Graphene Double-Layer Capacitor with ac Line-Filtering Performance

See allHide authors and affiliations

Science  24 Sep 2010:
Vol. 329, Issue 5999, pp. 1637-1639
DOI: 10.1126/science.1194372


Electric double-layer capacitors (DLCs) can have high storage capacity, but their porous electrodes cause them to perform like resistors in filter circuits that remove ripple from rectified direct current. We have demonstrated efficient filtering of 120-hertz current with DLCs with electrodes made from vertically oriented graphene nanosheets grown directly on metal current collectors. This design minimized electronic and ionic resistances and produced capacitors with RC time constants of less than 200 microseconds, in contrast with ~1 second for typical DLCs. Graphene nanosheets have a preponderance of exposed edge planes that greatly increases charge storage as compared with that of designs that rely on basal plane surfaces. Capacitors constructed with these electrodes could be smaller than the low-voltage aluminum electrolyte capacitors that are typically used in electronic devices.

Electric double-layer capacitors (DLCs), also called supercapacitors or ultracapacitors, store charge in the double layer formed at an electrolyte-electrode interface when voltage is applied. The electrodes are generally composed of high-surface-area conductive material, usually activated carbon (1). DLCs typically store more than an order of magnitude more energy per unit volume than conventional capacitors (2) but, like batteries, are low-voltage devices, so they often must be connected in series to meet system voltage requirements. Electric double-layer charge storage was first observed more than 100 years ago, but DLC products did not reach the market until 1978 (3). The introduction of volatile computer memory created the need for a power source that could be charged repeatedly and then reliably deliver low levels of dc power over a long time. Kilofarad-sized capacitors became available starting in the mid-1990s to “load-level” the power profile of battery-powered electric vehicles that were then under development (4) and, more recently, to capture and store energy derived from kinetic energy harvesting—for instance, the braking energy of a hybrid vehicle (5).

The delayed introduction of DLCs was caused by a limited market for capacitors that could only store charge but performed poorly at their other main task: filtering voltage ripple (6). The typical resistor-capacitor (RC) time constant for a DLC is ~1 s—far too long to be useful for the common application of 120-Hz filtering (8.3 ms period), which entails smoothing the leftover ac ripple on dc voltage busses found in most line-powered electronics. Sixty-Hz ac power is full-wave rectified and then filtered to create pure dc voltage. Filtering today is performed primarily by means of aluminum electrolytic capacitors, which usually are among the largest components found in any electronic circuit: Smaller-sized filtering capacitors may allow system size reductions, which is particularly valuable in some portable electronics applications.

Present DLCs have an impedance phase-angle at 120 Hz that is near 0°, which is far from the –90° value needed for filtering. This drawback is a direct result of using porous electrodes, which store charge in a distributed fashion. When driven by 120-Hz ac, they respond like a transmission line (7); that is, they behave like a resistor rather than a capacitor. Thus, high-surface-area materials with less inherent porosity have been examined. DLCs fabricated from felt-like electrodes comprising entangled, multiwall carbon nanotubes (CNTs) did establish a frequency response record in 1997: 6 Hz for an impedance phase angle of –45° (8). (Resistance and reactance have equal magnitudes at a phase angle of –45°, making this frequency convenient for comparison purposes.) These DLCs had thin electrodes and used a high-conductivity aqueous electrolyte but were still incapable of filtering at 120 Hz for two reasons. First, the felt-like structure created porosity with a pore length equal to the electrode thickness, thus creating distributed charge storage with associated ionic resistance. Second, the structure relied on manifold electrical contacts among the individual CNTs and against the current collector, contributing to the electronic resistance. Although these capacitors were unable to filter, they showed that DLC frequency response could be improved by using electrode materials with external rather than internal surface area.

Further improvements in frequency response were reported from electrodes of multiwall CNTs that were deposited onto a metal current collector, cleared of oxygen surface groups, and bonded to the collector with a hydrogen furnace treatment (911). These capacitors reached a –45° phase angle at 636 Hz, which is more than 100 times the 1997 record. The phase angle at 120 Hz was –65°, and thus they could not efficiently filter at this frequency. Multi-wall, vertically aligned CNTs grown directly on a metal current collector were also investigated as DLC electrodes. One such study showed impedance data with a –45° phase angle at 443 Hz (12). Yet another study bonded vertically aligned CNT electrodes to an aluminum current collector, but this also was unable to efficiently filter 120-Hz ripple (13).

Agglomerated, chemically modified graphene material was evaluated as a DLC electrode and showed a considerable degree of unwanted porous-electrode behavior (14). Apparently, direct access to the charge-storage surfaces was severely restricted by the structure, adding substantial ionic resistance. Furthermore, the bonded electrode relied on multiple electrical contacts among the graphene agglomerates, similar to the contacts present in all particulate electrodes, thus contributing additional electronic resistance. Capacitors constructed with electrodes of agglomerated graphene electrode as reported were not capable of 120-Hz filtering.

