• Carbon nanotubes;
  • Electrodes;
  • Energy storage;
  • Supercapacitors

Supercapacitors are electrochemical energy storage systems that store energy directly and physically as charge, whereas batteries, for example Li-ion cells, store energy in chemical reactants capable of generating charge.1 Accordingly, the energy density of supercapacitors (<10 Wh kg−1) is lower than batteries (>100 Wh kg−1). However, their power is significantly higher and their lifetime longer. As such, supercapacitors are expected to play a crucial role where superior power performance is required. The importance of supercapacitors is highlighted by a report from the US Department of Energy assigning equal importance to batteries and supercapacitors.2 Examples of envisioned large-scale applications of supercapacitors are load-leveling in solar, wind, and other energy sources and energy recovery from regenerative braking in automobiles.2, 3 To emerge as an important energy storage technology in the future, advanced supercapacitors must be developed with higher operating voltage and higher energy and power delivery, while maintaining high cyclability.

Hitherto, activated carbon (AC) has been the electrode material of choice due to its high surface area (1000–2000 m2 g−1).4, 5 However, the particulate nature, uncontrolled surface functional groups, and ill-defined structure of AC powders hinder their capacitance and result in long-term degradation. Additionally, to make electrode sheets from AC powders, binders and conducting agents are needed, further limiting the operating voltage range and compromising device lifetime. On the other hand, nanomaterials, as exemplified by carbon nanotubes (CNTs), can have controlled chemical composition and tailored physical architectures down to nanoscale dimensions, and thus address the fundamental issues plaguing AC.6

CNTs, particularly single-walled carbon nanotubes (SWNTs), satisfy all the fundamental requirements for supercapacitor electrodes. They have well-defined structures with high surface area, consist of only carbon with no surface functional groups, are conductive and as one-dimensional fibers transformable into electrode sheets without any binders. Since the first report7 that showed promising high power density (8 kW kg−1), there has been increasing interest in CNTs as supercapacitor electrodes. Thus far, CNT electrodes have not only achieved notable energy and power performance (7 Wh kg−1, 20 kW kg−1),8 they have also enabled new functionalities such as flexible9 and transparent10 supercapacitors. Moreover, nanostructured CNT composites11 have shown exceptionally high capacitance (100 F g−1) at high discharge rate (77 A g−1). Although each of these studies has shown advantageous individual properties of CNTs for supercapacitors, a comprehensive demonstration of their full potential as electrodes meeting all relevant criteria for practical devices has yet to be presented. Only through such evaluation can the true potential of CNT electrodes for supercapacitors be known. Moreover, previous studies of CNT electrodes have suffered from low surface area and chemical impurities hindering their performance.

In the work reported here, we made supercapacitor electrodes solely from the purest as-grown SWNTs available to achieve the full potential of CNT electrodes. By fabricating electrodes from SWNTs with high carbon purity (99.98%), exclusive SWNT selectivity (>99%), negligible carbonaceous impurity (<2% amorphous carbon),12 and near-ideal specific surface area (1300 m2 g−1),13, 14 we achieved operation at a higher voltage (4 V) while maintaining durable full charge–discharge cyclability, with an energy density (94 Wh kg−1, 47 Wh L−1) and power density (210 kW kg−1, 105 kW L−1) far exceeding those of AC both gravimetrically (33 Wh kg−1, 60 kW kg−1) and volumetrically (19.8 Wh L−1, 36 kW L−1).

In order to address the performance limits of CNTs, the key is to minimize impurities. Impurities are common in as-grown CNTsamples, and have greatly plagued the progress of CNT applications in the past. A truly pure CNT sample is difficult to achieve because metallic catalytic and carbonaceous impurities are generated by most synthesis methods and removal of these through purification introduces functional groups as new impurities. Such impurities are detrimental to supercapacitor performance as metallic particles and uncontrolled functional groups can cause parasitic reactions that limit the lifetime and carbonaceous impurities reduce the specific surface area, thus diminishing capacitance.

