pubs.acs.org/NanoLett
Electrochemical Properties and Myocyte Interaction of Carbon Nanotube Microelectrodes Andrew O. Fung,*,† Christos Tsiokos,† Omeed Paydar,† Li Han Chen,# Sungho Jin,# Yibin Wang,§ and Jack W. Judy†,| †
Biomedical Engineering Interdepartmental Program, University of California, Los Angeles, Los Angeles, California 90095, United States, # Department of Mechanical Engineering, University of California, San Diego, San Diego, California 92093, United States, § Departments of Anesthesiology, Physiology, Medicine Division of Cardiology, David Geffen School of Medicine at University of California, Los Angeles, Los Angeles, California 90095, United States, and | Department of Electrical Engineering, University of California, Los Angeles, Los Angeles, California 90095, United States ABSTRACT Arrays of carbon nanotube (CNT) microelectrodes (nominal geometric surface areas 20-200 µm2) were fabricated by photolithography with chemical vapor deposition of randomly oriented CNTs. Raman spectroscopy showed strong peak intensities in both G and D bands (G/D ) 0.86), indicative of significant disorder in the graphitic layers of the randomly oriented CNTs. The impedance spectra of gold and CNT microelectrodes were compared using equivalent circuit models. Compared to planar gold surfaces, pristine nanotubes lowered the overall electrode impedance at 1 kHz by 75%, while nanotubes treated in O2 plasma reduced the impedance by 95%. Cyclic voltammetry in potassium ferricyanide showed potential peak separations of 133 and 198 mV for gold and carbon nanotube electrodes, respectively. The interaction of cultured cardiac myocytes with randomly oriented and vertically aligned CNTs was investigated by the sectioning of myocytes using focused-ion-beam milling. Vertically aligned nanotubes deposited by plasma-enhanced chemical vapor deposition (PECVD) were observed to penetrate the membrane of neonatal-rat ventricular myocytes, while randomly oriented CNTs remained external to the cells. These results demonstrated that CNT electrodes can be leveraged to reduce impedance and enhance biological interfaces for microelectrodes of subcellular size. KEYWORDS Microelectrode, carbon nanotubes, Raman spectroscopy, impedance spectroscopy, cyclic voltammetry, myocyte
N
eural and cardiac electrode arrays are used in a variety of applications for electrophysiological recording and stimulation. Although many applications call for electrode arrays to record signals from the tissue level to the multicell level, others require multisite extracellular recording at the single-cell and subcellular scales. For example, the elucidation of arrhythmia in cardiac myocytes highlights the significance of correlating electrophysiology with genomic events. Ruan et al. verified a link between an inducible gene and local defects in the intracellular cycling of calcium ions,1 which are critical to muscular contraction. However, in order to correlate calcium defects with overall abnormality of action potentials, the surface electrophysiological heterogeneity must measured at multiple sites on the cell. For typical cardiac myocytes, which have a width of 20 µm and a length of 120 µm, this degree of spatial resolution would correspond to electrodes with a surface area less than 300 µm2. The mechanics of glass micropipet electrodes preclude simultaneous probing at multiple locations on a single myocyte. Microfabricated multielectrode arrays (MEAs) can
achieve the necessary spatial density, but miniaturization faces technical hurdles. Microelectrode arrays are commonly based on thin-film metal, such as gold, platinum, and iridium, because of their biocompatibility and chemical inertness. At subcellular dimensions, however, the topology and biochemistry of planar surfaces do not approximate a biomimetic environment amenable to a high-integrity cell-electrode interface. The efficiency of charge transfer during stimulation and recording decreases as the gap between the electrode and cell widens.2 Another obstacle to miniaturization arises in that electrode impedance scales inversely with surface area. This scaling effect is evident in a survey of the literature (Figure 1). Higher impedance increases Johnson (i.e., thermal) noise in recording and decreases the capacity for current injection in stimulation. As a result, this trade-off limits the minimum size of electrode that will yield acceptable signal quality. Researchers have sought to mitigate this trade-off by coating electrode sites with additional materials that reduce impedance. Coatings of platinum-black, activated iridium oxide films (AIROF), and carbon nanotubes (CNT) have all been reported to achieve lower area-specific impedances than bare metals by contributing a significant nanoscale topography to electrode surfaces. However, nanotubes pos-
