Sie sind auf Seite 1von 8

Carbon 40 (2002) 307314

Grafted and crosslinked carbon black as an electrode material for double layer capacitors
Roy Richner, Stefan Muller*, Alexander Wokaun
Paul Scherrer Institut, Electrochemistry Laboratory, Villigen PSI, Switzerland Received 16 January 2001; accepted 29 March 2001

Abstract Isocyanate prepolymers can readily react with oxidic functional groups on carbon black. Different prepolymers were examined for their grafting behaviour. In carbon black grafted with di-isocyanates, reactive isocyanate groups are available for cross-linking to a polyurethane system. This crosslinked carbon black was designed as a new active material for electrochemical electrodes. The successful crosslinking of carbon black was demonstrated by determinations of the grafted groups via titration and via thermogravimetric analysis. Products of this type were tested as an active material for electric double layer capacitor application. Active material for electric double layer capacitor electrodes was produced which had a specic capacitance of up to 200 F / g. Crosslinking efciencies of up to 58% of the utilized polymers were achieved. 2002 Elsevier Science Ltd. All rights reserved.
Keywords: A. Carbon black; Graphite; B. Chemical treatment; D. Functional groups; Surface properties

1. Introduction Carbon black has received increasing attention for application in catalysis and electrochemistry in recent years. It is used as catalyst carrier in fuel cell electrodes [1], in bifunctional oxygen electrodes for secondary metal air batteries [2,3], in insertion electrodes for lithium batteries [4] as well as in electrical double-layer capacitors (EDLC) [5]. Commercial EDLCs depend on the high surface area available in carbon electrodes and on the resulting double-layer capacitance of such electrodes in electrolyte solution. The electrodes consisting of high surface area carbon-made active material used in these devices are separated by a suitable porous matrix impregnated with an electrolyte [6]. According to Qu and Shi [7] the specic double layer capacitance is proportional to the surface of the meso- and macropores of the carbon. Apart from their high surface area, carbon blacks possess chemical surface groups. Their nature and number depends on the production process and on pretreatment. Carbonoxygen compounds are the most important surface groups inuencing the physicochemical properties of carbon blacks [8]. The effect of carbon surface groups in
*Corresponding author. Fax: 141-56-310-4415. E-mail address: stefan.mueller@psi.ch (S. Muller).

EDLC electrodes is controversial. In certain applications the presence of these groups appears desirable owing to the contribution provided to the total capacitance by the pseudocapacitance of the corresponding redox couples [9]. In a contrasting view, it is assumed that the double-layer capacitance only depends on the size of the surface area, while surface groups play a minor role and should actually be avoided because of their negative inuence on EDLC stability [10]. However, these surface groups can be utilised for the grafting of polymeric precursors or prepolymers. Reactions of carbon bres with polymers are already used to enhance the interactions between matrix and ber in ber reinforced plastics (FRP) [11]. The grafting and subsequent build-up of polymers on carbon surfaces have been studied thoroughly [12]. The grafting of diisocyanate prepolymers are of particular interest [13,14]. Conventional active materials for carbon-based capacitors are prepared from a suspension of carbonaceous materials in a binder solution or with suspended binder, which is applied to a current collector by a coating technique to form electrodes [5,6]. An adhesive binder will always tend to block a large part of the carbons high surface area. With larger amounts of binder, a higher mechanical stability is attained but the capacitance and thus the measured performance of the electrode is strongly

0008-6223 / 02 / $ see front matter 2002 Elsevier Science Ltd. All rights reserved. PII: S0008-6223( 01 )00100-2

308

R. Richner et al. / Carbon 40 (2002) 307 314

diminished by occlusion of the ne pores which provide the major contribution to a high double-layer capacitance. Prior investigations performed at our laboratory with Black Pearls 2000 and PTFE had conrmed these limitations of the conventional system [15,16]. In this paper we describe a new concept according to which prepolymers are grafted onto carbon black. These prepolymers are then crosslinked to form an active material. From grafted diisocyanate prepolymers and polydiols a polyurethane network can be built-up which incorporates the carbon black via chemical bonds. The problem that an important fraction of the pores becomes occluded in the process can be circumvented by selecting prepolymers of an appropriate size. Crosslinked active material is expected to have some major advantages over the conventional active material. We examine the reactions of appropriate prepolymers with carbon black powder with a view to achieving enhanced adhesion of the binder and forming electrode materials retaining the unique surface area and porosity of the constituent carbons, which is a prerequisite for the application in EDLCs. The degree of grafting of poly(1,4butanediol), tolylene 2,4-diisocyanate terminated (TDIPBG) was determined via amine consumption of unreacted isocyanates and back-titration of excess amine by acid base titration. The degree of grafting was veried by thermogravimetric analysis. Scanning electron microscopy (SEM) was used to compare extruded with cast carbon black network material.

