An electrically actuated, carbon nanotube-based biomimetic ion pump

Abstract

Single-walled carbon nanotubes (SWCNTs) are well-established transporters of electronic current, electrolyte, and ions. In this work, we demonstrate an electrically actuated biomimetic ion pump by combining these electronic and nanofluidic transport capabilities within an individual SWCNT device. Ion pumping is driven by a solid-state electronic input, as Coulomb drag coupling transduces electrical energy from solid-state charge along the SWCNT shell to electrolyte inside the SWCNT core. Short-circuit ionic currents, measured without an electrolyte potential difference, exceed 1 nA and scale larger with increasing ion concentrations through 1 M, demonstrating applicability in physiological (~140 mM) and saltwater (~600 mM) conditions. The interlayer coupling allows ionic currents to be tuned with the source-drain potential difference and electronic currents to be tuned with the electrolyte potential difference. This combined electronic-nanofluidic SWCNT device presents intriguing applications as a biomimetic ion pump or component of an artificial membrane.

Graphical Abstract

Single-walled carbon nanotubes (SWCNTs) are well-established platforms for transporting electronic charge and electrolyte. They can be used as semiconducting channels in electronic field-effect transistors,^1,2 where SWCNT shells have been shown to ballistically transport charge,^3,4 or as electrolyte channels in ionic field-effect transistors,⁵ where SWCNT cores efficiently transport fluid.^5–9 In electronic circuits, SWCNTs show tunable conductivity through a back gate potential, based on modulation of Schottky barriers at metal-SWCNT junctions.^10,11 In ionic circuits, SWCNTs show tunable conductivity through the same gate potential, based on modulation of electrolyte charge within the core.⁵ Despite their promise, these electronic and ionic transport capabilities have yet to be integrated into a single SWCNT. Such a device could afford new opportunities in biomimetic ion pumping¹² or artificial membrane engineering.¹³

SWCNTs already have shown promise as synthetic ion channels, where they resemble biological counterparts.^7,9,14 The atomically-small SWCNT core rejects macromolecules and chemical functionalization of SWCNT ends yields selectivity for small counter-ions.^7,9,15,16 In these prior works, potential differences or pressure gradients drive electrolyte and ions with anomalously large flows, enhanced mobility, and single-file transport.^{5,7,9,14–17} In our work, integrating ionic and electronic transistors into individual SWCNTs allows us to transport electrolyte using a solid-state electronic input. This electronic input pumps ions through Coulomb drag energy transfer from electronic currents traversing the shell to electrolyte within the core.

Coulomb drag is an effect where mobile charge across a conductor induces a short-circuit current or open-circuit voltage across a nearby yet electrically isolated conductor.^18–23 Mobile charge in the “drive layer” transfers energy to quiescent charge in the “drag layer,” inducing a drag layer current that flows without a supporting potential difference. Coulomb drag has been theoretically^24,25 and experimentally^26,27 demonstrated in SWCNTs by exposing them to fluid flow and measuring corresponding open-circuit voltages. Potential differences across SWCNTs can likewise drive fluid flow, due to the mutual nature of Coulomb interactions.²⁷ We engineer these Coulomb drag effects to pump ions through SWCNTs without requiring electrolyte potential differences or pressure gradients.

Fig. 1a shows the simultaneous employment of a single SWCNT as an ionic and an electronic field-effect transistor. The SWCNT connects two electrolyte reservoirs containing Ag/AgCl pseudo-reference electrodes for applying electrolyte potential differences (V_ionic) and measuring ionic currents (I_ionic). Metal electrodes contact the SWCNT for applying source-drain potential differences (V_ds) and measuring electronic currents (I_d). During operation, the source terminal is maintained at equipotential with the electrolyte reservoir on the same side of the SWCNT; V_ds and V_ionic are applied with respect to these equipotential electrodes, through the drain terminal and opposing electrolyte reservoir (Fig. 1b). Interlayer Coulomb coupling arises due to the proximity of the ionic and electronic circuits, such that I_ionic signals are tunable with V_ds and I_d signals are tunable with V_ionic.

Figure 1.

Experimental overview. a, A single SWCNT connects two electrically isolated circuits. Electronic current (I_d) flows along the SWCNT shell and through gold electrodes, while ionic current (I_ionic) flows through the SWCNT core and Ag/AgCl reference electrodes. The proximity of the circuits results in Coulomb drag energy transfer between layers (orange arrows), driving I_ionic to flow without an electrolyte potential difference. Currents primarily contain positive charge (holes, cations), such that Coulomb drag coupling is repulsive. b, Device schematic, illustrating SWCNT location and application of source-drain (V_ds), electrolyte (V_ionic), and back-gate (V_gs) potential differences. c, Scanning electron micrograph image of isolated SWCNT with source and drain electrodes. d, SWCNT exhibits ion selectivity comparable to biological ion channels, excluding anions and macromolecules. e, Photograph of a fabricated device. Note: 100-nm gold electrodes appear gray due to thickness.