* To whom correspondence should be addressed,
[email protected]. Received for review: 04/20/2010 Published on Web: 10/18/2010 © 2010 American Chemical Society
4321
DOI: 10.1021/nl1013986 | Nano Lett. 2010, 10, 4321–4327
FIGURE 1. Survey of microelectrode impedance at 1 kHz vs geometric surface area. Reports were surveyed for measurements of electrode impedance in saline. The dashed box encloses data points for this work. Materials include gold,3-22 platinum,3,10,14,19,23-26 platinum-black,3,10,14,23,26 iridium27,28 (including activated iridium oxide films (AIROFs)29,30), and carbon nanotubes (CNTs).23,31-40
sess a unique combination of properties that can be leveraged for higher-performance biopotential electrodes: (1) enhanced electron transfer (ET),41 (2) high surface-area-tovolume ratio, (3) ease of biomolecular modification,42 and (4) three-dimensional (3-D) structure that enhnaces cellular interfaces.43 Despite these promising advantages of carbon nanotube electrodes (CNTEs), it is clear from Figure 1 that their application at single- and subcellular scales remains relatively unexplored. While studies have revealed that some neurons show an extraordinary affinity for randomly oriented CNTs,43,44 it remains to be seen how cultured myocytes interact with CNT substrates. If a similar phenomenon occurs with myocytes, it could enable greatly improved signal-to-noise ratio over traditional extracellular planar metallic electrodes. In the work presented here, we report the fabrication of very small gold MEAs and the subsequent localized growth of CNTs using titanium nitride (TiN) as a diffusion barrier. We investigated the effect of electrode size and material composition on interfacial impedance, ET kinetics, and the interaction of CNTs with cultured myocytes. A detailed understanding of electrochemical properties and cellular interaction is required to realize the maximum potential of CNTs in microelectrode technology. Microfabrication of MEAs for in vitro cell cultures allows precise control of material properties and geometries of the recording sites. An array of individually addressable probes, with nominal geometric surface areas from 20 to 200 µm2, was fabricated using a typical three-layer configuration for a thin-film microelectrode (Figure 2). Synthesis of CNTs by chemical-vapor deposition (CVD) enabled selective growth of nanotubes to form 3-D electrode structures.45,46 The choice of underlaying supporting material was critical for the © 2010 American Chemical Society
FIGURE 2. Fabrication process of the microelectrode array. (a) Schematic diagram of the microelectrode layout (top view). (b) Crosssectional diagram of the fabrication process illustrating self-aligned deposition of TiN/Ti/Ni at the active site. (c) Carbon nanotubes were synthesized in situ by CVD. Bond pads were exposed in a separate oxide etch. (d) SEM micrograph showing array of 32 carbon nanotube microelectrodes.
successful deposition of CNTs. For use in an electrophysiological electrode system, all support layers had to be electrically conductive and compatible with microfabrication process conditions. It was desirable that the work functions of all supporting layers match that of the CNTs, ≈4.9 eV.47 In order to maintain the integrity of the gold traces and bond pads, we opted for lower CVD temperatures notwithstanding the possibility of CNT growth with a reduced yield, rate, and uniformity. High-purity Ni was chosen as the catalyst because it is more suitable than Fe for CNT growth at lower temperatures. Titanium was deposited as a support layer because of its superior wetting properties and relatively low electrical resistance.48 Titanium is also less prone than Cr to contribute undesirable electrochemical reactions. A barrier layer of TiN was required to prevent interdiffusion of gold with the catalyst.49,50 4322
DOI: 10.1021/nl1013986 | Nano Lett. 2010, 10, 4321-–4327
FIGURE 3. Randomly oriented carbon nanotubes synthesized by thermal chemical vapor deposition. (a) Scanning electron micrograph of the cross section of the nanotube film. (b) Raman spectrum of carbon nanotubes. G/D peak intensity ratio is 0.86.