In this work we used vertically aligned SWNTs (forests) synthesized by water-assisted chemical vapor deposition,15 denoted as “supergrowth”, to achieve the lowest possible amount of impurities and a very high specific surface area. SWNT forests (3%–4% volumetric occupancy, mass density 0.03 g cm−3) were synthesized from iron catalysts (1 nm) sputtered on silicon wafers using ethylene and water. These forests achieved a high carbon purity (>99.98%), because the highly efficient growth process requires a minimal amount of catalyst, and the catalyst particles remain on the growth substrate during post-growth removal of the forest. Quantitative elemental analysis based on X-ray fluorescence spectrometry has shown the removed forests’ noncarbon content to be only a minute amount of iron (0.013 wt%).15 The catalyst layer thickness was precisely tuned to selectively (ca. 99%) grow SWNTs16 with a minimal amount of multiwalled CNTs, which have lower surface area. Also the growth condition was tuned with great care to synthesize a SWNT forest with minimal carbonaceous impurities, thus achieving an absolute purity (SWNT weight percent of sample) of 98.3%.12 These high purity levels meant that the forest was almost solely composed of single-walled carbon nanotubes with minimum impurities.

In fabricating the electrode sheet from these forests without using binders, we took advantage of the one-dimensional fiber-like nature of SWNTs to assemble them into sheets, similar to how paper is made from cellulous fibers. Figure1a schematically illustrates the key processes in fabricating the electrode solely from pure SWNTs. Specifically, the sparse forest removed from the growth substrate was sheared between glass slides and subsequently wetted by electrolyte. This process flattened and densified the forest similar to deflating an air mattress to pack the SWNTs into a highly densely packed solid form (density 0.5 g cm−3, area 1 cm2, thickness ca. 100 μm).17 This fabrication process is scalable, as demonstrated by the recent development of large-area synthesis of SWNT forests.18

thumbnail image

Figure 1. Electrode fabrication and characterization. a) SWNT electrode fabrication. b) SEM image of SWNT electrode showing ordered pore structure and alignment. c) SEM image of AC depicting random pore structure. d) Typical cell assembly. e) IR spectra of SWNTs showing presence of no (IR-active) surface functional groups. f) Thermogravimetric analysis of SWNT electrode heated (5 °C min−1) in nitrogen environment showing no weight loss.

Download figure to PowerPoint

For comparison, AC electrode sheets (density 0.6 g cm−3, area 1 cm2, thickness ca. 100 μm) were also made by a conventional process, that is, mixing and kneading of AC powder (YP17: 1640 m2 g−1) with a binder (poly(tetrafluoroethylene), PTFE) and a conducting agent (carbon black). Scanning electron microscopy (SEM) reveals the difference between the aligned well-defined structure of SWNT electrodes (Figure 1b) and the random irregular structure of AC electrodes (Figure 1c). In addition, the SWNT electrode is mesoporous, with a maximum pore size of 10 nm,13 and has similar surface area to that of the forest (1250 m2 g−1). Infrared spectroscopy (Figure 1e) and thermogravimetric analyses15 (Figure 1f) further confirmed the presence of no chemicals other than carbon within the SWNT. In the infrared spectra, the peak after 0.3 e. V. is related to the S11 transition associated with the bandgap energy of the semiconducting SWNTs present in the forest.13 The SWNT electrodes fabricated have comparable density and thickness to conventional electrodes while preserving the intrinsic chemical purity and high surface area of the SWNT forest. While electrodes with lower densities and/or thinner thickness would inherently have higher power capability due to enhanced ion transport, such electrodes would be unable to meet the weight and volume constrains of practical applications. AC and SWNT supercapacitors were assembled by stacking a separator in between two electrodes supported with platinum mesh current-collectors (MCCs, Figure 1d). In addition, SWNT devices without MCCs were made to utilize the higher conductivity of the SWNTs (21 S cm−1) compared to AC electrodes (0.3 S cm−1) to explore the dual functionality of SWNTs as electrodes and current-collectors.

First, and central to this work, we show that the SWNT electrodes with and without MCCs had an operating voltage range of 4 V, as demonstrated by the identical symmetric trapezoid-shaped cyclical voltammograms (CVs) (Figure2a). This is the highest reported voltage range for any (single-cell) battery or supercapacitor using conventional organic solvent-based electrolytes. In contrast the AC electrodes were limited to 3 V, as evidenced by the peak after 3 V, indicating parasitic chemical reactions initiated by impurities within the electrode. The higher operating voltage range of the SWNT electrodes is attributed to its purity, that is, no conducting agents or binders or surface functional groups. Operation at higher voltage significantly boosts the overall performance of the supercapacitor because both the energy and power are proportional to the square of the voltage.4, 5 A voltage range of 4 V would increase energy and power by at least 77% compared to 3 V, the operating limit of AC. In addition, operating at higher voltage offers yet another advantage for practical applications, by reducing the number of devices in series required to reach the desired output voltage.

