Letter pubs.acs.org/NanoLett
Electrochemical Charge-Transfer Resistance in Carbon Nanotube Composites Brad L. Corso, Israel Perez, Tatyana Sheps, Patrick C. Sims, O. Tolga Gül, and Philip G. Collins* Department of Physics and Astronomy, University of California at Irvine, Irvine, California 92697, United States ABSTRACT: Using a model system of single, isolated carbon nanotubes loaded with high-capacitance metal-oxide films, we have quantitatively investigated electrochemical composites on the single-nanotube scale. Electrochemical charging and discharging of a model MnO2 storage material was used to probe interfacial charge transfer and surface impedances at the nanotube interface. We found that one single-walled carbon nanotube has an apparent surface resistivity of 30 mΩ cm2, approximately 4 times smaller than for a multiwalled carbon nanotube and 50 times smaller than the 1.5 Ω cm2 resistivity of Pt or graphite films. The improvement originates in the electrochemical-transport properties of microelectrodes shrunk to a nanotube’s dimensions rather than any unique nanotube property like curvature, bandstructure, or surface chemistry. In explaining the enhanced performance of certain nanotube-containing composites, the results overturn widely held assumptions about nanotubes’ roles while also providing guidelines for optimizing effective composites. KEYWORDS: Carbon nanotube, charge transfer, supercapacitor, heterogeneous composite
H
eterogeneous composites of nanostructured materials provide opportunities to design bulk materials with superior optical, electronic, or chemical properties. Nanocarbon materials like carbon nanotubes (CNTs) and graphene are being extensively investigated as electrically conductive additives to insulating and semiconducting materials. Antistatic plastics and Li-ion batteries are two examples of widespread technologies that have long used CNTs as conductive fillers,1 and the research literature includes a wide range of prototype heterogeneous composites for electrochemical and energystorage applications like solar cells,2−5 supercapacitors,6−15 and catalyst supports.16−18 Electrochemical composites seem to gain special benefits from the addition of nanocarbons like CNTs and graphene, but the responsible mechanisms remain unclear. With certainty, CNTs and graphene have beneficial properties like low mass density, high strength, and high conductivity,1 but these materials also suffer from low out-of-plane electrochemical charge-transfer rates.19−22 Electrical and electrochemical investigations both indicate that heterogeneous transport from graphitic carbon to a surrounding electrolyte or chargestorage material falls by at least 4 orders of magnitude19−22 as the carbon surface becomes more highly ordered, pristine, and defect free. Because this type of charge transfer is critical for effective composites, the enhanced power densities achieved using nanocarbon additives presents a paradox. Empirically, intentional surface functionalization improves charge-transfer rates by making electrochemically inert carbons more active, and an abundance of publications highlight the added benefits of damaging CNTs and graphene chemically and mechanically, although often at the cost of requiring greater carbon mass fractions.12,23−26 © 2014 American Chemical Society
Quantitative research into this trade-off is rare, leaving the possible limitations of carbon’s low electrochemical-transfer rates widely acknowledged but poorly characterized, especially in composites research. Highly heterogeneous composites are difficult to characterize precisely, and the imperfections believed to dominate charge-transfer rates cannot be quantified. Step edges, grain boundaries, and oxygenating functional groups are ubiquitous, even on high-quality graphite (HOPG) and graphene, making it difficult to distinguish the intrinsic properties of the pristine basal plane from the rest of the carbon in a composite.20,27 Individual single-walled carbon nanotubes (SWNTs) provide a unique material solution to this problem of heterogeneity because they are edge-free and, in some instances, perfectly defect-free. By fabricating SWNT transistor devices and using electrical characterization to select SWNTs that were defectfree, we obtained unique carbon current collectors in which the contributions of edges, ends, and defects could be definitively ruled out. Here, we use this model carbon system to investigate the fundamental impedance of carbon−metal oxide interfaces for electrochemical-storage composites.9,10,28 Specifically, we selected single, defect-free SWNTs, coated those devices with porous LixMnO2 as a generic charge-storage cathode material,15,29,30 and then probed the electrochemical pseudocapacitive performance of the composite structures. We reproduced the LixMnO2 capacitances of larger-area films and bulk composites Received: November 22, 2013 Revised: February 3, 2014 Published: February 14, 2014 1329
dx.doi.org/10.1021/nl404349g | Nano Lett. 2014, 14, 1329−1336
Nano Letters
Letter
(even though measurements on extremely small material quantities often inflate estimates of specific capacitance31) and obtained sufficiently concentrated current densities to probe the series resistances of the SWNT interfaces directly. The small surface area of an individual SWNT produced a high resistance compared to larger-area electrodes, but the areanormalized resistivity proved to be anomalously small, with no apparent difference between defect-free SWNTs and more disordered ones. We uncover the cause of this anomaly and conclude with a discussion of how it enhances the properties of heterogeneous SWNT composites.
