is associated with some of the SWNT vibrational modes. The peak at 1340 cm-1 is associated with disordered carbon. One can see that the major phase in the sample ball-milled for 10 minutes still consists of SWNTs. This is in contrast with the 20 minutes ball-milled sample. It seems that ball-milling increases the Crev of SWNTs, but ball-milling too long gives the opposite effect. Ball-milling the SWNTs for 10 minutes gives the best results for Li intercalation.

However, the exact mechanism for the enhanced Li intercalation capacity in ball-milled SWNTs is not clear. It is suggested that the enhanced capacity is related to the degree of disorder within the bundle, and lithium diffusion into the inner cores of the fractured nanotubes.

Figure 5-10: Raman spectra of purified [a], 10 minute [b] and 20 minute [c] ball-milled SWNT materials. 70

H. Shimoda et al.95 also used SWNT bundles synthesised by laser ablation. Their purified samples contain over 90% SWNT bundles with a length of 10 mm and a bundle diameter of 30-50 nm, while the average diameter of an individual nanotube was estimated at 1.4 nm. The purified bundles were processed to shorter ones by sonication in a solution of H2SO4 and HNO3 for 10-24 hours. The average lengths of the bundles were reduced to 4 mm after 10 hours and to 0.5 mm after 24 hours of processing.

After chemical etching, the samples (the purified sample, A, also) were electrochemically reacted with lithium. Then, the reversible capacity of the samples was analysed. This capacity turned out to be Li2C6 (744 mAh/g) for the sample with a bundle length of 4 mm (sample B) and Li2.1C6 (781 mAh/g) for the sample with a bundle length of 0.5 mm (sample C). This is an increase of a factor 2 with regard to sample C (Figure 5-11). This effect is attributed to Li filling of the interior space of the chemically etched SWNTs. However, the difference in capacity between samples B and C is not very big despite their difference in length. So, it seems that etching for more than 10 hours has no additional value.

The voltage hysteresis was 0.5 V lower in sample C compared with sample A. This indicates a shorter diffusion path or diffusion barrier in the etched samples, which is consistent with the observation of smaller bundle diameter and length in sample C. It is clear that etching for more than 10 hours gives a better effect in this case.

It can be concluded that processed SWNTs have a better reversible capacity concerning lithium intercalation with regard to simple purified SWNTs. So, SWNT bundles seem to be attractive host materials for energy storage.

Figure 5-11: Second cycle intercalation (discharge) and deintercalation (charge) data collected from the purified and etched SWNTs. 95

MWNTs could also be low-cost, high-performance anode materials for rechargeable lithium-ion batteries since they show an excellent reversible capacity and cycle ability during lithium insertion and extraction.96

5.2.2 Modelling97;98

As stated earlier CNTs could be used in Li ion batteries. Lithium can intercalate in CNTs but it is not always clear how lithium ions get inside the SWNTs or SWNT bundles and if the inside or outside of the tubes is preferable for lithium. These questions can be answered by theoretical calculations. This is what we call modelling.

Because of the large size of CNTs ab initio theoretical calculations of the full system are impractical. For that reason it is possible to use a good model with fewer atoms, which mimics the system quite well. With nanotubes the insertion of lithium takes place through the hexagonal rings in the tube wall. In a model, this can be simplified by considering only a benzene molecule. Insertion of the Li ions takes place through the centre of this benzene ring. However, not only hexagons (6R) are considered, but also pentagons (5R), heptagons (7R) and octagons (8R), which are common defects within CNTs.

The initial calculations on these simple models have been carried out using HF, MP2 and DFT-B3LYP methods using a wide range of basis sets from minimal to extended. For the reliability of the models, lithium ion intercalation into two different zigzag nanotubes with different length has been studied. The electronic binding energies of Li and the benzene ring at different levels of theory are summarised in Table 5-2. Geometries are optimised at each level of theory. Energy values obtained from the highest level of calculations (6-311++G** basis sets) are in good agreement with the experimental binding energy of 1.70 eV.

Table 5-2: Binding energies and equilibrium distances between Li+ and the centre of the benzene ring.98

From studies on Li-intercalated carbon nanotube ropes it was found that little deformation of the structure is present as a result of intercalation. So, in our model the C-C distances are kept fixed at the normal 1.4 Å for nanotubes. Then the energy associated with moving the lithium ion along an axis perpendicular to the benzene ring is measured. This energy is showed for the different models in Figure 5-18. It can be seen that there is a minimum at 1.8 Å from the centre of the hexagon. This is the equilibrium distance. At the centre of the ring, the energy is at a maximum. The lithium ion would have to cross a high energy barrier to enter the nanotube.

