Nanogap-Based Electrochemical Measurements at Double-Carbon-Fiber Ultramicroelectrodes

Abstract

Electrochemical measurements with unprecedentedly high sensitivity, selectivity, and kinetic resolution have been enabled by a pair of electrodes separated by a nanometer-wide gap. The fabrication of nanogap electrodes, however, requires extensive nanolithography or nanoscale electrode positioning, thereby preventing the full exploration of this powerful method in electrode design and application. Herein, we report the simple fabrication of double-carbon-fiber ultramicroelectrodes (UMEs) with nanometer-wide gaps not only to facilitate nanogap-based electrochemical measurements, but also to gain high time resolution, signal-to-background ratio, and kinetic selectivity for dopamine against ascorbic acid. Specifically, ~7 μm-diameter carbon fibers are inserted into a double-bore glass capillary, heat-pulled, and milled by focused-ion-beam technology to yield ~50 μm-long double-cylinder UMEs. The redox cycling of the Ru(NH₃)₆^3+/2+ couple across a nanogap between voltammetric generator and amperometric collector electrodes reaches quasi-steady states at fast scan rates of 100 V/s as demonstrated experimentally and even 1000 V/s as predicted theoretically. The transient background of the amperometric collector response is suppressed ~100 times in comparison with that of the voltammetric generator response. Nanogap voltammograms based on the collector response against the cycled generator potential are quantitatively analyzed without background subtraction to reproducibly yield nanogap widths of ~0.18 μm and a standard electron-transfer rate constant of 0.9 cm/s. Moreover, nanogap-mediated redox cycling can be initiated by dopamine oxidation at the generator electrode to largely improve the dopamine selectivity of the collector response against ascorbic acid, which is also oxidized at the generator electrode to immediately and irreversibly produce a redox-inactive species.

Graphical Abstract

A pair of electrodes with nanometer-wide separation constitutes a nanogap cell to enable unprecedented electrochemical measurements.^1–3 Nanogap-based electrochemical cells were used for the highly sensitive detection of analytes at down to the single molecule level,^4,5 the kinetically selective elimination of interfering species,^6,7 and the measurement of extremely fast electron-transfer kinetics.^8,9 The fabrication of nanogap cells, however, is based on the extensive nanolithography of chip-type electrodes^5,10 or nanoscale electrode positioning by scanning electrochemical microscopy (SECM),^11,12 thereby preventing a wider range of applications. Moreover, the power of nanogap electrodes has not been fully explored owing to limited designs. Recently, we employed finite element simulation to predict that a parallel pair of cylindrical ultramicroelectrodes (UMEs) with nanometer-wide separation enables quasi-steady-state voltammetry at fast scan rates without serious interference from transient background¹³ in contrast to fast-scan cyclic voltammetry with single-cylinder UMEs.¹⁴ In addition, the probe-type design of nanogap-based double-cylinder UMEs is potentially useful for implantation into tissues with minimal damage¹⁵ to enable in-vivo measurements with high temporal and spatial resolution as practiced by using single-cylinder UMEs.¹⁴ There, however, have been reports of double-cylinder UMEs only with micrometer-wide gaps,¹⁶ which were used to assess steady-state theories¹⁷ at slow scan rates. Moreover, nanogap cells were characterized voltammetrically only at slow scan rates of ≤0.1 V/s.^2,3

Herein, we report the simple fabrication of UMEs based on a pair of carbon fibers (CFs) with nanometer-wide separation (Figure 1A) to voltammetrically achieve the high temporal resolution, low background current, and high kinetic selectivity at fast scan rates. In this work, the potential of the generator electrode is scanned to reduce Ru(NH₃)₆³⁺ to Ru(NH₃)₆²⁺, which is amperometrically detected at the collector electrode at a constant potential (Figure 1B) to regenerate Ru(NH₃)₆³⁺ and, subsequently, complete redox cycling (red and blue arrows in Figure 1C). We demonstrate experimentally and theoretically that the efficient diffusion of the Ru(NH₃)₆^3+/2+ couple across a nanogap establishes quasi-steady-state redox cycling when the generator potential is cycled at 100 V/s and 1000 V/s, respectively. Moreover, the transient background of amperometric collector response is ~100 times lower than that of the generator voltammogram based on the plot of the generator current against the generator potential. Accordingly, background subtraction is not needed to quantitatively analyze the resultant nanogap voltammogram based on the plot of the amperometric collector response against the generator potential by the finite element method,¹³ which yields reproducible gap widths of ~180 nm. Preliminarily, we apply double-CF UMEs to detect dopamine (DA) with improved kinetic selectivity against ascorbic acid (AA) at fast scan rates of up to 100 V/s.

Figure 1.