■
EXPERIMENTAL METHODS SWNT current collectors were grown in place on a SiO2 surface using catalyst-assisted chemical vapor deposition (CVD). The CVD catalyst consisted of FexMoy nanoparticles32,33 diluted in ethanol (108 Ω, indicating that the conductors’ longitudinal resistance was immaterial. Depleted semiconducting SWNTs can have resistances much higher than 106 Ω, but the gating conditions to deplete a SWNT were far outside the electrochemical window used here for MnO2 charge and discharge. Therefore, the semiconducting SWNTs in these experiments were always doped to a conductive state, and we observed no effects of the back gate upon Ri or CS.
increases. For instance, the ionic impedance, Rliq, of the electrolyte and the electronic impedances, RMnO2, of the metal oxide might have rate-independent components. However, these terms should not be affected by varying the current collector, and we did not observe any effects from changing the electrolyte concentration. Furthermore, the collector-independent component, z/(mvα), in Table 1 reproduced the accepted mass- and rate-dependent impedances for bulk nanostructured MnO2.55 When multiplied by the total capacitance, mCS, ZW produced a mass-independent time constant τ that was characteristic of MnO2 charging, with a value of τ = 0.3 s at v = 100 mV/s. After considering all of the resistive elements depicted in Figure 1c, we concluded that careful separation of Ri from the MnO2-dependent parameters ZW, Cdl, and Co enabled a quantitative analysis of differences between collector types and that the entire increase ΔRi observed for MWNTs and SWNTs over Pt collectors could be attributed to the collector interface. Further analysis showed that the entire increase in Ri was a consequence of different collector areas, although not in the conventional sense. Conventional planar electrodes have a material-specific, intrinsic resistivity, RiA, but Table 1 shows that RiA produces apparent resistivities of only 30 mΩ cm2 for a SWNT, approximately one-third of the value of MWNTs and almost 50 times smaller than for Pt or HOPG films. Although some literature has attributed unique electrochemical properties to SWNT surfaces, we rejected the RiA variation among sp2 carbons as unphysical. Instead, we lithographically fabricated additional Pt electrodes with areas ranging from 0.1 to 10 μm2 to confirm that Ri was proportional to A−1/2 instead of A−1. Table 1 demonstrates that all three types of collectors shared the same impedance Ri = (90 ± 10 Ω m)A−1/2 when normalized in this manner. The A−1/2 factor resembles the current-limiting behavior of ultramicroelectrodes in the cyclic voltammetry of redox-active solutions.56 When electrode dimensions are smaller than the length scale of concentration gradients that an electrolyte can support, diffusional limitations of mass transport constrain currents to a maximum value that is proportional to A1/2 rather than the microelectrode area, A. The effect has been extensively studied for well-defined, Pt electrode−electrolyte interfaces57,58 and SWNT−electrolyte systems,43,45,59−61 but it has not previously been considered a factor in heterogeneous composites. The MnO2-coated current collectors in this work are certainly within the typical size range for ultramicroelectrodes. The A−1/2 factor was unexpected, however, for a purely electronic current collector. The ionic impedance associated with MnO2 charging and discharging is contained within ZW and is proportional to the active surface area of the entire
■
DISCUSSION After accounting for the limitations described above, Table 1 may be fairly generalized as showing no dependence of ZW on the current collector, either in the charging or the discharging directions. Ri was the only parameter with systematic variations, increasing by 0.4 and 1.3 GΩ for MWNT and SWNT collectors, respectively, compared to Pt devices. In principle, various impedance components might contribute to these Ri 1333
dx.doi.org/10.1021/nl404349g | Nano Lett. 2014, 14, 1329−1336
Nano Letters
Letter
with their much larger, interfacial charge-transfer resistance. It is the latter charge-transfer resistance that benefits from SWNTs’ small dimensions. Throughout the SWNT-composite literature, discussion of photovoltaic efficiency2−5 and charge-transfer impedance6−18 uniformly use estimates to the total SWNT surface area instead of A−1/2, leading to apparent improvements shown as RiA in Table 1. Finally, we conclude by addressing the time constant, τ, for charging and discharging MnO2 films through SWNT collectors. As described above, the ZW term of the impedance produced a mass-independent component. τ1, that was typical of double-layer electrochemical storage, where capacitance is generally proportional to mass (if the material is sufficiently porous) and resistance is inversely proportional to it. The additional impedance, Ri, increases the time constant by an additional term, τ2 = mCSRi, that is proportional to mA−1/2. Instead of being independent of mass, τ2 grows in proportion to film thickness and collector dimensions, with equal sensitivity to both. Essentially, loading more mass onto a fixed-area collector does not change the surface current density of the material’s fast-storage mechanisms (τ1), but it does drive larger currents across the impedance of the collector interface (τ2). The latter mechanism becomes dominant as the collector area shrinks. Table 1 includes calculated values of both components at typical experimental conditions to illustrate their relative values. For the large-area Pt collectors, τ1 > τ2, but the small area of SWNT collectors leads to much larger τ2 values. For devices to make effective use of any metal oxide’s charging kinetics, the current collector should add negligible resistance to the system. This design criteria can be expressed as τ1 ≫ τ2 for the parameters described here. SWNTs clearly do not meet this criteria in these experiments, which instead maximized τ2 for the sake of characterization. Armed with welldefined Ri and ZW, we may now express quantitative design rules for more appropriate mass loadings in heterogeneous composites. For the slow kinetics of MnO2 charging, mass loadings of 1.0 pg per μm of SWNT achieved τ1 ≈ τ2. Smaller mass loadings of 0.1 pg/μm would result in a more acceptable τ1 > 10τ2. To achieve the same ratio during discharge, which is intrinsically much faster in MnO2, the mass loading must decrease by another factor of 5 to 10. In practice, this means that a SWNT can support a cylinder of MnO2 up to 50 nm in radius without compromising either charge or discharge kinetics. Annealed MnO255 or alternative oxides with even faster rates (i.e., smaller ZW components) would require proportionally smaller mass loadings. At very small loadings, the diffusional concentration gradients in the material also overlap significantly, causing the advantageous microelectrode feature of SWNTs to be lost.