Scanning electron microscopy (SEM) micrographs of the CNTs deposited by thermal CVD (Figure 3a) revealed a randomly oriented CNT film, approximately 2.5 µm thick. The CNTs were kinked and had diameters ranging from 25 to 60 nm. These features are typical of CNTs nucleated by Ni at relatively low process temperatures (600 °C).51-53 Efforts to deposit CNTs have been mostly limited to nonmetallic substrates54 that do not deactivate the catalyst; it is noteworthy in this work that patterned CNTs were deposited using a TiN diffusion barrier on gold traces. Raman spectra reveal much about the structure and properties of nanotubes.55 The E2g modes of crystalline graphite are Raman-active. Specifically, ordered graphite is indicated by the G peak at 1582 cm-1. Disorder in polycrystalline graphite gives rise to a D peak at 1350 cm-1. The ratio of the G/D peak intensities measures the relative responses of graphitic carbon and defective/amorphous carbon and was used here as a semiquantitative metric of the quality of the CNTs. The first-order Raman spectrum of the CNTs showed a sharp G peak at 1583 cm-1 (Figure 3b). Characteristic of multiwalled CNTs, an additional D-peak was located at 1350 cm-1. This strong peak is not present in highly ordered pyrolytic graphite (HOPG) but is qualitatively similar to the spectrum of polycrystalline graphite.56,57 The presence of a sharp G peak along with an enhanced D peak agrees with the report of Li et al.58 and was attributed to curvature and disorder in the graphitic layers.59 This interpretation was supported by a strong peak at 2700 cm-1 (i.e., 2 × 1350 cm-1) in the second-order Raman spectrum. The ratio of intensities, G/D ) 0.86, indicated a high degree of disorder in the graphene sheets, as could be expected for CNTs grown at low temperature. Nyquist plots of potentio electrochemical impedance spectroscopy (PEIS) data from the planar gold microelectrodes in phosphate-buffered saline (PBS) followed a semicircular locus (Figure 4a). Data were fit by a simple capacitive equivalent circuit, which consists of a series solution resis© 2010 American Chemical Society
tance, Rs, a double-layer capacitance, Cdl, and a chargetransfer resistance, Rct. A phase shift closer to zero was required for an accurate measurement of Rs. The double layer is discharged with a time constant, τ, given by τ ) RsCdl ) πκrC0/4, where r is the microelectrode radius, C0 is the area-specific capacitance, and κ is the conductivity of the electrolyte (74 Ω·cm for normal saline).60 At these sizes of microelectrodes, τ is less than the period of the sinusoidal signal at the highest measured frequency. Therefore, Rs was estimated by extrapolation of the impedance data at high frequencies. Since interfacial impedance can deviate from the ideal capacitance predicted by Cdl, the equivalent circuit models were adapted by substituting each capacitor by a constant-phase element (CPE),61 with impedance given by
ZCPE(ω) )
1 K(jω)α
(1)
where ω is the angular frequency, K is a measure of the magnitude of ZCPE, and R is a fractional power related to the phase angle. Both K and R are fitting parameters, and the CPE is identical to an ideal capacitor for R ) 1. When R < 1, the center of the semicircular impedance locus is translated below the real axis. Typically, 0.8 < R < 1 for biomedical electrodes due to the frequency dispersion effects of surface roughness.62 The use of CPEs improved the fit, but the gold microelectrodes generally showed near-ideal capacitance (R > 0.96). The CNTE was modeled as a metal electrode with a porous coating (Figure 4b,c), using an equivalent circuit commonly employed to study the impedance properties of painted metal.63 The resistor, RCNT, was interpreted to represent the sum of CNT-substrate contact resistance and pore resistance due to electrolyte penetration into the CNT film. The capacitor, CCNT, was interpreted as the capacitance of the CNT film. Diffusion processes within the CNT film were modeled by a CPE, Zdiff, in series with Rct. The equiva4323
DOI: 10.1021/nl1013986 | Nano Lett. 2010, 10, 4321-–4327
FIGURE 4. Nyquist plots of potentioelectrochemical impedance spectra for microelectrodes (114 µm2 geometric surface area) of (a) gold, (b) pristine carbon nanotubes, and (c) nanotubes treated with O2 plasma. Markers show measured data, and solid lines show the fitting curve of the equivalent circuit below each plot. Each inset details the respective plot at lower impedances. TABLE 1. Representative Fitting Parameters for Gold, Pristine Nanotube, and Plasma-Treated Nanotube Electrodes electrode material gold pristine CNT O2 plasma CNT
Rs (Ω) 244