Electrodes with vertically oriented graphene nanosheets coating carbon fibers have also been examined (15). Capacitor designs were developed for a 14,900 F device, but no two-terminal electrical responses were reported. Such vertically oriented graphene nanosheets are believed to offer a near-ideal structure for DLC electrodes capable of high-frequency operation. First, they have a preponderance of edge planes that provide capacitance of 50 to 70 μF/cm2 in comparison with that of basal planes, which provide capacitance of only ~3 μF/cm2 (16). Second, charge-storage edge planes are exposed and directly accessible, minimizing the distributed nature of the charge storage. Third, porosity effects are minimal because of the open structure, reducing ionic resistances. And fourth, graphene nanosheets themselves have extremely high conductivity and can be grown from a conductive surface, minimizing electronic resistances. These factors, we expect, should provide high levels of charge storage that is accessible through minimum series resistance and thus allow creation of a DLC capable of high-frequency operation.

We synthesized vertically oriented graphene nanosheets directly on heated nickel (Ni) substrates using radio frequency (RF) plasma–enhanced chemical vapor deposition (17, 18). After pumpdown, the substrates were plasma-etched for 10 min in 40% argon (Ar) + 60% H2 (total pressure of 50 mT). After the Ar was shut off, the Ni substrate was heated to 1000°C in H2, methane (CH4) was added, and a 1000-W plasma was generated by using 40% CH4 and 60% H2 at a total pressure of 85 mT. Initially, graphitic islands formed on the Ni substrate (Volmer-Weber planar growth), with two-dimensional growth continuing until impingement with other islands. This basal layer was approximately 10 to 15 nm thick. After 20 min of growth, the resulting coating was composed of vertically oriented graphitic nanosheets approximately 600 nm high with a cross section <1 nm thick, but often terminating with an edge of a graphene sheet (one atomic layer).

A plan view of vertically oriented graphene nanosheet electrode (Fig. 1A) shows the irregular morphology of the surface (exposed edge planes and the random expansive open areas) created by defects that arise from stress and hydrogen incorporation. The structures grow vertically from the metal substrate and are electronically connected to it. The measured surface area was ~1100 m2/g. Similar graphene growth is shown in Fig. 1B, but on a circular fiber so as to display both plan and shallow-angle views of the surface. Raman intensity measurements of the coating had a D/G band intensity ratio (ID/IG) of 0.67 because of defects.

Fig. 1

(A) Plan SEM micrograph of coated Ni electrode. (B) SEM micrograph of a coated fiber, showing plan and shallow-angle views.

Capacitors were fabricated to measure the electrical performance of the vertically oriented graphene nanosheet electrodes. Two identical 2.5-cm-diameter, 75-μm-thick Ni disks with graphene nanosheet coatings grown over their 1.6-cm-diameter central region (2.0 cm2 area) were separated by a 25-μm-thick microporous separator. The coatings and the separator were wetted with the aqueous electrolyte [25% potassium hydroxide (KOH)] before sealing the perimeter of the disks with a thermoplastic by use of an impulse-heat-seal apparatus. These packaged prototypes, 2.5 cm diameter by ~175 μm thick, had a mass of ~0.8 g. The active material-coating thickness on each electrode was approximately 0.6 μm, which is negligible compared with device dimensions and mass. Electrical connection was made to the back surface of each Ni disk.

Impedance phase angle data of a prototype capacitor fabricated with the vertically oriented graphene nanosheet electrodes is shown in Fig. 2. Data from a commercial DLC and from an aluminum electrolytic capacitor are included for comparison. All sets of data show capacitive behavior (near –90° phase angle) at low frequency and inductive behavior at high frequency. However, the transition between these two limits occurs over a single decade in frequency for the graphene nanosheet capacitor, whereas it occurs over approximately seven decades in frequency for the activated carbon capacitor. This disparity is due almost entirely to electrode porosity differences. The impedance phase angle of the graphene nanosheet capacitor reached –45° at ~15,000 Hz in comparison with 0.15 Hz for the activated carbon capacitor and ~30,000 Hz for the electrolytic capacitor. At 120 Hz, the impedance phase angle of the graphene nanosheet capacitor was approximately –82° as compared with ~0° for the activated carbon capacitor and approximately –83° for the aluminum electrolytic capacitor. The phase angle for a blank (bare Ni electrode prototype) was –85°.

Fig. 2

Impedance phase angle versus frequency for the graphene nanosheet DLC. Measurements from a commercial DLC having an activated carbon electrode and an aluminum electrolytic capacitor are shown for comparison.

A complex plane plot of the impedance data obtained from the vertically oriented graphene nanosheet capacitor is shown in Fig. 3, with an expanded view in the inset. There is no evidence of porous electrode behavior, which would manifest itself by a line that intersects the real axis at a near-45° angle. There are also no features associated with a series-passive layer (high-frequency semicircle), which would add series resistance. Data fit a near-vertical line as produced by a series-RC circuit. Raistrick modeled the impedance of rough, saw-tooth-surface electrodes having various height-to-separation ratios (19). Fig. 3 follows such rough-surface behavior with an aspect ratio of <2, which is consistent with the observed scanning electron microscopy (SEM) micrograph structure.