Figure 5-12: Variation of energies by moving Li+ from 2.6 Å to the centre of the benzene ring.98

The same can be done with the other three rings. The intercalation energy is lowered drastically from 24 eV to 2 eV as the ring size increases from pentagon to octagon. This is showed in Figure 5-13. It is now supposed that intercalation of lithium in nanotubes becomes easier when the structural defects are more present, because of the greater ring sizes and thus lower energy barriers. The equilibrium distance of 1.8 Å seems independent of the ring size.

Figure 5-13: Variation of MP2/6-31G* energies by moving Li+ to the centre of the 5-, 6-, 7- and 8-rings.98

Additional calculations have been performed on ring geometries that are allowed to expand. In this case the energy barriers are significantly lower, especially for the smaller rings. So, allowing the rings to deform facilitates the passage.

However, one single ring is only an approximation for the potential experienced by the ion outside of the nanotube. For an approximation inside the tube, two 6-membered rings with a lithium ion in between could be considered. This can be seen in Figure 5-14a. The distance L gives the diameter of the tube or the interlayer distance in a bundle of nanotubes. With L = 4 Å, all three methods indicate a situation in which the ion prefers its equilibrium position in the centre of the tube. At longer L, the lithium ion crosses two minima around 2 Å from each ring. A small barrier is encountered in the middle of the two rings. This barrier height increases with the distance L (Figure 5-14b). The difficulty of moving the lithium ion within the tubes seems to depend strongly on the diameters of the tubes.

Figure 5-14: (a) A hexagon model where L is the distance between 2 benzene rings. (b) Variation of energies (MP2/6-31G*) by moving Li+ between two benzene rings98.

The length of the tube is not expected to be an important factor. Comparing two tubes with different lengths but same diameter can test this assumption. For this test two zigzag nanotubes of different length are taken. For the two different tubes there are nearly no differences in energy with approach and intercalation of the lithium ion. However, the height of the energy barrier inside the tube depends on the length of the tube. The longest tube has the lowest energy barrier inside the tube.

From the model and the actual calculations it is suggested that the lithium ion prefers positions as well inside as outside the tube. It is possible now to look at the several combinations of inside and outside positions for two lithium ions in a nanotube. The electronic binding energy of the two ions strongly depends on the positions of those ions. The least stable state is found when both ions are inside the tube. The DE value then is -1.80 eV and the distance from the wall is 1.934 Å. The stability increases if one ion is moved outside the tube. The most stable state is the situation in which both ions are outside the tube with a DE value of -2.96 eV. This state is even more stable than the single ion system. The different possible states with their binding energies are given in Figure 5-15.

Figure 5-15: HF/3-21G optimised structures and binding energies (in eV) of different isomers of one Li+
(a and b) and two Li+ ions at a nanotube.98

Another situation that can be modelled is the situation of multiple lithium ions in MWNTs. This can be seen in Figure 5-16a. In Figure 5-16b the energy of moving Li1 to the centre of the middle ring and the position of Li2 compared to X2 are shown. What happens is that Li1 has its equilibrium state at 2 Å from X1. If we move Li1 towards X2 the energy increases. Li2 is at that moment in its equilibrium state compared to X2. When Li1 approaches X2 further, Li2 shifts sharply to its second preferred position at about 2 Å from X3. The energy obtained by moving Li1 towards X2 now decreases suddenly but increases to a maximum by moving further towards X2. The barrier height for Li1 is
2.3 eV and this is double that of the two benzene rings and one lithium ion model.

The last case that could be modelled is that of an opened tube. There are three possible paths for insertion: along the central axis of the tube or at 2 Å from the tube wall inside or outside the tube. The barrier height for the path along the central axis is significantly lower than for the other pathways. So, it seems that the lithium ions enter the tube through the central axis and then may spread inside the tube close to 2 Å from the tube wall.

Figure 5-16: (a) Three benzene rings and two Li+ ions model. (b) Variation of energies (HF/6-31G*) and distance between Li2 and X2 by moving Li1 from the centre of the leftmost ring (X1).98

For the application of CNTs in Li-rechargeable batteries it can be concluded that the barrier height of the intercalation process is a crucial factor in battery activity. Insertion of lithium ions through the sidewall of the nanotubes seems energetically unfavourable unless there are structural defects. Release of the ion during discharge process has to cross a very high barrier. This depends on the size of the rings. The electronic binding energies of the lithium ion at its equilibrium distance also decrease as the ring size increases. Thus, it seems that ions outside the tubes may more easily take part in battery activities. For the situation of two lithium ions, the binding energy strongly depends on the position of the lithium ions compared to each other. The most stable configuration is the one where both the ions are outside the tube.