Because our device leverages Coulomb drag coupling between ionic and electronic currents, it operates as an electrically actuated, biomimetic ion pump. When applying V_ds, I_d carriers traversing the SWCNT shell transfer energy to ions within the SWCNT core through Coulomb drag. In turn, ions are pumped without a V_ionic potential difference (Fig. 1a, orange arrows). The observed Coulomb coupling is consistent with reciprocal energy transfer, showing similar performance when employing I_d or I_ionic as the drive layer signal. Coulomb drag interactions are repulsive due to I_d being hole-dominated and I_ionic being cation dominated.^7,9,16,28,29 In addition to Coulomb drag, interlayer coupling exists through field-effect gating of I_ionic with V_ds and of I_d with V_ionic.

Fig. 1b depicts a chip-level schematic of the device (side-view in Fig. S1). Initially, chemical vapor deposition SWCNTs nucleate and grow from iron catalyst particles over highly doped silicon with thermal oxide. Raman spectroscopy confirms that growth conditions are tuned to yield single-walled carbon nanotubes (Fig. S2). After growth, an isolated nanotube is located with scanning electron microscopy and confirmed to be single-walled with atomic force microscopy (Fig. S3). Average diameters are 1 nm (Table S1) and lengths are hundreds of μm. Metal electrodes are deposited over suitable SWCNTs (Fig. 1c), followed by PMMA patterning to form electrolyte reservoirs and expose SWCNTs to oxygen plasma. Plasma treatment truncates SWCNTs at electrolyte reservoir edges, shortening SWCNTs to tens of μm and functionalizing the outer rings with carboxylic acid groups. These outer rings become negatively charged in solution, yielding selective transport of small cations similar to biological ion channels (Fig. 1d). A PDMS cover interfaces the electrolyte reservoirs with external tubing. Electrical contact to the silicon substrate allows for modulation of SWCNT conductivity through the back-gate (V_gs) configuration, though V_gs is only applied during device verification. A fully fabricated device photograph is provided in Fig. 1e.

Control tests verify leakage-free ionic and electronic circuit pathways. Before introducing electrolyte, we verify the electronic circuit by sweeping V_gs from −10 V to 10 V with V_ds = 0.1 V. SWCNTs demonstrate unipolar (p-type) or ambipolar conductivity (Fig. S4), findings consistent with intrinsically p-type SWCNTs and electric-field gating of the Schottky barriers at the metal-nanotube contacts.^11,28 Schottky barrier variation across devices is such that I_d ranges from several nA to several uA. Next, we verify the ionic circuit based on observing linear ionic conductance (G_ionic) that scales semi-logarithmically with electrolyte concentration (c) as G_ionic ≈ c^b. We extract b from linear fits to log-log conductance-concentration data (Fig. S5), where G_ionic is calculated from linear I_ionic-V_ionic measurements (Fig. S6, red). Prior SWCNT studies measure b ≈ 1/3^5,6 and our experimental values range from 0.22 to 0.37 (Table S1). These measurements confirm that ionic transport through the SWCNT core is observed, rather than through possible leakage paths.

The coupling between the ionic and electronic circuits becomes evident during ion circuit verification. Linear ionic conductance measured with floating metal electrodes (electronic amplifier off) becomes nonlinear when setting V_ds = 0 (Fig. S6, blue). The rectified I_ionic signal is attributed to mismatched Schottky barriers at metal-SWCNT contacts, which are highly sensitive to fabrication conditions.²⁹ Due to this barrier height mismatch, applying V_ds induces asymmetric charge distribution along the SWCNT shell, rectifying I_ionic-V_ionic and I_d-V_ds curves.³⁰ I_ionic magnitudes increase when V_ds = 0 (Fig. S6), demonstrating that the charge redistribution lowers the energy barrier to ion transport.

To study Coulomb drag ion pumping and interlayer coupling, electrolyte reservoirs are flushed with KCl solutions and V_ionic and V_ds are swept from −0.3 V to 0.3 V in 0.1 V increments (with V_gs = 0). At every condition, currents are recorded as averages with minimum integration times of two seconds, yielding signal-to-noise ratios above nine. The subsequent discussion focuses on three representative devices; characterizations of all working devices in 10 mM electrolyte are provided in Table S1. Ionic leakage paths through a polymer layer eventually develop in many of these devices, and are evidenced by abrupt orders-of-magnitude increases in ionic current magnitudes. All reported measurements are recorded before these leakage currents appear.