Scheme of (A) a double-CF UME and (B) waveforms for nanogap voltammetry. E_g and E_c are potentials of generator and collector electrodes, respectively. Red dotted line in part A corresponds to the cross section shown in part C for redox cycling of the Ru(NH₃)₆^3+/2+ couple across the nanogap (red and blue arrows) as well as the reduction of Ru(NH₃)₆³⁺ at the side of the generator electrode far from the nanogap (black arrows).

We obtained double-CF UMEs without extensive nanolithography^5,10 or nanoscale SECM^11,12 by adapting a fabrication procedure established for single-CF UMEs (Supporting Information), which have been widely used for in-vivo electrochemical measurements.¹⁴ In fact, double-CF UMEs were prepared previously by heat-pulling a theta glass capillary filled with CFs, but were further polished to yield double-disk CF UMEs with gap distances of <1 μm.¹⁸ In this work, we obtained ~50 μm-long CFs exposed from the glass capillary (Figure 2A) by milling as-pulled double-CF UMEs using focused ion-beam (FIB) technology to remove the rest of CFs that eventually contacted with each other (Figures S-1A and S-1B). Importantly, a narrow gap between the resultant short CFs (Figure 2B) was obtained inherently and was not prepared by FIB. We employed FIB to achieve ~100% success of milling CFs to the desirable length, which is long enough to be manually cut with a lower success rate as demonstrated for the fabrication of single-CF UMEs.¹⁴ The overall yield of good double-CF UMEs was limited by direct contact and subsequent current flow between generator and collector electrodes (Figure S-2). So far, 72 double-CFM UMEs have been milled by FIB and characterized electrochemically to find that 45 of these devices (62.5%) yielded independent current responses at generator and collector electrodes without contact. It should be noted that we used a double-bore glass capillary with disk-shaped channels instead of a theta glass capillary with semielliptical channels¹⁸ to seal the gap between CF and glass wall without using a glue, which filled the gap between CFs. A wider gap can be obtained by using a capillary with a wider bore–bore separation (Figures S-1C and S-1D).

Figure 2.

Scanning electron microscopy of (A) a double-CF UME sealed in a glass capillary and (B) a nanogap between CFs.

Double-CF UMEs were characterized voltammetrically using a commercial bipotentiostat with some modifications (Supporting Information) to measure both generator and collector responses when the generator potential was cycled at up to 100 V/s (Figure 3). As the scan rate was decreased from 100 V/s to 2 V/s, the voltammetric generator response changed from a transient response with a pair of anodic and cathodic peaks to a quasi-steady-state response with a sigmoidal shape and a hysteresis (top panel of Figure 3A). The transient generator response is attributed to a current response based on the diffusion of the Ru(NH₃)₆^3+/2+ couple from the solution to the side of the generator surface far from the nanogap (black arrows in Figure 1C).¹³ By contrast, the corresponding amperometric collector response maintained a sigmoidal dependence on the generator potential even at 100 V/s (bottom panel of Figure 3A), thereby ensuring quasi-steady-state redox cycling across the nanometer-wide gap between generator and collector electrodes¹³ (blue and red arrows in Figure 1C). Importantly, the capacitive background of the amperometric collector response was negligible in comparison with that of the voltammetric generator response. Eventually, both generator and collector electrodes gave quasi-steady-state responses with low background to yield high collection efficiencies of ~0.90 as expected for double-cylinder UMEs under steady states¹⁹ when the scan rate of generator potential was lowered to 0.02 V/s (Figure 3B).

Figure 3.

Generator (top panel) and collector (bottom panel) current responses of double-CF UMEs at (A) 2-100 and (B) 0.02-1 V/s in 10 mM Ru(NH₃)₆³⁺ and 1 M KCl. Solid and dashed lines represent forward and reverse scans, respectively. Collector potential at −0.05 V.

We analyzed nanogap voltammograms by the finite element method based on a 2D diffusion model of double-cylinder UMEs¹³ (Supporting Information) to estimate gap widths of 0.18 ± 0.03 μm with five electrodes. In this analysis, we fixed a CF radius of 3.5 μm to fit experimental nanogap voltammograms at 100 V/s with those simulated by adjusting nanogap width, CF length, and standard electron-transfer rate constant, k⁰, at the generator electrode (Figure 4A). The hysteresis of a nanogap voltammogram served as a measure of gap width, whereas the limiting current was based on the convolution of gap width and CF length.¹³ The CF length and k⁰ value thus determined at 100 V/s yielded good fits of nanogap voltammograms at slower scan rates (Figures 4B, S-3A, and S-3B) with those simulated by adjusting gap width. In addition, parameters determined from nanogap voltammograms were used to simulate the generator response, which required background subtraction. A comparison of a collector response with a background-subtracted generator response at the switching potential yielded higher collection efficiencies of 0.7 ± 0.1, 0.83 ± 0.05, 0.89 ± 0.04, and 0.91 ± 0.03 with five electrodes as the scan rate decreased from 100 V to 10, 1, and 0.1 V, respectively, as expected theoretically.¹³ We also extended the finite element simulation to demonstrate that the gap of the double-CF UMEs developed in this study is narrow enough to establish quasi-steady-state redox cycling even at 1000 V/s (Figure S-5). Furthermore, 3D model was used to find that efficient redox cycling across the nanogap enhanced generator responses by a factor of ~2.5 and ~10 at 100 and 0.1 V/s, respectively, in comparison with the corresponding transient responses of single CF UMEs with identical sizes (Figure S-7).