submitted for TGA as described below. For crosslinking, a solution of 0.24 g p(THF) and 0.1 ml TMP-EO (Mn 5 740 g / mol) in 10 ml THF was added to the above slurry. After several hours of reaction, the new slurry could be used for electrode preparation. Electrodes were prepared by techniques such as casting on a current collector, or extruding. Instead of TDI-PBG, 0.3 g of TDI-PPG 1k or 0.69 g of TDI-PPG 2k were used in some grafting experiments.

2.3. Titration
An amount of analyte containing ca. 0.8 mmol of free isocyanate groups was slurried with 75 ml THF. Then, 10.00 ml of a 0.5 mol / l n-butylamine solution in THF (e.g. 1.8285 g butylamine in 50.0 ml THF) were added. Isopropanol (60 ml) was also added to make the slurry miscible with the aqueous acidic solution. The butylamine not converted was then back-titrated with 0.1 molar aqueous HCl-solution. A combined glass-electrode was used to determine the endpoint potentiometrically. The original isocyanate group content of each prepolymer was determined by titrating some quantity of prepolymer alone. The blank of butylamine consuming surface groups of Black Pearls 2000 and KS44, respectively, was also titrated. For the titration of grafted carbonaceous material the reaction slurry in anhydrous THF was quenched with 10.00 ml of the n-butylamine solution and then treated as described above. Calculation of the grafted fraction (F ) from the equivalents (Eq) of butylamine consumed and the mass (m) of the relevant species in the respective titration: F (%) 5 EqCB 3 m CB( 3 ) EqDI 3 m DI( 3 ) 2 3 ]]]] 1 ]]]] 2 EqGr m CB( 1 ) m DI( 2 ) ]]]]]]]]]]]] 3 100% EqDI 3 m DI( 3 ) ]]]] m DI( 2 ) where CB is carbon black, DI is diisocyanate, Gr is grafted sample, (1) is carbon black blind titration, (2) is diisocyanate prepolymer titration and (3) is grafted sample titration. The diisocyanate groups on one end are expected to decrease proportionally to the grafted amount. When 50% of the titratable diisocyanate groups are left, all prepolymer molecules are grafted to the carbonaceous material on one end. The obtained carbonaceous material is expected to have a variety of butylamine consuming groups which do not change signicantly after addition of the diisocyanate. The blank measured for the carbonaceous material is therefore taken into account additively. The weight gain was obtained from the titration results by multiplying the grafting efciency with the percentage of polymer utilised.

2. Experimental

2.1. Materials and electrodes


Carbon black and graphite used were Black Pearls 2000 (Cabot, 1475 m 2 / g BET) and KS44 (Timcal, 7 m 2 / g BET), respectively. Prior to grafting or crosslinking, the carbonaceous materials were dried overnight in a vacuum at 1008C and then stored under nitrogen. Prepolymers used were poly(1,4-butanediol), tolylene 2,4-diisocyanate terminated, number average molecular weight (Mn ) 5 1800 g / mol (TDI-PBG), poly(propylene glycol), tolylene 2,4diisocyanate terminated, Mn 5 1000 g / mol (TDI-PPG 1k) and Mn 5 2300 g / mol (TDI-PPG 2k). Extenders used were poly(tetrahydrofuran), Mn 5 1400 g / mol p(THF) and trimethylolpropanole ethoxylate, Mn 5 740 g / mol (TMPEO).