Fig. 2 shows I_ionic versus V_ionic when varying V_ds (Fig. 2a) and I_d versus V_ds when varying V_ionic (Fig. 2b), for the smallest-diameter SWCNT (Device A) in 10 mM electrolyte. Notably, I_ionic is nonzero at V_ionic = 0 and I_d is nonzero at V_ds = 0 (Figs. 2a,b, boxed regions) due to Coulomb drag energy transfer. These short-circuit Coulomb drag currents, denoted $I_{\mathit{\text{ionic}}}^{*}$ and $I_d^{*}$, are plotted in Fig. 2c versus the corresponding drive layer voltages. Because Coulomb drag currents oppose drive layer voltages, the interlayer interactions are repulsive. The nonlinear dependences of Coulomb drag currents on drive layer voltages are consistent with Fig. 2a,b: the larger $I_d^{*}$ when V_ionic > 0 results from the larger driving current (I_ionic) when V_ionic > 0 (Fig. 2a), while the larger $I_{\mathit{\text{ionic}}}^{*}$ when V_ds < 0 results from the larger driving current (I_d) when V_ds < 0 (Fig. 2b). In other words, with the electronic circuit as the drive layer, negative I_d exceeds positive I_d (Fig. 2b, gray curve), such that $I_{\mathit{\text{ionic}}}^{*}$ is largest for negative drive voltages (Fig. 2c, red curve). With the ionic circuit as the drive layer, positive I_ionic exceeds negative I_ionic (Fig. 2a, gray curve), such that $I_d^{*}$ is largest for positive drive voltages (Fig. 2c, blue curve).

Figure 2.

Mutual Coulomb drag energy transfer in Device A with 10 mM electrolyte. a, Modulation of I_ionic by sweeping V_ds. b, Modulation of I_d by sweeping V_ionic. a, b, Coulomb drag results in short-circuit currents ($I_{\mathit{\text{ionic}}}^{*}$, $I_d^{*}$) being recorded with V_ionic (V_ds) = 0 V (boxed regions). The potential difference applied across the alternate conduction layer is the drive layer voltage. c, $I_{\mathit{\text{ionic}}}^{*}$ (red) and $I_d^{*}$ (blue) as functions of drive voltages. d, Schematic illustration of Coulomb drag interactions. With a voltage applied across the drive layer and the drag layer short-circuited, mobile charge in the drive layer induces transport in quiescent charge in the drag layer. Coulomb drag is repulsive due to currents being primarily comprised of positive charges (cations, holes).

This observation that larger drive currents induce larger drag currents is consistent with the proposed Coulomb drag energy transfer mechanism. The top (bottom) of Fig. 2d qualitatively depicts these repulsive Coulomb interactions driving $I_{\mathit{\text{ionic}}}^{*}$ and $I_d^{*}$. Upon short-circuiting the ionic (electronic) circuit and applying a potential difference across the electronic (ionic) circuit, energy transfers from the mobile electronic (ionic) charge to the quiescent ionic (electronic) charge, dragging a current through the short-circuited layer. Drive and drag layer currents oppose each other due to cations being the primary current carrier in the SWCNT core and holes being the primary current carrier along the SWCNT shell.

Device A shows mutual Coulomb drag because I_d and I_ionic are of comparable magnitude. It is more typical that I_d >> I_ionic due to the high electrical conductivity of SWCNTs, as illustrated by Device B in 10 mM electrolyte. For this device, Fig. 3 shows I_ionic versus V_ionic when varying V_ds (Fig. 3a) and I_d versus V_ds when varying V_ionic (Fig. 3b). Under these I_d >> I_ionic conditions, I_d drives $I_{\mathit{\text{ionic}}}^{*}$ (Fig. 3a, inset) but I_ionic is too small to drive $I_d^{*}$ above the noise floor. Thus, the I_ionic-V_ionic characteristics of Devices A and B show a similar V_ds dependence, while the I_d-V_ds characteristics of Device B show minimal dependence on V_ionic.

Figure 3.

One-directional Coulomb drag energy transfer in Device B with 10 mM electrolyte. a, Modulation of I_ionic by sweeping V_ds. (Inset) $I_{\mathit{\text{ionic}}}^{*}$ versus V_ds. b, When I_d >> I_ionic, I_d is minimally modulated with V_ionic and $I_d^{*}$ is below the noise floor.