thumbnail image

Figure 2. Electrochemical performance of AC and SWNT electrodes at 4 V. a) Cyclic voltammograms of the SWNT and AC electrodes measured at 1 mV s−1. The SWNT electrodes show stable operation up to 4 V as evidenced by the symmetry between the charge and discharge sweeps. In contrast, the peaked shaped CV of AC suggests irreversible parasitic reactions. b) Galvanostatic discharge at 1 A g−1. The SWNT electrodes possess higher capacitance/energy compared to AC as demonstrated by the longer discharge times. c) Capacitance vs. discharge current. At any discharge rate the SWNT electrodes provide higher capacitance compared to AC. d) Potential drop associated with the cell internal resistance (IR loss) vs. discharge current. The SWNT cells have lower IR losses compared to AC. (Values are normalized with respect to combined dry weight of working and counter electrodes.)

Download figure to PowerPoint

To evaluate the performance of the electrodes at 4 V, we measured galvanostatic discharges (Figure 2b) at various rates. The capacitance (Figure 2c) and the internal resistance values were then determined from the slope and the initial voltage drop (Figure 2d) of the discharge curves, respectively. First, the specific capacitance of the SWNT electrodes with MCC (blue) reached 160 F g−1, did not vary with increasing discharge rates, and was much higher than the 100 F g−1 specific capacitance of the AC electrodes, which also decreased further with higher discharge rates. Previous studies at ca. 3 V have reported comparable or lower capacitance for as-grown SWNTs compared to AC electrodes,17, 19 as the high surface area of SWNTs is comparable to or lower than that of AC. Here, for operation at 4 V, the capacitance of SWNTs has significantly increased (from 68 F g−1 at 2 V to 160 F g−1 at 4 V), exceeding that of AC. The increase of capacitance with voltage is due to electrochemical doping,20, 21 an important additional advantage of higher voltage operation for SWNT electrodes. The unvarying capacitance of SWNT electrodes with increasing discharge rates means fast ion transport, thus high power capability. While the capacitance of the SWNT electrode without MCC is lower than when a MCC is used, it still exceeds that of AC electrodes. This is due to the conductivity of the SWNT electrode being higher than that of the AC electrode yet still lower than for bulk metals. The important role of the MCC is to equilibrate the potential across the electrode plane, thus only a potential distribution in the thickness direction exists. However, without the MCC, the potential distribution would extend throughout the electrode, resulting in a lower capacitance.

Second, the internal resistance was the lowest for the SWNT electrodes with MCC (blue) (IRdrop[V] = 0.099375 + 0.036354I), increased without MCC (IRdrop[V] = 0.24924 + 0.075286I), and was the highest for AC (IRdrop[V] = 0.7609 + 0.087492I). High voltage operation and low internal resistance provided a higher maximum discharge power rating for the SWNT electrodes with MCC (210 kW kg−1) compared to without MCC (93 kW kg−1) and AC (60 kW kg−1). The ordered pore structure of SWNT electrodes results in a lower tortuosity, enabling fast ion transport and thus higher power capability. CNT electrodes operational without MCC have been demonstrated previously,22 albeit using thinner electrodes (40 μm), yet delivering lower power density (70 kW kg−1).

Operation at higher voltages is known to compromise device lifetime and durability. To address this important issue, we performed 1000 complete charge–discharge cycles at 1 A g−1 (Figure3a) from 0 V to 4 V and back to 0 V. This test is more severe than typical lifetime studies performed at half-depth discharge; still the SWNT electrodes showed excellent durability, as demonstrated by the small decline in capacitance (3.6% with or without MCC). In contrast the AC electrodes declined significantly (46%). A simple exponential fit suggests the SWNT electrodes would operate for more than 20000 cycles before reaching the same degradation level as AC electrodes reached over just 1000 cycles. While impressively long lifetime has been previously reported, for example, degradations limited to 2% over 10000 cycles,23 such reports were based on lower cell voltage ranges. Fundamentally, supercapacitors are expected to have long lifetime as energy is stored physically. However, device lifetime is still limited by parasitic chemical reactions and mechanical breakdown due to swelling of the electrode during charge–discharge, with both effects greatly amplified at higher voltages. Here the high purity of SWNTs minimized parasitic chemical reactions and the network of flexible and fiber-like SWNTs accommodated swelling while preserving mechanical integrity and electronic/ionic conduction paths of the electrode. In contrast, for the AC electrodes, the binder adhesion wears off with swelling, eventually leading to mechanical breakdown, and the presence of uncontrolled functional groups and impurities leads to parasitic reactions degrading capacitance.

thumbnail image

Figure 3. Dynamic lifetime. Over 1000 cycles of charge–discharge at 1 A g−1, the capacitance of the SWNT cell with or without MCC operated at 4 V declines by only 3.6%. In contrast the AC electrode suffers a decline of 46% when operated at 3.5 V.