5.3 Gas phase intercalation of hydrogen

5.3.1 Experimental studies

Gas phase intercalation of hydrogen in CNTs concerns the adsorption of H2, called physisorption instead of chemisorption (involving H+ and chemical bonds). This adsorption of H2 (other gases are possible too) on the surface of CNTs is a consequence of the field force at the surface of the solid, called the adsorbent, which attracts the molecules of the gas or vapour, called adsorbate. The forces of attraction emanating from a solid can be either physical (Van der Waals) or chemical (thus chemisorption, involving the electrochemical storage of hydrogen). This section is about the storage due to the physical forces.

Carbon nanotubes have attracted considerable interest due to several reports of high hydrogen storage capacities at room temperatures, even higher than the goals set for vehicular storage by the Department Of Energy (being an H2-storage capacity of 6.5 wt% and 62 kg H2/m3). However, conflicting reports indicate that room temperature storage capacities do not exceed ~ 0 wt%. Reporting capacities range from 0 to 60 wt%. Several reports on hydrogen storage exist, which can be divided according to their results. The most important results found are:

  1. Hydrogen storage with a H/C ratio larger than 2, thought to be incorrect (> 14 wt%). For example, a storage capacity of 67 wt% in GNFs (see Table 5-1) is reported, but this seems highly irrational because this corresponds to an average of about 24 hydrogen atoms for each carbon atom;
  2. Hydrogen storage consistent with expectations based on findings for activated and other, conventional high-surface area carbons (0 -2 wt%);
  3. Hydrogen storage results are in the intermediate range and are not obviously incorrect (2-14 wt%).99

There are several possibilities for the hydrogen to be stored in the CNT samples. For hydrogen storage in closed CNTs the structure has two possible sites: inside the tubes and in the interstitial sites between the tube array (Figure 5-17). In case of a long SWNT closed with fullerene-like end caps, the hydrogen can only get access to the tube interior via the hexagons of the graphite-like tube wall. An opened tube with removed caps gives an easier access for the hydrogen molecules to the tube. Typically, the tubes are very long and therefore a good diffusivity of the hydrogen inside the tube will be required in order to fill the whole tube volume. Cutting the rope in shorter pieces may therefore help to improve hydrogen storage and its kinetics.100

Figure 5-17: Schematic representation showing potential sites for hydrogen adsorption within a nanotubes bundle: (left) hydrogen atoms occupying the interstitial spaces between the tubes, and (right) hydrogen atoms inside the tube interior.101

Mainly, three different techniques are applied to study the hydrogen storage in solids: (1) the volumetric method, (2) the gravimetric method and (3) thermal desorption spectroscopy.

(1) The volumetric method measures the pressure drop owing to hydrogen adsorption after applying a hydrogen pressure to the specimen, contained in a constant volume. Similarly the pressure increase due to desorption can be measured. For good accuracy and reliable results, this method requires typically specimen masses of 500 mg or higher. Leakage or temperature instability of the apparatus may give rise to large experimental errors. The advantage of this technique is that both adsorption and desorption can be measured.

(2) The gravimetric method measures the weight change of the specimen due to absorption or desorption of hydrogen. A high accuracy can be achieved even at sample masses of 10 mg. However, this technique is sensitive to all gasses adsorbed or desorbed since it is only based on weighing.

(3) Thermal desorption spectroscopy (TDS) measures only the hydrogen desorption in high vacuum utilising mass spectrometry. This method is selective and highly sensitive, which can be improved by using deuterium-loaded specimens. In this case no disturbing background from water or other hydrogen containing adsorbents occurs.100

Table 5-3 shows some of the reported storage capacities, up to 2002:

Table 5-3: Overview of reported storage capacities of hydrogen in carbon nanotubes, up to 2002, of which most at room temperature.99

None of these experimental results have been confirmed by independent research groups. Obtaining activated CNT hydrogen storage materials with highly reproducible adsorption capacities has not yet been achieved. A possible reason is that hydrogen storage is only optimised for a very specific and narrow distribution of CNTs of distinct types and diameters. Although theoretical calculations predict that a range of 4-14 wt% hydrogen adsorption in carbon nanotubes is possible, they do not clearly distinguish between chemisorption and physisorption. Chemisorption (the covalent bonding of hydrogen) would require a high temperature and high activation energy for hydrogen release, whereas any practical fuel cell application would require low adsorption and desorption energies.101