We identify the influences of SWCNT diameter and field-effect gating on Coulomb drag-induced ion pumping by comparing $I_{\mathit{\text{ionic}}}^{*}$ signals between Devices A and B. Because the reductions in $I_{\mathit{\text{ionic}}}^{*}$ between Fig. 2c and Fig. 3a (inset) are similar to the diameter ratio between the SWCNTs (0.6 nm and 1.8 nm, respectively; Table S1), we propose that Coulomb drag coupling is enhanced in smaller-diameter SWCNTs. However, the correlation between I_d and $I_{\mathit{\text{ionic}}}^{*}$ magnitudes in Device A is not reproduced in Device B, where positive I_d exceeds negative I_d (Figure 3b), yet negative $I_{\mathit{\text{ionic}}}^{*}$ exceeds positive $I_{\mathit{\text{ionic}}}^{*}$ (Figure 3a, inset). We attribute this discrepancy to field-effect gating of I_ionic by V_ds, in which the V_ds field modulates cation concentration within the SWCNT.⁵ In this model, a negative V_ds signal increases the cation concentration within the SWCNT core while a positive V_ds signal reduces it, complicating the relationship between drive and drag layer currents.

We study the concentration dependence of Coulomb drag ion pumping with Device C (10 mM characterization in Fig. S7), which is similar in diameter to Device B and measured through 1 M electrolyte conditions. Fig. 4a depicts $I_{\mathit{\text{ionic}}}^{*}$ versus V_ds in 10 mM, 100 mM, and 1 M electrolytes. Coulomb drag scales beyond conditions where Debye screening lengths are comparable to or less than SWCNT radius (0.75 nm, Table S1), as evidenced by $I_{\mathit{\text{ionic}}}^{*}$ increasing with concentration across all drive voltages. The field-effect gating of the SWCNT core with V_ds is preserved, as positive $I_{\mathit{\text{ionic}}}^{*}$ magnitudes exceed negative magnitudes under all conditions, consistent with the trends in Figs. 2c and 3a. Recalling the relationship G_ionic ≈ c^b, we compare Coulomb drag current scaling to ionic conductivity scaling. Fig. 4b plots $I_{\mathit{\text{ionic}}}^{*}$ magnitudes versus electrolyte concentrations for V_ds = ± 0.3 V, with the slopes of linear fits to the log-log data denoted b*. The scaling factor for positive $I_{\mathit{\text{ionic}}}^{*}$ is ~250% greater than for G_ionic (Figure 4b, red curve, b* = 0.64 versus b = 0.25), while the scaling factor for negative $I_{\mathit{\text{ionic}}}^{*}$ is ~50% greater (Figure 4b, blue curve, b* = 0.36). The dependence of $I_{\mathit{\text{ionic}}}^{*}$ scaling on the sign of V_ds is also attributed to the V_ds-induced field-effect gating of electrolyte concentration.

Figure 4.

Concentration-dependent scaling of Coulomb drag currents in Device C. a, $I_{\mathit{\text{ionic}}}^{*}$ versus V_ds at varying electrolyte concentrations. $I_{\mathit{\text{ionic}}}^{*}$ scales through concentrations where Debye screening lengths are shorter than SWCNT radius (0.75 nm). b, $\left|I_{\mathit{\text{ionic}}}^{*}\right|$ versus concentration, with V_ds= +/− 0.3 V. $I_{\mathit{\text{ionic}}}^{*}$ scales faster when V_ds= −0.3 V (b* = 0.64) than when V_ds= 0.3 V (b* = 0.36) due to field-effect where negative (positive) V_ds draws more (fewer) cations into the SWCNT core. $I_{\mathit{\text{ionic}}}^{*}$ scales faster with concentration than ionic conductance (b = 0.25, Table S1).

To establish our SWCNT device as a biomimetic ion pump, it must actively transport ions against electrolyte potential differences. Active transport is achieved when Coulomb drag is substantial enough that I_ionic opposes V_ionic. The energy transduction under these conditions is equal to the product of I_ionic and V_ionic, and operating conditions for maximum efficiency ($I_{\mathit{\text{ionic}}}^{\mathit{\text{MP}}}$, $V_{\mathit{\text{ionic}}}^{\mathit{\text{MP}}}$) are determined by a maximum power rectangle (Fig. 5a). In Fig. 5b, we determine the maximum power transfer efficiencies for Devices A-C at 10 mM as a function of V_ds, defining efficiency as $\eta =\frac{{ V_{\mathit{\text{ionic}}}^{\mathit{\text{MP}}}I}_{\mathit{\text{ionic}}}^{\mathit{\text{MP}}}}{V_{\mathit{\text{ds}}}{I}_d}$. The narrowest SWCNT (Device A) shows three orders of magnitude better efficiency than the wider Device B, suggesting that single-file transport is most conducive to Coulomb drag.^9,17 The efficiency differences between similarly sized Devices B and C suggest that η is strongly affected by parameters other than SWCNT diameter, including Schottky barrier heights, chemical functionalization of SWCNT ends, and SWCNT impurities or defects. We predict that efficiency would most easily improve by increasing the spacing between the source and drain electrodes, which would provide greater surface area for Coulomb interactions.