2.2. Grafting and crosslinking


The suspension used to produce the grafted material was prepared by stirring up 1.5 g of Black Pearls 2000 or KS44 in a solution of 0.54 g TDI-PBG in 30 ml tetrahydrofuran (THF). The components were allowed to react for 1 h at room temperature, then this slurry could be titrated or

R. Richner et al. / Carbon 40 (2002) 307 314

309

2.4. Continuous extraction /thermogravimetric analysis tests


Black Pearls 2000 was grafted (and crosslinked respectively) and the reaction slurry was quenched with the required amount of butylamine solution after the grafting reaction was complete. After evaporation of the solvent and drying, the residue was divided in two parts, one was immediately used for TGA while the other part was continuously extracted for a period of 7 days with THF in a soxhlet apparatus and then dried and analysed by TGA. The weight gain was obtained by heating the sample with a ratio of 58C / min to 4758C under a nitrogen atmosphere.

taken with a TOPCON ABT-60 with an acceleration voltage of 10 kV. The samples were xed to the sample holders with a contacting graphite glue.

2.7. Pore size distribution


Pore size distributions (PSD) of the respective carbonaceous particles were obtained by low angle light scattering (LALLS), laser diffraction with a Malvern MasterSizer X laser diffractometer with optics of 45 mm focal length. The measurements were calculated to distributions with a very polydispersive model of spherical particles which behave according to the Fraunhofer approximation. Carbon black was measured in 2-propanol. The PSD was measured before, directly after 1 min of ultrasonic agitation and 4 h after the ultrasonic agitation.

2.5. Electrochemical impedance spectroscopy


The suspension of grafted and crosslinked carbon in THF was processed to different electrode types: extruded as lament, cast in nickel foam or rolled into a nickel net. (i) Extruded lament: the suspension was lled into a disposable syringe and left to settle. Then, the supernatant solution was poured off. The remaining, highly viscous suspension was extruded forming long, thin laments with a diameter of 0.5 mm, which rapidly dried. (ii) Cast nickel foam: the suspension was ltered through nickel foam, which retained the active material. Afterwards, the paste (incorporated in the foam) was dried. The foam had a thickness of 2 mm and a porosity of 45 pores per inch (ppi). (iii) Rolled nickel net: the extruded laments of (i) were rolled into a nickel net before drying, when still ductile. Impedance measurements were carried out with a Zahner EIS Module in a three electrode arrangement with a 1 M solution of tetraethyleneammonium bortetrauoride (TEABTF) in acetonitrile as the electrolyte. The counter electrode was a glassy carbon disc and a Ag /AgNO 3 wire served as the reference electrode. The impedance spectra were measured over the frequency range of 10000.01 Hz with an amplitude of 0.01 V.

3. Results and discussion

3.1. Grafted carbon black


Isocyanate-terminated polyethers were reacted with carbon black and graphite as the carbonaceous materials as described in Section 2. The reaction sites are oxidic surface functions such as phenol or carboxylic acid groups, as indicated in Scheme 1. The graphite and carbon black have been chosen in regard to the oxidic surface groups available. According to its morphology and surface the graphite is expected to have nearly no surface groups in contrast to the carbon black which should contain a sufcient amount of these groups to yield a reasonable grafting efciency. Results for the thermogravimetric analysis and the titration of grafting reactions are given in Table 1. Column 1 shows the proportion of prepolymer used relative to the total amount of carbonaceous material and prepolymer. Column 2 contains the grafted fraction calculated as given in Section 2.3. The weight gain in column 3 is calculated from the value of column 2 multiplied with the respective value of column 1. The weight gain in column 4 was received from the thermogravimetric analysis after extraction as described in Section 2.

2.6. Scanning electron microscopy


The scanning electron microscopy (SEM) pictures were

Scheme 1.

310

R. Richner et al. / Carbon 40 (2002) 307 314

Table 1 Grafted fractions and weight gains of grafted carbons System Prepolymer used (wt.%) 16.7 31.5 26.5 16.7 31.5 26.5 Grafted fraction calculated from titration (%) 55 33 85 16 7 16 Weight gain calculated from titration (wt.%) 10 13 23 3 2 4 Weight gain calculated from TGA after extraction (wt.%) 8 10 18 0 0 0

Black Pearls 2000 TDI-PPG 1k Black Pearls 2000 TDI-PPG 2.3k Black Pearls 2000 TDI-PBG 1.8k KS44 TDI-PPG 1k KS44 TDI-PPG 2.3k KS44 TDI-PBG 1.8k