Work by Liu et al. indicates that SWNTs are highly promising for H2 adsorption, even at room temperature. They measured the H2 storage capacity of SWNTs synthesised by a hydrogen arc-discharge method, with a relatively large sample quantity (about 500 mg) at ambient temperature under a modestly high pressure, which was soaked in hydrochloric acid and then heat-treated in vacuum. A H2 uptake of 4.2 wt%, which corresponds to a H/C atom ratio of 0.52, was obtained by these SWNTs with an estimated purity of 50 to 60 wt%. Also, 78.3 % of the adsorbed hydrogen (3.3 wt%) could be released under ambient pressure at room temperature, while the release of the residual stored hydrogen (0.9 wt%) required some heating of the sample. Furthermore, they found that after four cycles of adsorption/desorption, the H2 uptake capacity of the SWNTs remained unchanged and that less than 1 wt% could be stored at pressures lower than 5 MPa. The researchers argued that this relatively high H2 adsorption capacity of their SWNTs could be related to their larger mean diameter of 1.85 ± 0.05 nm, while typical SWNT diameters are in the range of 1.2 to 1.4 nm.102

Careful work at NREL (National Renewable Energy Laboratory) by Dillon et al. indicates a maximum capacity for adsorption of hydrogen on SWNTs is ~8 wt%. Samples displaying this maximum value were prepared by sonicating purified SWNTs in a dilute nitric acid solution with a high-energy probe. In this process SWNTs are cut into shorter segments and this introduces a Ti6AI4V alloy due to the disintegration of the ultrasonic probe. The hydrogen adsorption on many of the NREL samples can be explained fully by adsorption on the incorporated alloy. However, one-half of the samples show capacities that are too high to be explained by the presence of the alloy alone. Assuming that the alloy in the SWNT sample behaved like the pure probe-generated alloy sample of Ti-6AI-4V, the hydrogen uptake on the SWNT fraction still reached ~ 8 wt%. Cross-calibration of the experimental apparatuses with three different standards establish the validity of the measurements, and repeated measurements on a given sample yield reproducible results, but nominally similar sample preparation procedures do not repeatedly produce samples that exhibit the same hydrogen storage capacities. A possible explanation is that the degree of tube cutting, and how, where and in what form the metal particles are incorporated can vary dramatically even if identical sonication parameters are employed. Earlier, in 1997, they found that the desorption of hydrogen fitted first order kinetics and the activation energy for desorption was measured to be 19.6 kJ/mol. Also, adsorption measurements on samples having differing distributions indicate a link between SWNT size and type and hydrogen capacity. More recently, they developed a controlled dry cutting technique that does not employ ultrasonication (and so does not simultaneously incorporate a metal hydride alloy). Transmission electron microscopy analyses and Raman spectroscopy showed that significant cutting occurred without extensive damage to the SWNTs.99

5.3.2 Modelling

Gas phase intercalation of hydrogen in CNTs is mainly being analysed by grand canonical ensemble Monte Carlo simulations (GCEMC) and molecular mechanics (MM) classical algorithms, in which only nonbonding interactions, mostly being described by the Lennard-Jones potential, are included in the potential functions.103 In a simulation loop, the motion of hydrogen molecules in a given pore volume for a fixed temperature T and a chemical potential m is calculated. The carbon pore is built up by M carbon atoms being located at the surface of any desired pore geometry. The most commonly used potentials U(r) for the hydrogen-hydrogen and hydrogen-carbon interaction energy are described below. The total potential energy of a particular hydrogen molecule is then given by summing up all interaction energies between neighbouring hydrogen molecules and pore wall carbon atoms.

During the simulation, the particle number N within the pore fluctuates owing to particle displacement, creation and destruction steps, which are executed with equal frequency. In a creation step, the position of the new particle is chosen randomly within pore volume V and its potential energy U is calculated. This creation step is accepted with the probability104:

Analogous destruction, respectively displacement steps are accepted with the probability:

A simple and ideal model of nanotube materials consists of parallel SWNTs with equal diameters D on a two-dimensional triangular lattice where the minimal distance between the nanotube walls is equal to d. In the cylindrical wall obtained by rolling up a basal graphite plane, the carbon atoms form a two-dimensional hexagonal lattice where each carbon atom has three nearest neighbours at a distance c = 1.42 Å. The diameter D of a nanotube is equal to

D =

where n is the number of hexagons along the tube perimeter. The pair wise interaction energy between two hydrogen molecules is mostly calculated by the Lennard-Jones potential . In addition to this pair potential, the hydrogen molecules interact by a quadrupole-quadrupole interaction which is taken into account by the Coulomb interactions of effective electric charges: two charge q = 0.4829e located on the protons and one charge -2q at the centre of mass. The interaction between hydrogen molecules and carbon atoms in the walls commonly uses the Lennard-Jones potential too.