Figure 4.

Experimental (lines) and simulated (circles) current responses of generator (blue) and collector (red) electrodes in 10 mM Ru(NH₃)₆³⁺ and 1 M KCl at (A) 100 and (B) 0.1 V/s. Only experimental generator responses were background-subtracted. Solid lines and closed circles represent forward scans. Dashed lines and open circles correspond to reverse scans. Simulation employed k⁰ = 0.9 cm/s, formal potential of −0.185 V, diffusion coefficient of 7.0 × 10⁻⁶ cm²/s, CF length of 51.1 μm, and gap widths of 0.15 ± 0.02 μm.

Excellent fits between experimental and simulated voltammograms yielded k⁰ at the generator electrode in addition to gap width and CF length (~50 μm). Both generator and collector responses were quasi-reversible in 1 M KCl solution of extremely pure water with low total organic carbon (TOC) of 2–3 ppb to yield a k⁰ value of 0.9 cm/s, which is close to a k⁰ value of 2.2 cm/s predicted for the Ru(NH₃)₆^3+/2+ couple by Marcus theory for adiabatic outer-sphere electron transfer.²⁰ By contrast, other electrode materials yielded k⁰ values of ≥10 cm/s for the Ru(NH₃)₆^3+/2+ couple to exceed the theoretical value.²¹ These anomalously high k⁰ values are attributed to inner-sphere electron transfer as discussed qualitatively in our recent work²¹ and quantitatively in Appendix of Supporting Information to estimate k⁰ values of 30–46 cm/s, which are comparable to those measured in low-TOC ultrapure water (e.g., 36 ± 4 cm/s at single Pt nanoparticles²²). Without the use of such pure water, k⁰ values (e.g., 10 ± 5 cm/s at metallic carbon nanotubes²³) still exceed the value predicted by Marcus theory, but go below the values estimated with inner-sphere electron transfer as attributed to adventitious surface contamination in the Appendix.

Preliminarily, we tested double-CF UMEs to detect 50 and 100 μM DA in Tris buffer at pH 7.4 that mimics a cerebellum fluid (Supporting Information). Colletcor responses at 0.1 V/s were similar to generator responses (Figure 5A), where the large background was subtracted for the latters to yield high collection efficiencies of 0.87 ± 0.03 at forward peak potentials using four electrodes (Figure S-8). This result indicates that dopamine-o-quinone (DOQ) was oxidatively generated from DA at the generator electrode and was reduced to DA at the collector electrode to efficiently complete quasi-steady-state redox cycling. Noticeably, generator and collector responses were lower during the reverse scan than the forward scan. This hysteresis, however, was not due to electrode fouling. Both generator and collector responses during the forward scan of the next cycle were recovered (red and green lines in Figure 5A) to continuously yield nearly identical voltammograms. We estimate quantitatively that the cyclization of DOQ (Figure S-9) is too slow²⁴ to produce leucodopaminochrome as the precoursor of a blocking polymer film²⁵ before DOQ is transported far away from the generator electrode or reduced at the collector electrode (Supporting Information).

Figure 5.

Generator and collector current responses of double-CF UMEs at (A) 0.1 and (B) 100 V/s in Tris buffer containing 0 (black), 50 (blue), and 100 (red and green for first and second cycles, respectively) μM DA. Collector potential at −0.3 V.

DA was detectable also at 100 V/s (Figure 5B) to complete redox cycling based on DA oxidation at the generator electrode and DOQ reduction at the collector electrode. DA molcules adsorbed on the CF surface were oxidized at the generator electrode to yield an enhanced peak-shaped response, which was proportional to the scan rate.²⁶ The collector response maintained a sigmoidal shape, but also increased significantly in comparison with that at 0.1 V/s owing to the reduction of DOQ desorbed from the generator electrode. The collector response, however, was several times lower than the peak current response of the generator electrode, which suggests that only a small fraction of DOQ was desorbed from the generator electrode during the fast potential cycle. In fact, the time scale of DOQ desorption from the CF surface (sub-seconds²⁶) is much slower than the time scale of the potential cycle (~30 ms) at 100 V/s. Noticeably, collector responses were distorted by capactive coupling with generator responses as observed typically with nanogap-based electrochemical cells.²⁷ Most clearly, the peak-shaped ganerator response based on the oxidation of adsorbed DA was coupled with a peak-shaped collector response at 0.45 V. Moreover, a non-zero background collector response with a spike at 1.0 V was coupled with a capacitive background response of the generator electrode. Importantly, the reduction of DOQ at the collector electrode dominated a sigmoidal collector response, which was not observed when the collector potential was too positive to reduce DOQ (Figure S-12).