These data reveal a high grafting efciency of 85% for the Black Pearls 2000 / TDI-PBG system. Both the titration of the carbon and the TGA analysis after extraction of the adsorbed polymer from the Black Pearls 2000 revealed weight gains for the TDI-PPG system in the order of 10%. Grafting efciencies calculated from the weight gain deviated somewhat from the ones determined by titration, both for Black Pearls 2000 and for KS44. The results obtained in titrations of grafted samples of graphite KS44 yield a reasonable explanation for this result. In fact, this graphite is expected to have practically no surface groups, and therefore all isocyanate groups added to the slurry should be back titratable in these samples. According to the titrations, some isocyanate groups remain adsorbed on the graphite, and are responsible for the non-zero result. Wu [14] postulated the formation of loops (ring formation of diisocyanate) responsible for higher values of grafting efciencies obtained in titration analysis of grafted carbon black. On the other hand the results of the TGA analysis for KS44 shows that this is a more systematic error of the titration analysis.

3.2. Crosslinked carbon black


The prepolymer molecules used for grafting are known

to undergo cross-linking with diols and triols. Therefore attempts were made to crosslink correspondingly grafted carbonaceous materials. A diol / triol mixture was added to the slurry of grafted carbon prepared as above. The remaining isocyanate reaction sites were then expected to react with the hydroxylic groups as shown in Scheme 2. This reaction sequence was performed with different prepolymers, both on carbon black and on graphite. The slurry obtained after reaction was cast and an extraction analysis was performed as described above. The results are listed in Table 2. Column 1 shows the proportion of used polymer (prepolymer and diol / triol) relative to the total amount of carbonaceous material and polymer. The total immobilised fraction of polymers (column 2) was calculated from the proportion of the total bound polymer remaining after extraction (weight gain, column 3) to the total used polymer (column 1) received through TGA. Up to 58% of the binder utilised was integrated into the Black Pearls 2000 structure. The overall immobilised or bound fraction (column 2) of the utilised polymer was higher than the grafted fractions obtained through titration for the TDI-PPG 2.3k prepolymer, in the same region for the TDI-PPG 1k prepolymer and somewhat lower for the TDI-PBG prepolymer. A higher total immobilised fraction of polymer than the grafted fraction was expected for all

Scheme 2.

R. Richner et al. / Carbon 40 (2002) 307 314 Table 2 Immobilised fractions and weight gains of crosslinked carbons System Polymer used (wt.%) 28.2 33.9 39.5 27.4 38.0 17 32 Total immobilised fraction of polymers by extraction / TGA of cast samples (%) 53 56 51 58 55 0 0

311

Weight gain calculated from extraction / TGA of cast samples (wt.%) 15 19 20 16 21 0 0

Black Pearls 2000 TDI-PPG 1k (a) (b) Black Pearls 2000 TDI-PPG 2.3k Black Pearls 2000 TDI-PBG 1.8k (a) (b) KS44 TDI-PPG 1k KS44 TDI-PPG 2.3k (a), (b), two distinct experiments

systems. The reason for this expectation was the presence of residual ungrafted free diisocyanate and extender (diol / triol). The extender should link not only to the isocyanate sites grafted onto the carbonaceous material, but also to ungrafted free diisocyanate still present in the material, thereby building up a crosslinked network. Too much is unknown to make an estimation of the amount of bound extender in the crosslinking step. The size and extent of the side reactions are unknown (i.e. formation of ureas and allophanates). Besides this, the residual amount of ungrafted diisocyanate from the grafting step is ambiguous. Hence the estimation of free diisocyanate for the crosslinking step is very unprecise. Nevertheless, the weight gain of the respective crosslinked sample has increased as compared to the grafted sample. Added extenders are denitely incorporated, while at the same time immobilizing additional free diisocyanate molecules. The increase of the amount of immobilised polymer with the crosslinking can be regarded from an engineering point of view as the part of the binding responsible for the cohesion of this active material. It was observed that the grafted carbonaceous material never showed a mechanical stability even when large amounts of diisocyanate were applied. The produced active material only had a sufcient mechanical stability when the extenders were applied. The weight gain of KS44 samples submitted to the crosslinking reaction were zero as expected. A single extruded lament made from the reaction slurry using the TDI-PBG prepolymer and a total weight fraction of 33.9% of polymer in the active material was also submitted to TGA. The weight gain of this extracted lament, which had not been subjected to continuous extraction, was equal to 25% as compared to values around 20% for cast and extracted material in Table 2. This shows, that most of the unbound polymer is pressed out with the excess solvent during extrusion. Only about 40%

of the unbound polymer remains in the extruded laments acting as an ordinary binder and disturbing the porosity of the active material.