Due to the small mass of hydrogen and the confinement of the molecules inside the nanotubes and the interstitial space between the tubes, the quantum effects are expected to contribute significantly to the adsorption process. Some investigators take these effects into account in simulations by the path-integral Monte Carlo method. But in the domain of temperature above 70 K considered in most studies, the use of a Feynman-Hibbs effective potential which estimates the quantum effects exactly to the order of is sufficient (the dimensionless parameter giving the magnitude of the quantum effects, with mr the reduced mass of a pair of hydrogen molecules and kB the Boltzmann constant, is equal to ~0.07 at T = 70 K). The expression for the Feynmann-Hibbs effective potential is105:

with ULJ(r) given by .

The potentials described above are not the only potentials that are used to model the H2-H2 and H2-C interaction energy. An alternative potential for the H2-H2 interaction energy is the Silvera-Goldman potential, which is a semi-empirical pair potential UP(r) that includes a pair-wise effective three-body term. This potential is given by 106:

with:

and rm being the position of the minimum of UP(r).

An alternative potential for the hydrogen-nanotube interaction is a model by the Crowell-Brown potential for hydrogen interacting with carbon atoms in an oriented graphitic sheet. This hydrogen-carbon potential is given by 107:

where r is the distance from a carbon atom in the tube, to the hydrogen molecule (which could be either inside or outside the tube), and f is the angle between the axis normal to the tube surface and a line connecting the hydrogen and carbon atom. EH and EC are the atomic energies for hydrogen and carbon respectively, P// and are atomic polarisabilities, where the subscripts denote orientations parallel to and perpendicular to the graphite c axis and PH is the polarisability of hydrogen.107 A thorough description of these potentials can be found in the referred articles.

Apart from the hydrogen-hydrogen and hydrogen-carbon interaction energy, several other features can be modelled in different ways, leading to differences of adsorption amounts in published articles. These differences are mainly attributed to different models/values for:

- single- or multi-walled carbon nanotubes;

- open or closed tubes, which are important for the adsorption phenomenon (open tubes enable gas adsorption both inside and outside the tubes whereas closed tubes do not);

- the reactive surface of the adsorbent material considered. This value is dependent on the distance considered between consecutive tubes, the so-called inter-tube spacing;

- the sampling method used in simulations;

- the gas thermodynamic conditions (in particular, the adsorption is temperature-dependent due to the high attractive forces at 80 K and high thermal effects occurring at room temperature at which the hydrogen molecules dispose an important kinetic energy);

- the adsorbent configurations (including the values of the tube diameters, the tube lengths and the inter-tube spacing) that permit the hydrogen gas either to move in a large adsorption volume or just to be confined in the interstitial pores;108

In spite of the use of different calculation methods, the following conclusions could be reached:

- adsorption of hydrogen does possibly occur in sites of different binding energy in SWNTs. It is suggested that the outer surface of SWNT ropes in the simulation of physisorption is important, not only for geometrical reasons (high surface area), but also for energetic reasons;

- the inner-tube cavity has high adsorption potential for hydrogen, compared with the planar surface and slit-pores of similar size;

- interstitial adsorption constitutes a significant fraction of the total amount adsorbed hydrogen for a tube of larger diameter such as the (18, 18) tube array (2.44 nm). By comparison, in the smaller interstices of the (9,9)tube array, adsorption is negligible due to the quantum effect. Interstitial adsorption accounts for at most 14 % of the total adsorption for the (18, 18) tube array at 77 K;107

- the packing geometry of the SWNTs plays an important role in hydrogen adsorption;

Most of the performed simulations do not confirm the high hydrogen uptake capacity obtained experimentally for similar systems of SWNTs. However, Williams et al. reported the results of Monte Carlo simulations for the physisorption of H2 in finite-diameter ropes of parallel SWNTs is consistent with the experimental results obtained by Ye et al. at 77 K. Their results show that the maximum gravimetric adsorption capacity of hydrogen onto an isolated (10, 10) nanotube can reach 9.6 wt% under cryogenic temperatures (77K) and a pressure of 10 MPa. However, their results cannot explain the hydrogen adsorption at room temperature.109;110

Levesque et al. have studied hydrogen storage in CNTs by Monte Carlo simulations in the pressure domain from 0.1 to 20 MPa at temperatures of 77, 150 and 293 K. The hydrogen-hydrogen and hydrogen-carbon interaction energies were both described by with =36.7 K, = 2.958 Å and = 32.05 K, = 3.18 Å respectively. The quantum effects are modelled by .