Figure 5.

Efficiency of active Coulomb drag ion transport. a, Schematic depiction of active transport region, where I_ionic opposes V_ionic due to Coulomb drag current. Transport energy is transduced from the electronic regime to the ionic regime. The operating conditions for maximum transport efficiency ($V_{\mathit{\text{ionic}}}^{\mathit{\text{MP}}},I_{\mathit{\text{ionic}}}^{\mathit{\text{MP}}}$) are given by the maximum power rectangle. b, Maximum transport efficiency (η) for Devices A-C, defined as $\eta =\frac{{ V_{\mathit{\text{ionic}}}^{\mathit{\text{MP}}}I}_{\mathit{\text{ionic}}}^{\mathit{\text{MP}}}}{V_{\mathit{\text{ds}}}{I}_d}$.

In conclusion, we demonstrate mutual Coulomb drag coupling between ionic and electronic charge within a single SWCNT. Electronic currents along the SWCNT shell drag ions through the SWCNT core, yielding an electrically actuated biomimetic ion pump that can transport electrolyte without requiring an electrolyte potential difference or pressure gradient. The solid-state electronic input controls the direction and magnitude of ion pumping, with Coulomb drag coupling scaling through 1 M electrolyte concentrations. Transport through the SWCNT core resembles biological ion channel transport, rejecting macromolecules and capable of selectively transporting cations or anions, based on chemical functionalization. As a controllable and selective ion pump, the SWCNT device is a promising tool for biomimetic ion transport in biosensing, nanofluidic, and filtration applications.

Methods.

CNT growth and device fabrication.

Carbon nanotubes were grown with iron-stored ferritin catalysts (Sigma-Aldrich, from equine spleen CAS number 9007-73-2) and chemical vapor deposition, on 2-cm-by-2-cm, highly doped 0.001–0.005 Ω-cm Si chips coated with 285nm SiO₂. After an air anneal, the system was flushed with argon and hydrogen gas to reduce and activate the catalyst. The furnace temperature was then raised to 890 C, after which controlled rates of argon and hydrogen gas flowed over and vaporized ice cold ethanol to provide a carbon source for the nanotube growth process, which lasted one hour. SWCNT growth was imaged in a scanning electron microscope until a suitable nanotube was found, e.g. one that was single-walled, sufficiently long, and not overlapping other nanotubes.

After finding an appropriate SWCNT, atomic force microscopy was used to determine its diameter. Source-drain electrodes with 4 μm width and 10 μm spacing were patterned using electron beam lithography (Nanobeam NB4 tool). 100nm gold electrodes were deposited using electron beam evaporation (Angstrom Evovac deposition system) with adhesive layers of chrome (2nm) and palladium (10nm). A subsequent layer of PMMA (Microchem) was spun and patterned as a resist mask, to remove all other nanotubes with oxygen plasma (15s duration).

Finally, two reservoirs were patterned in another layer of PMMA resist. The extremities of the SWCNT were exposed to oxygen plasma (5s duration) to truncate and functionalize the ends at known locations. Approximately one in three plasma treatments successfully opened SWCNTs without creating a leakage path. A pre-prepared system of fluid ducts embedded in polydimethylsiloxane (PDMS) (Sylgard 184) was stamped over the chip, such that the channels were aligned over the PMMA reservoirs and the source-drain electrode contacts remained exposed. Ag/AgCl electrodes (BASi systems) were used to access the KCl reservoirs (Fisher Scientific, NH, USA, part number: 7447-40-7) via tubing through the PDMS, and the entire chip was placed in a 3D printed mold. The mold interfaced pins with the electrodes inside the CNT device, breaking out the electronic and ionic signals to benchtop electronics.

Data acquisition and analysis.

Two low noise current preamplifiers (Stanford Applied Research, SR570) were employed to apply V_ds and V_ionic to electrodes and measure the resulting currents. Offsets were compensated to ensure I_d and I_ionic were zero at V_ds = V_ionic = 0 V. Preamplified current values were acquired using a DAQ card (National Instruments) sampled at 20 kHz. Data was subsequently analyzed using MATLAB.