3.3. Electrodes from crosslinked carbon black


The next step was the production of electrodes from this active material to measure its electrochemical properties by EIS. Black Pearls 2000 was grafted as described in Section 2. The grafted carbon was then converted with a diol / triol mixture to a crosslinked active material. The slurry obtained after reaction was cast into metal foam (current collector) or extruded. The extruded laments were contacted with a graphite binder to a titanium foil as a current collector. A third possibility of electrode preparation was the pasting of freshly extruded laments (still wet and ductile) into a nickel net. Electrochemical impedance spectrometry (EIS) was used to study the electrochemical properties of these electrodes. Impedance spectra for laments and for pasted elctrodes (Ni-net current collector) are shown in Fig. 1. The diagrams reveal a higher grain to grain resistance for extruded laments than for rolled electrode samples, observable by the sharp increase in resistance in the frequency region of 110 Hz. This can be explained by the longer pathway of the current through the lament (35 mm length), which is in contact with the current collector at one end only. In contrast, the rolled electrodes only have a thickness of 0.3 mm. The plot of calculated specic capacitance versus frequency is shown in Fig. 2. The high specic capacitances demonstrate the potential of the crosslinked active material for supercapacitor applications. Electrodes with the crosslinked carbon black incorporated in a metallic foam as the current collector exhibited the highest specic capacitances, of up to 200 F / g at 0.01 Hz. For a practical realisation, the mechanical stability and electrical conductivity of this electrode material must still be improved.

312

R. Richner et al. / Carbon 40 (2002) 307 314

Fig. 1. Electrochemical impedance spectra of electrodes made from crosslinked carbon black Black Pearls 2000.

To explain this behaviour the physical structure of the different electrodes prepared in the present work was investigated by scanning electron microscopy (SEM). The SEM pictures of cast and extruded samples are presented in Fig. 3. Some distinct differences can be discerned between fracture sites in cast and extruded carbon materials. The cast material clearly reveals carbon black aggre-

gates as if suspended in the solvent. The shearing energy imparted to the slurry by stirring had little effect on the grain-size distribution in cast samples. The grains retain the same shape as in the slurry but are now held together by the binder. In the extruded sample of the crosslinked carbon, where substantial shear forces have acted upon the slurry during extrusion, the mass is quite homogeneous, and no grain structure can be observed. As a result of thixotropy exhibited by the carbon grains under shear stress, the binder surrounding the carbon and the carbon itself are smeared into a homogeneous compound retaining its shape after extrusion. To understand these observations the morphology of carbon black has to be considered. Black Pearls 2000 used as carbon black consists of primary nanoparticles fused together to build aggregates or clusters. These clusters in turn are held together by weak forces or in electrodes by the binder system, either adhesively or by bound polyurethane. The loosely cohering clusters can be fragmented or agglomerated in suspension by appropriate techniques such as ultrasonic agitation or strong shearing by mechanical homogenisers. The clusters are also deformed through the shearing energy occurring during extrusion. The inuence of ultrasonic agitation on the particle size distribution (PSD) of carbon black suspended in 2-propanol was therefore further investigated by laser diffraction. The PSD of the as-received Black Pearls 2000 is shown in Fig. 4 (thick line). A major part of the particles are agglomerated to clusters larger than 50 mm in diameter. When these agglomerates are exposed to ultrasonic agitation, increasing fragmentation occurs, as shown in the broken line of Fig. 4. A broad peak of smaller fragments is observed at 18 mm. These fragments tend to reagglomerate in the suspension, as shown in the thin line of Fig. 4. The

Fig. 2. Plots of specic capacitance vs. frequency for electrodes made with crosslinked Black Pearls 2000 of different preparation.

R. Richner et al. / Carbon 40 (2002) 307 314

313

Fig. 3. SEM picture of (a) cast and (b) extruded lament active material of crosslinked carbon black Black Pearls 2000.

observations of the agglomerating behaviour received by these experiments are thought to be transferrable to the observations made with the crosslinked carbon. The thixotropic behaviour of the carbon black suspension may be responsible for the differences in capacitance exhibited on one hand by the rolled and extruded electrodes and on the other hand by the cast samples. The observed rearrangement probably makes most of the micropores and even a fraction of the mesopores inaccessible to the electrolyte solution. The capacitance plots support this assumption. In the range of frequencies above 1 Hz where micro and mesopores hardly contribute to the double-layer capaci-

tance, the specic capacitance becomes independent of the procedure used to prepare the electrodes.