Figure 5-18: Projection on the xz-plane of adsorbed hydrogen molecules at 10 MPa and 293K in a triangular lattice of closed and opened nanotubes respectively with diameter D = 13.3 Å with d = 3.4 Å (left) and D = d = 6 Å (right). The open circles give the locations of nanotube walls, the black dots those of hydrogen molecules. The lengths of the sides of the simulation cell are in angstroms. 105

Figure 5-18 presents the projections of typical configurations of adsorbed hydrogen molecules on the xz-plane. Figure 5-18(left) shows the small extension of the domain where the molecules can be inserted when the nanotubes have closed ends and are close packed; Figure 5-18(right) shows that the contribution of the interstitial space and external surface of nanotubes to the adsorption process is similar to that due to the internal volumes and surfaces of nanotubes, from which it can be concluded that the opening of CNTs is essential for hydrogen adsorption. In the domain of temperatures and pressures considered, the values D = 13.3 Å and d = 6 Å correspond to adsorption close to the best possible properties. By contrast, for d = 3.4 Å, the adsorption capacity of the nanotube material is low, but it is useful to notice that this value of the distance between nanotube walls could be that existing in the ropes of synthesised nanotubes. Furthermore, the adsorption at temperatures 150 and 293 K was several times less than the observed uptake at 77 K.105

Dodziuk et al. have carried out molecular mechanics calculations and molecular dynamics simulations for systems consisting of (5, 5) armchair, (9, 0) zigzag and (7, 3) chiral nanotubes. Contrary to many other calculations in which the condition of the nanotubes rigidity is imposed and hydrogen molecules are treated as spheres, they used both CVFF (consistent valence force field) parameterisations and ESFF (extensible systematic force field) parameterisations, which uses a rule based algorithm to determine a potential parameter set for a given system. These force fields do not imply such limitations. These calculations indicate that there is no essential difference among armchair, zigzag and chiral nanotubes as concerns their ability to host hydrogen molecules inside them. Furthermore, it is concluded that the total amount of hydrogen inside the nanotubes is very small and H2 molecules are not being adsorbed at higher temperatures. This result agrees with several other reports. From this, it is concluded that the literature reports on the very high hydrogen uptake cannot be obtained by physisorption process only.103


5.4 Supercapacitors

5.4.1 Introduction

In the future, supercapacitors might become an excellent means of certain kinds of energy storage. These electrochemical capacitors have a long durability (over 106 cycles), don't suffer from short circuit conditions, have a complete discharge and possess a high power density. Loading of a supercap can be performed at high current densities, which decreases the loading time needed. However, their energy density is lower than for conventional batteries, which is a possible drawback for possible applications. Typical electrochemical accumulators, in which compounds only take place in redox reactions, cannot fulfil these good characteristics that electrochemical capacitors have. Supercapacitors have already been applied in small-scale energy storage devices, such as in memory backup devices. Now the capability of supercapacitors with a high power density is increasing, potential applications extend to hybrid battery/supercapacitor systems.111

Carbon in general, and especially nanotubes, form an attractive material for electrochemical applications as they have a large active surface area. In addition, carbon is a relatively cheap, low density, environmentally friendly and highly polarisable material which makes application even more attractive.112

At first this section deals with basic processes in supercapacitors. Then, the determination of supercapacitor properties is explained briefly. Thereafter, supercapacitors based on CNTs are investigated after which attention is given to the modification of CNTs in supercapacitors to improve their properties.

5.4.2 Basic principles of supercapacitors

The basic principle of energy storage in a supercapacitor is based on creating a charge-separated state in an electrochemical double layer. In this case, energy storage is based on the separation of charges in the double layer across the electrode/electrolyte interface. The positive electrode is electron deficient whereas the negative electrode contains a surplus of electrons. The energy (W) stored in a capacitor as a function of the voltage applied (U) and the capacity (C) is equal to: .

Figure 5-19: Scheme of an electrochemical capacitor.

The electrodes of a supercapacitor must be electrochemically stable, which is the case for chemically unmodified carbon. The decomposition voltage of the electrolyte determines the maximum operating voltage of a supercapacitor. For the generation of high voltages, aprotic electrolytes with a decomposition range between 3 and 5 V should be used. However, these liquids only have a fraction of the conductivity that water has. In addition, the use of an aprotic electrolyte has technological, economical and safety barriers.112 The final electrolyte choice depends on the demanded specific power and energy values.

Electrochemical capacitors based on carbon are of two different types depending on the type of energy storage. The first type is the electrical double layer capacitor (EDLC) where only a pure electrostatic attraction between ions and the charged surface of an electrode takes place. The second type is a supercapacitor (SC), which is additionally based on faradaic supercapacitance reactions. The total capacitance C is determined by the series capacitances of the anode (CA) and cathode (CC) according to the equation112: .