porous deposit. The collector, by comparison, has a much smaller area that is treated throughout the literature as an electronic interface. Our main conclusion is that ionic diffusion ultimately determines the electronic resistance of nanotube interfaces, even when coated by hundreds of nanometers of metal oxide. This loading demands a relatively high electronic current at the collector that remains coupled to the limitations of ionic transport regardless of whether the MnO2 porosity allows the electrolyte to approach the collector interface. This conclusion challenges assumptions used throughout the MnO2 and SWNT-composite research fields, and it demonstrates how understanding the coupled transport of electronic and ionic carriers remains a grand challenge in electrochemical energy storage.28,51 Other than being determined by A−1/2, Ri did not depend on the collector material, and we observed no reproducible variations between different devices. We did not distinguish any reproducible differences between semiconducting and metallic SWNTs suggested by others43,59 or any differences among MWNTs with different diameters. In the extremely confined geometry surrounding a SWNT, we suspected that polarization effects and enhanced electric fields might improve charge transfer over planar Pt films, but no such improvements were found. Carefully selected, defect-free SWNTs behaved the same as MWNTs, which generally have much more surface disorder. Ultimately, comparisons among SWNTs with different degrees of disorder proved that sidewall defects had no measurable effect on Ri, opposite to what might be expected from the HOPG literature.19−22 On average, Ri values may have decreased slightly for charging (70 ± 20 Ω m) compared to discharging (90 ± 10 Ω m), but the effect was within one standard deviation, indicating that Ri was not particularly sensitive to the charge state of the interfacial MnO2. All of these attributes are common to outer-sphere redox chemistry,56 reinforcing the conclusion that the interfacial resistance of nanotubes is not a purely electronic parameter. SWNTs and MWNTs do have a unique attribute, which is their ability to maintain an exquisitely small surface area while forming an effective conductive network. There is no practical way to interconnect arrays of Pt microelectrodes in a heterogeneous composite material and take advantage of their unusual resistivity scaling. CNTs, however, can be randomly dispersed into an active material in a concentration that not only creates a percolative network but also maintains a small total area. When optimally dispersed, the resistance of a SWNT-network current collector still falls within the microelectrode regime, increasing proportionally to A−1/2 instead of A and providing an opportunity to design and obtain smaller electrochemical interface resistances. The net effect is represented by the resistivities in Table 1, for which a conventional estimate using RiA suggested SWNTs to have a remarkably low effective surface resistivity for their area. We propose that this effect may explain the improvements reported for nanotube-containing energy-storage composites, improvements that include low internal dissipation, enhanced charge/discharge rates, and the avoidance of capacitance loss that normally accompanies high rates. Any of these factors can be a primary power-limiting factor for energy-storage applications, especially when material properties like ZW would otherwise allow for high power (e.g., during MnO2 discharge). The literature often credits the improvements of SWNTs to their low longitudinal resistance, but it seems certain that longitudinal resistance is a minimal component in series
■
SUMMARY In electrochemical storage, electronic transport is almost always secondary to the limitations of ionic transport. By using a cylindrical, nanocollector geometry of MnO2 on a single SWNT, we reversed this situation to investigate quantitatively the electronic impedances at a current collector during electrochemical cycling. The ideal geometry allowed us to concentrate a known electrical current onto well-characterized SWNT surfaces. Precise knowledge of a SWNT’s surface area and chemical state uniquely enabled a quantitative measure of charge-transfer resistivity. The small surface area of SWNTs resulted in a high impedance that was shown to be a primary power-limiting 1334
dx.doi.org/10.1021/nl404349g | Nano Lett. 2014, 14, 1329−1336
Nano Letters
■
mechanism for MnO2 charge and discharge. Poor power performance of thick films is generally blamed on ionic tortuosity or the low electrical conductivity of bulk MnO2, but here we have shown that the collector impedance remained a primary factor up to oxide thicknesses of 500 nm. To reduce the effect of the collector to the point that its impedance is insignificant requires lower mass-loading ratios, such as film thicknesses