Download figure to PowerPoint

To demonstrate the operational characteristics, the energy E discharged at different average powers Pav were calculated from the constant current discharge curves according to E = ∫IV(t)dt and Pav = IV/2. At any discharge rate (power) the SWNT electrodes discharge more energy (>50 Wh kg−1 at the power limit of AC, i.e., 40 kW kg−1) compared with the AC electrode, meaning that the SWNT electrodes were superior in terms of both energy and power (Figure4a). Even without MCC, the SWNTs’ performance is better than AC. The performance of the SWNT electrodes appears to be primarily limited only by the organic electrolyte used. The relationship between charge Q, capacitance C, and voltage V described by the basic ideal model Q = CV suggests that the average charge concentration required (1.7 M) exceeds the bulk concentration of the electrolyte in use (1 M). Molecular dynamics simulations24 have shown that at the double-layer interface, the ionic concentrations are even significantly higher. Thus higher concentration electrolytes may enable higher capacitance, hence higher energy; they may also improve lifetime as bulk electrolyte concentration changes during charge–discharge cycles would be diminished.

thumbnail image

Figure 4. Electrode and device performance. a) Ragone plot of AC and SWNT electrode performance. b) Performance comparison of different electrodes. 4 V operation of the SWNT electrode presented in this study achieves a higher performance than most other CNT-based supercapacitors.8, 9, 22, 25–29 c) When the entire device is considered, the SWNT device without metal current-collectors is at the cusp of bridging the energy gap between commercial supercapacitors and batteries. (Data from.4 B1: Li-ion, B2: NiHD, B3: Pb-acid; S1: Panasonic 2000F, S2: Superfarad 250F, S3: Saft Gen2, S4: Saft Gen3, S5: Maxwell 2700F, S6: Panasonic 800F.)

Download figure to PowerPoint

We have compared the energy density and maximum power density of these SWNT electrodes to those of other reported CNT electrodes (Figure 4b). The combination of the energy density and maximum power density rating of the SWNT electrodes (94 Wh kg−1, 47 Wh L−1; 210 kW kg−1, 105 kW L−1) surpasses most other CNT electrodes reported. Recently low-density plasma-treated multiwalled CNTs used as electrodes with an ionic liquid electrolyte29 have shown high performance at 4 V; however, the charge/discharge cycles were limited to the range from 2 V to 4 V and no lifetime performance has been reported yet.

Finally, we evaluate the performance of the SWNT electrodes as supercapacitor devices and compare them with commercial batteries and supercapacitors (Figure 4c) to evaluate their potential. The comparison given here should be considered provisional as packaged devices were not constructed. Even for packaged devices the total device weight is dependent on the cell-size and not readily scalable. Considering the total device weight (electrodes, separator, current-collector, and estimated packaging), for the conventional AC cell and SWNT cell (with MCC) the electrodes’ dry weight accounts for only 12% and 10%, respectively, of the total device weight. These percentages are typical for commercial devices. For the SWNT cell without MCC, the electrodes account for 26% of the total device weight, thus achieving a much higher device performance with an energy density of 17 Wh kg−1 with a maximum power density rating of 24 kW kg−1. This energy density would still be 10× lower than a typical cell-phone Li-ion battery, however, the power output would be 100× higher. The SWNT device without MCC might find applications where lightweight and compact energy storage devices are required, such as portable electronic gadgets. For large-scale applications such as backup power, the critical issues are energy efficiency and absolute power rather than weight or volume. Here SWNT electrodes with MCC would be the preferred choice because of their lower internal resistance hence higher energy efficiency and power.

Although we have scientifically shown a practical reliable performance of SWNT electrodes far exceeding that of AC electrodes, for commercialization cost is a major consideration. We believe recent advances made in SWNT mass production18 will reduce the cost by a factor of several hundred, thus eventually reaching a similar cost to AC.