Fig. 3

Complex plane plot of the impedance of the graphene nanosheet capacitor showing vertical intersection with the real axis and not the usual porous electrode behavior shown by DLCs.

We used a series-RC circuit model in which resistance is the real part of the impedance and capacitance is calculated as C = –1/(2πfZ″), where f is frequency in Hz and Z″ is the imaginary part of the impedance. Capacitance versus frequency from this equation is plotted in Fig. 4 for the graphene nanosheet capacitor shown in the previous figures. At 120 Hz, the derived capacitance value is 175 μF, and the measured resistance is 1.1 ohms, yielding an RC time constant of less than 200 μs. Divergent behavior near 20 kHz is an artifact of the model (where Z″ passes through zero) and should be ignored. As shown, the capacitance value increases from ~175 μF in the 10- to 104-Hz band up to ~375 μF at frequencies below ~1 Hz. The increase occurs in the region of the phase angle dip shown in Fig. 3. Capacitance saturation (horizontal behavior) was not observed even at 0.01 Hz, although this device exhibited incredibly fast response, suggesting the involvement of a second, much-lower-rate charge-storage process. Three-electrode measurements revealed that the positive electrode had greater capacitance increase at low frequency than did the negative electrode. Capacitors of the same design fabricated by using bare Ni electrodes had capacitance values of less than 25 μF at 120 Hz. We attribute the marked capacitance increase at low frequency to pseudocapacitance derived from ion intercalation into the exposed edge planes of the graphene structure. Pseudocapacitance by this mechanism has been reported for exfoliated carbon fibers with sulfuric acid electrolyte (20).

Fig. 4

Capacitance versus frequency of the graphene nanosheet DLC, assuming a series-RC circuit model. Capacitive behavior is shown up to ~104 Hz.

A similar vertically oriented graphene nanosheet capacitor fabricated with an organic electrolyte (1 M tetraethylammonium-tetrafluoroborate salt in propylene carbonate) showed more definite but still incomplete saturation of capacitance at low frequency (fig. S1). The organic electrolyte capacitance was ~50% higher than the value from aqueous electrolyte prototypes, which is opposite to typical activated carbon behavior. This boost in capacitance using the organic electrolyte may result from more complete wetting of the graphene nanosheet surface with this electrolyte.

Capacitance at 120 Hz was ~175 μF for the graphene nanosheet capacitor (with KOH electrolyte), which corresponds to a capacitance density of the 0.6-μm-thick active layer of ~3 F/cm3. The graphene active layer stored ~1.5 FV/cm3 with the aqueous electrolyte (0.5 V) and ~5.5 FV/cm3 with the organic electrolyte (1.25 V). Aluminum electrolytic capacitor foil is highly etched before being anodized so as to create charge storage throughout its volume. Low-voltage aluminum anode foil (KDK, Tokyo, Japan) has CV/volume values up to ~0.14 FV/cm3. Thus, a graphene DLC in principal could have substantially smaller volume than a comparably rated low-voltage aluminum electrolytic capacitor.

DLC designs would differ from the spiral-wound construction commonly used with aluminum electrolytic capacitors because of operating voltage differences. A bipolar design (cell stacking), as presently used by several DLC manufacturers, offers volumetric efficiency and may be optimal. With stacked cells, DLC capacitance density scales like the inverse of the square of the number of series-connected cells: ~1/V2, where V is the device operating voltage. Capacitance density of an electrolytic capacitor scales like 1/V because the CV product of anodic dielectric is approximately constant. Thus, DLC capacitance density advantages will disappear at a particular voltage. Using prototype performance measurements with conventional construction materials in a practical design (fig. S2), a single-cell graphene DLC (~2.5 V operation) should offer a sixfold or greater volume advantage over an aluminum electrolytic capacitor of the same rating. A two-cell graphene DLC (~5 V operation) should offer a twofold or greater volume advantage over an equivalently rated aluminum electrolytic capacitor. No volumetric advantages are expected at higher voltage. Aluminum electrolytic capacitors rated at 2 V or higher are widely used today for bypass and filtering in portable electronics equipment.

Graphene nanosheet electrodes could be manufactured by using standard semiconductor process equipment. The fabrication of the electrode and the choice of electrolyte have not been optimized, and increases in capacitance density through further optimization appear likely. Cell operating voltage may be increased with ionic liquid electrolytes or by use of an asymmetric design, with either approach expanding the voltage region in which the graphene DLC capacitance density exceeds that of present aluminum electrolytic capacitor technology.

Supporting Online Material

Figs. S1 and S2

References and Notes

  1. This work was supported in part by DARPA Defense Sciences Office, 3701 N. Fairfax Dr., Arlington, VA 22203.
View Abstract

Stay Connected to Science

Navigate This Article