In the EDLC, the contact between the electrode surface and the electrolyte plays an important role and determines the amount of charge stored. The capacitance C is proportional to the surface area S and the permitivity e of the electrolyte and reciprocally dependent on the distance of charge separation: . In practice, the surface area determined by gas adsorption differs from the active surface area available for charged species. When ions are solvated by water molecules, their mean diameter is approximately 15 Å. This explains the need for a relatively large pore size of the electrode material for good interaction of ions with the electrode. Thermal treatment of the electrodes results in significant alteration of the pore size distribution and thus to enhanced interaction.1 Increasing the capacitance values 10 to 100 times is possible by using pseudocapacitance effects. These depend on the surface functionality of carbon and/or on the presence of electro-active species.112

As said before, in supercapacitors faradaic reactions similar to processes in accumulators occur. Pseudocapacitances arise when the charge q, required for the progression of an electrode process, is a continuously changing function of potential U. These pseudocapacitance effects, for example electrosorption of H or metal ad-atoms and redox reactions of electroactive species, strongly depended on the chemical affinity of carbon materials to the ions sorbed on the electrode surface.112

An ideal double layer capacitance results in ideally rectangular shaped cyclovoltammetry diagrams. This phenomenon is ideal if the current density is independent of the potential applied and if this effect is purely electrostatic of nature. Due to redox peaks, pseudocapacitances result in deviations from this ideal shape as is shown in Figure 5-20.

ECDL capacitors can be represented in an equivalent electrical circuit as is shown in

Figure 5-21. This circuit consists of a capacitance C, a parallel resistance RF which is responsible for the self-discharge, an inductance L and an equivalent series resistance RS that models the internal resistance. A maximum value of C and a minimum value of RS result in a high value for the power density and energy value.

Figure 5-20: Typical charge/discharge voltammetry characteristics of an electrochemical capacitor112

The time constant for the charging and discharging cycles is equal to RS C, while the self-discharge time constant is equal to RF C. In order to minimise self-discharge the value for RF should be as large as possible.

Figure 5-21 Equivalent circuit for a real ECDL capacitor

5.4.3 Determination of supercapacitor properties

Key techniques to determine capacities are cyclic voltammetry, galvanostatic charge/discharge, external resistor discharge and impedance spectroscopy. Each technique reveals other specific information about the capacitor performance. Usually, a three-electrode set-up is used for material characterisation. In the resistor discharge method, the time needed to completely discharge through a certain resistor is measured.

The term "impedance spectroscopy" implies the dependence of impedance, or a kind of "generalised" resistance (V / I), on a certain frequency. This frequency is that of the power source used. For various reasons, the current might have a phase delay with regard to the voltage. Both this phase difference as the magnitude of the impedance (Z) play a key role in determining electrochemical mechanisms and electrode characteristics. In impedance spectroscopy, these parameters are determined as a the logarithm of the angular frequency (log(w)).113

Periodic processes, such as those in impedance spectroscopy, can be described in terms of real and imaginary parts of the parameters investigated. These complex plots, also called Cole-Cole plots, provide information about key parameters such as internal resistances and the kinetics of different processes.

5.4.4 Modification of CNTs

Carbon nanotubes have been proposed as electrodes for supercapacitors. Different values of capacitance mainly depend on the kind and purity of the samples. For purified nanotubes specific capacitance varies from 5 to 80 F/g. Pure carbon nanotubes have a moderate surface area (120 to 400 m2/g) because of their highly mesoporous character. The more graphitised nanotubes show smaller values of capacitance. However, presence of defects causes an increase of ability for accumulation of charges.114

There is a great difference in capacitance between SWNTs and MWNTs. However, the many different ways of producing the nanotubes also create differences in capacitance values. This can be seen in Table 5-4. Surface groups noticeably enhance the measured capacitance value, as is the case with MWNTsCo/700 °C modified by HNO3.

Table 5-4: Specific capacitances of differently produced carbon nanotubes in F/g.115

To increase the capacitance of nanotubes it is possible to increase the electrode surface area or to increase the pseudocapacitance effects obtained by addition of special oxides or electrically conducting polymers (ECP) like polypyrrole (PPy). The ECPs have the advantage of lower costs compared to oxides. Another advantage is that the pseudocapacitance effects of ECPs are quite stable.116

Charging of the electrical double layer proceeds mainly in the micropores. These micropores are just limited present in the pure CNTs. So, we have to modify them to get mesopores as well as micropores. This is possible by activation of the nanotubes by KOH. The exact mechanism of this activation by KOH is not clear yet. Both activation and coating increases the capacitance of the nanotubes.117;118

The mesoporous character of as-produced nanotubes essentially determines their electrochemical properties. For charging of the double layer a developed surface area is needed, thus the presence of micropores is necessary. With chemical activation of pure MWNTs by KOH, microporosity is introduced (Figure 5-22).