In conclusion, we have fabricated a supercapacitor electrode from SWNTs possessing all the critical properties that impact device performance as summarized in Table1. Specifically, durable operation with a range of 4 V was made possible by the monolithic chemical composition of the SWNT electrode, that is, the absence of surface functional groups, conducting agents, and binders. In addition, the fiber-like structure of SWNTs contributed to the lifetime by preserving the mechanical integrity of the electrodes. High energy and power performance was achieved by operating at the higher voltage range of 4 V. The combination of high surface area and electrochemical doping enabled a high specific capacitance of 160 F g−1, further enhancing energy storage. The low tortuosity of the SWNT electrode, which enhances ion transport, contributed to high power capability. Finally, the electronic conductivity of SWNTs enabled dual functionality as both electrode and current-collector. These results reveal the full potential of SWNTs as ideal supercapacitor electrodes.

Table 1.. Attributes of SWNT and AC electrodes and their impact on supercapacitor performance.
Electrode propertySWNTActivated carbonImpact on device
StructureHollow fiberParticleLifetime
Surface area [m2 g−1]12501000–2000Energy
CompositionCarbon onlyCarbon and functional groupsVoltage, lifetime
Conductivity [S cm−1]High (21)Low (0.3)Current-collector (weight)

Experimental Section

  1. Top of page
  2. Experimental Section
  3. Acknowledgements

SWNT Synthesis: SWNTs were synthesized in a fully automatic 1 inch tube furnace by water-assisted chemical vapor deposition at 750 °C using ethylene with a parts per million (ppm) level of water on a silicon substrate with a thin film iron catalyst (1 nm) supported on an alumina layer (10 nm).15

AC Electrode Fabrication: Commercially available activated carbon powder YP17 (Kuraray Co., Ltd.) was used. Electrode sheets of YP17 with a thickness of ca. 100 μm were prepared by mixing and kneading the YP17 powder with 10 wt% PTFE binder and 10 wt% carbon black conducting agent.

Supercapacitor Assembly and Testing: Apart from the electrolyte, all active components of the supercapacitors (electrodes, separators, and current-collectors) were vacuum dried at 150 °C for at least 8 h prior to device assembly. The cells were assembled in a dry argon environment, using platinum mesh (100 mesh, Nilaco Corp.) current-collectors, a porous proprietary cellulose-based separator (40 μm thickness, density ca. 0.6 g cm−3) and 1 M Et4NBF4/propylene carbonate as electrolyte (Tomiyama Pure Chemical). For the SWNT cell without MCC, contact was made by placing platinum mesh at the top edge of the electrodes, with less than 5% of the total electrode area covered.

Electrochemical characteristics were obtained using a VMP3 galvanostat/potentiostat/frequency response analyzer (Princeton Applied Research). The IR loss and capacitance values were estimated from the galvanostatic discharge curves by fitting with a simple series RC model:

  • equation image((1))

where Vcharged is the voltage of the charged cell, the 2IRs term represents the initial voltage drop, and the capacitive term is determined by the slope of the discharge curves. The discharge-energy values were calculated by numerically integrating the discharge curves:

  • equation image((2))

The maximum power density rating was calculated4 using the Rs values determined from a linear fit to the IRdrop values (Figure 2d):

  • equation image((3))

where a represents the difference between the 4 V applied potential and the charged potential of the capacitor, b represents double the value of the internal resistance Rs, and I is the discharge current,4 giving

  • equation image

The device performance calculations were made on the basis of a device consisting of the following:

  • 1)
    Two 100 μm thick electrodes (SWNT electrode density 0.5 g cm−3, AC electrode density 0.6 g cm−3) with an area of 1 cm2.
  • 2)
    One 40 μm thick separator (density 0.6 g cm−3) with an area of 1 cm2: 2.4 mg.
  • 3)
    Electrolyte (density 1.205 g cm−3) occupying half the combined electrodes and separator volume (50% porosity): 9.64 mg.
  • 4)
    Two Pt MCCs (100 mesh, 70 μm wire diameter), each weighing 31.95 mg (for the SWNT cell without MCC, 5% of this weight is used).
  • 5)
    Estimated packaging using polyimide (also known as Kapton) film (density of 1.42 g cm−3), with thickness 40 μm and total area of 2.2 cm2: 12.5 mg.

The resulting device weights are 100.44 mg for the AC cell, 98.44 mg for the SWNT cell, and 37.735 mg for the SWNT cell without MCC.


  1. Top of page
  2. Experimental Section
  3. Acknowledgements

A.I. acknowledges financial support from the Ministry of Education, Culture, Sports, Science and Technology (Monbukagakusho), Government of Japan. Partial support by the Core Research for Evolution Science and Technology (CREST) program of the Japan Science and Technology Agency (JST) is acknowledged.