4. Conclusions The conducted investigations clearly show the crosslinking of the carbon black with the assembled reactive polyurethane system. The incorporated carbon black Black Pearls 2000 was crosslinked by its reactive surface groups. With this procedure an electrode material with up to 20% of binder content could be generated with a specic

Fig. 4. Particle size distribution determined by laser diffraction in carbon slurries. Samples were exposed to ultrasonic agitation (US) for fragmentation.

314

R. Richner et al. / Carbon 40 (2002) 307 314 [6] Bonnefoi L, Simon P, Fauvarque JF, Sarrazin C, Sarrau JF, Dugast A. Electrode compositions for carbon power supercapacitors. J Power Sources 1999;80(12):14955. [7] Qu DY, Shi H. Studies of activated carbons used in doublelayer capacitors. J Power Sources 1998;74(1):99107. [8] Kinoshita K. Carbon, electrochemical and physicochemical properties, New York: John Wiley, 1988. [9] Sarangapani S, Tilak BV, Chen CP. Materials for electrochemical capacitors theoretical and experimental constraints. J Electrochem Soc 1996;143(11):37919. [10] Nakamura M, Nakanishi M, Yamamoto K. Inuence of physical properties of activated carbons on characteristics of electric double-layer capacitors. J Power Sources 1996;60(2):22531. [11] Chou T-W. Structure and properties of composites. In: Cahn RW, editor, Materials science and technology, vol. 13, Weinheim: VCH Verlagsgesellschaft, 1993. [12] Tsubokawa N. Functionalization of carbon black by surface grafting of polymers. Prog Polym Sci 1992;17:41770. [13] Tsubokawa N, Kobayashi K, Sone Y. Grafting onto carbon black by the reaction of reactive carbon black having masked isocyanate or acyl azide group with functional polymers. J Polym Sci 1988;26:22333. [14] Wu ZH, Pittman CU, Gardner SD. Grafting isocyanateterminated elastomers onto the surfaces of carbon bers: reaction of isocyanate with acidic surface functions. Carbon 1996;34(1):5967. [15] Richner R, Bartschi M, Muller S, Kotz R, Wokaun A. Comparison of specic energy and power for differently bound carbonaceous materials in supercapacitors. In: Extended abstracts, 51st ISE-Meeting, Warsaw, Poland, 1999, Abstract no. 639. [16] Richner R, Muller S, Kotz R, Wokaun A. Studies of activated carbon and carbon black for supercapacitor applications. In: PSI Annual Report, Annex V, 1998, p. 47.

capacitance of up to 200 F / g. The impedance spectra showed a low level of affection of the pore structure of the carbon black. The manufacturing process has to be optimised to enhance mechanical stability without occluding the micropores during the process which is a result of too high shearing forces. This also means that the pore size distribution can be directed by the manufacturing process. A graphite chosen for comparison showed no reactivity to the isocyanate system and could not be crosslinked as it had no surface groups. This active material is promising for application in electrochemical double-layer capacitors.

Acknowledgements Financial support by the Board of the Swiss Federal Institute of Technology is gratefully acknowledged. The authors thank Mr. D. Kunz for excellent technical support.

References
[1] Stonehart P. Carbon substrates for phosphoric acid fuel cell cathodes. Carbon 1984;22(4):42331. [2] Muller S, Holzer F, Arai H, Haas O. A study of carbon catalyst interaction in bifunctional air electrodes for zincair batteries. J New Mater Electrochem Systems 1999;2:22732. [3] Arai H, Muller S, Haas O. AC impedance analysis of bifunctional air electrodes for metalair batteries. J Electrochem Soc 2000;147(10):358491. [4] Winter M, Besenhard JO, Spahr ME. P. N. insertion electrode materials for rechargeable lithium batteries. Adv Mater 1998;10(10):725. [5] Kibi Y, Saito T, Kurata M, Tabuchi J, Ochi A. Fabrication of high-power electric double-layer capacitors. J Power Sources 1996;60(2):21924.