We investigated generator and collector responses with DA in the presence of AA (Figure 6), which is co-existent with DA in vivo and is oxidized as readily as DA.¹⁴ We employed nanogap voltammetry at 0.1 V/s (bottom panel of Figure 6A) to obtain low collection efficiencies of 0.02 ± 0.01 for 500 μM AA at the switching potential using seven electrodes. In fact, the collector response was largely increased by the addition of 100 μM DA. By contrast, the generator response was much higher for AA and was only slightly increased by the addition of 100 μM DA (top panel of Figure 6A). The collection efficiency of AA is low, because the oxidative product of AA, i.e., dehydroascorbic acid, is irreversibly and immediately hydrolyzed to a redox-inactive species.²⁸ More quantitatively, a low collection efficiency of 0.02 for 500 μM AA yields 10 μM of dehyroascorbic acid at the collector electrode and corresponds to a high first-order rate constant of 1.2 × 10⁵ s^–1 for the irreversible hydrolysis of dehyroascorbic acid (Figure S-10). Interestingly, a short-lived product was barely detectable at the colelctor electrode owing to the narrow gap. We estimate that a wider gap of ~0.3 μm will lower the collection efficieny of AA to 0.0001 to nearly eliminate the contribution of AA to the colletor response (Supporting Information). Importantly, narrow gaps with widths of either 0.18 or 0.3 μm prevent the relatively fast reduction of DOQ by AA²⁹ (Figure S-11) to yield a high collection efficiency for DA in the presence of AA (Supporting Information).

Figure 6.

Generator and collector current responses of double-CF UMEs at (A) 0.1 and (B) 100 V/s in Tris buffer containing 500 μM AA with (red and green for first and second cycles, respectively) and without (blue) 100 μM DA. Black lines represent background responses in the buffer. Collector potential at −0.3 V.

We were able to obtain a higher collector response to 100 μM DA than 500 μM AA even at 100 V/s, which is 1000 times faster than employed in the previou study of DA detection in the presence of AA with a nanogap cell.⁷ Quantitatively, a collector response with 500 μM AA was ~10 nA with respect to the background at >0.5 V during the forward scan and was much smaller than the corresponding change of ~80 nA in the collector response by the addition of 100 μM DA (bottom panel of Figure 6B). The collector response with DA and AA yielded peaks capactiviely coupled with generator responses to AA at 0.25 V during the forward scan as well as DA at 0.55 V and 0.25 V during forward and reverse scans, respectively (top panel of Figure 6B). Noticeably, sigmoidal generator (0.1 V/s) and collector (0.1 and 100 V/s) responses shifted to more positive potentials in the presence of AA (compare Figure 6 with Figure 5), which slowed DA oxidation at the generator electrode. Moreover, AA exerted a thermodynamic effect to shift the entire generator voltammogram at 100 V/s to slightly more positive potentials than without AA. These kinetic and thermodynamic effects did not affect the diffusion-limited current of the collector electrode at the generator potential of >0.5 V as a measure of DA concentration.

In summary, we demonstrated the simple fabrication of double-CF UMEs with nanogaps and their application for quasi-steady-state nanogap voltammetry at high scan rates with the suppressed background. Reproducible gap widths of 0.18 ± 0.03 μm were obtainable without the use of extensive nanolithography^5,10 or nanoscale SECM^11,12 as determined from nanogap voltammograms. The fastest scan rate of 100 V/s in this study far exceeded scan rates of ≤0.1 V/s employed previously for nanogap-based voltammetry^2,3 including the selective detection of DA in the presence of AA.⁷ The probe-type design of double-CF UMEs as well as their high DA selectivity against AA is attractive for in-vivo applications, where a wider gap will be useful not only to further reduce AA interference and possibly capacitive coupling, but also to maintain a nanogap without CF–CF contact. Such applications, however, will require us to better understand and optimize the surface-dependent electrochemistry of DA³⁰ at double-CF UMEs exposed to ion and electron beams as well as to develop a faster bipotentiostat for potential scans at ≥400 V/s as practiced for fast-scan cyclic voltammetry.¹⁴ More generally, the simple fabrication of double-CF UMEs will promote the exploration and application of nanogap-based electrochemical measurements to complement approaches based on nanolithography^5,10 and nanoscale SECM,^11,12 where fast-scan nanogap voltammetry will be also applicable.