This KOH activation is performed by Frackowiak et al.117 at 800 °C under argon flow. The KOH:C weight ratio is 4:1 during the entire process. Afterwards, the samples are washed with demineralised water.

Specific surface area is measured by nitrogen adsorption at 77 K after the samples are outgassed at 350 °C for 24 hours until pressure reaches 10-6 mbar.

The electrodes are prepared in the form of pellets with 85 wt% nanotubular material, 5 wt% acetylene black and 10 wt% polyvinylidene fluoride. The mass of the electrodes varies from 4 to 8 mg.

Three different types of electrolytic solutions are used: aqueous 6 M KOH, aqueous 1 M H2SO4 and organic 1.4 M TEABF4 in acetonitrile.

Two different types of MWNTs are used. One type is obtained by decomposition of acetylene at 700 °C on cobalt (12,5 %) supported on silica (A/CoSi700). The other type is prepared by decomposition of acetylene at 600 °C on Co particles from a solid solution of cobalt oxide and magnesium oxide (A/CoxMg(1-x)O).

The values of capacitance are estimated by voltammetry, galvanostatic charge/discharge cycling and impedance spectroscopy.

Figure 5-22: SEM images of normal CNTs (a) and activated CNTs (b).118

The specific surface of the A/CoSi700 nanotubes increased from 430 to 1035 m2/g and the micropore volume changed from nearly zero to 0.47 cm3/g after activation. For A/CoxMg(1-x)O nanotubes the specific surface area increased from 220 to 885 m2/g and the micropore volume increased to 0.40 cm3/g from almost zero. These are significant changes. For the latter type an additional effect is that the tips are opened by KOH activation. Microporosity is now enhanced by activation, but the mesoporous character is still present because of the entanglement and presence of a central canal.

The results of the capacitances of the different nanotubes in three different electrolytes are given in Table 5-5. The capacitances for the unactivated nanotubes are between 10 and 15 F/g.

The values of capacitance were estimated by voltammetry with a scan rate of potential from 1 to 10 mV/s and galvanostatic charge/discharge cycling from 0 to 0.6 V or higher voltage limitation until 1.2 V.

electrolyte

A/CoSi700

A/CoxMg(1-x)O

activated

activated

1 M H2SO4

95 F/g

85 F/g

6 M KOH

---

90 F/g

1.4 M TEABF4

65 F/g

65 F/g

not act.

act.

not act.

act.

surface area (m2/g)

430

1035

220

885

Table 5-5: Capacitances of the activated MWNTs. The unactivated nanotubes show capacitances between 10 and 15 F/g. The surface area of the activated and unactivated nanotubes is also given.

The modification of carbon material by a specific additive providing quick pseudo-capacitance redox reactions is another way to enhance capacitance. This is possible with metal oxides, but in this case the addition of ECP is used. ECP itself has a capacitance of about 90 F/g. Pseudocapacitance effects of ECP are relatively stable. If we coat a nanotube with, for instance, polypyrrole we take the profit of the good electronic conducting properties and keep the advantage of ionic conductivity in the opened mesoporous network of the nanotube. These are perfect conditions for a supercapacitor.

Jurewicz et al.116 took five nanotube samples, all of them made in a different way and measured the capacitance of them with and without a layer of PPy coated on the surface. The specific surface area of the nanotube materials was measured by nitrogen adsorption at 77 K. The chemical polymerisation of pyrrole on the nanotubes was performed with ammonium persulfate as an oxidant. The thickness of the PPy layer was about 5 nm.

Electrodes were either from bucky paper or pellets of a mixture of MWNTs (85 wt%), acetylene black (5 wt%) and polyvinylidene fluoride (10 wt%). The aqueous electrolyte was 1 M H2SO4.

The results of capacitance measurements on the different nanotubes are given in Table 5-6. It can be concluded that the nanotubes with electrochemically deposited polypyrrole give much higher values of capacitance than the untreated samples. This proofs that the properties of both materials are used in a nice way. In Figure 5-23 the difference in voltammograms of Hyperion with and without coating can be seen.

Figure 5-23: Potentiodynamics characteristics (2 mV/s) of a capacitor assembled in 1 M H2SO4 from Hyperion nanotubes without PPy (left) and with PPy (right). Mass of each electrode was 3.5 mg.115

Striking is the effect with the coated nanotubes of the type P/800Al. In this case it might be expected that a thin PPy film also covers the inner core which is also accessible for the electrolyte.

A couple of the capacitors have been cycled over 2000 cycles and the charge loss never exceeded 20%. So, coating of the nanotubes with polypyrrole seems a nice way to enhance capacitance and efficiency for a long durability.

Table 5-6: Capacitance values (F/g) of the nanotubular material in acidic medium (1 M H2SO4) with and without PPy.116