Electrically Controllable Single-Point Covalent Functionalization of Spin-Cast Carbon-Nanotube Field-Effect Transistor Arrays

Abstract

Single-point-functionalized carbon-nanotube field-effect transistors (CNTFETs) have been used to sense conformational changes and binding events in protein and nucleic acid structures from intrinsic molecular charge. The key to utilizing these devices as single-molecule sensors is the ability to attach a single probe molecule to an individual device. In contrast, with noncovalent attachment approaches such as those based on van der Waals interactions, covalent attachment approaches generally deliver higher stability but have traditionally been more difficult to control, resulting in low yield. Here, we present a single-point-functionalization method for CNTFET arrays based on electrochemical control of a diazonium reaction to create sp³ defects, combined with a scalable spin-casting method for fabricating large arrays of devices on arbitrary substrates. Attachment of probe DNA to the functionalized device enables single-molecule detection of DNA hybridization with complementary target, verifying the single-point functionalization. Overall, this method enables single-point defect generation with 80% yield.

Keywords: CNTFET array; DNA melting; carbon nanotube; diazonium; single-point defect; smFET; spin-cast; wafer scale.

Figure 1.

(a) Simulated percentages of zero, single, and multiple nanotube crossings are depicted. 500 electrode pairs are simulated for each density and width pairing. (b) Top-down photograph of fully processed 100 mm Si/SiO₂ wafer. (c) Representative SEM scans of single-nanotube crossings. Electrode width (w), gap (l), and height (h) are noted. Insets in each panel show a more magnified view of the nanotube crossing. (d) Statistical comparison of simulated and experimental yields for spin-cast nanotubes. Simulations were run ten times with a surface concentration of 0.13 CNT/μm² and an electrode width of 20 μm. Experimental yields were computed from SEM images of all electrode pairs on three chips from the same wafer.

Figure 2.

(a) Histograms of the number of electrically conducting devices across seven chips on one wafer. All zero crossings represent electrode pairs which did not conduct (even though they may have one or more nanotubes bridging them in the corresponding SEM image). SEM images are referenced to verify the presence of 1, 2, 3, or 4 CNT crossings. Error bars are calculated from seven chips from the same wafer. (b) Electronic sweeps in sodium phosphate buffer solution (100 mM, pH 8.0) of 49 single-crossing nanotube devices deposited on four separate chips. These devices not only have a single CNT bridge but also exceed threshold I_on values of 1 nA in I–V_lg characteristics at V_sd = 100 mV. Both semiconducting and metallic electrical behavior is evident.

Figure 3.

Liquid-gate-potential-dependent diazonium reaction. (a) One representative I-t trace (515 s) of one m-SWCNT device after introducing 72 μM FBDP solution (at t = 0). V_lg is fixed at −500 mV and V_sd of 50 mV is applied. One conductance level is evident and the corresponding current histograms show a single Gaussian distribution. (b) A representative I-t trace (90 s) of another m-SWCNT device after introducing 10 μM FBDP solution while V_sd of 50 mV and V_lg of 0 V are fixed. Red arrows indicate times at which discrete downward current steps are observed. Corresponding histograms of current show that the current drops are clearly quantized. Counts of the histogram are normalized by dividing by the sum of the total counts. Data in the I-t trace graphs are sampled at 10-ms intervals from 1 kHz raw-data.

Figure 4.

Feedback-controlled diazonium reaction for single-point functionalization. Initially liquid gate potential (V_lg) of −500 mV is applied and subsequently increased to promote the reaction. When a downward current step is detected, V_lg is immediately switched back to −500 mV to halt the reaction. Overlaid I-t and V_lg–t records of Device A1 during FBDP exposure, in which a device is exposed to 72 μM FBDP solution at V_sd of 100 mV (a; left). A zoomed-in view of the red box when V_lg reaches to the threshold reaction potential, in which three consecutive downward steps and one upward step are monitored while holding V_lg at 50 mV for approximately 15 s, indicating two ultimate sp³ defects (a; right). Overlaid time traces of single-point-functionalized devices. Device A2 in (b) is exposed to 72 μM FBDP solution at V_sd of 50 mV. A current step corresponding to 8.7 kΩ is observed when V_lg reaches 200 mV. Device B5 in (c) is exposed to 6 μM FBDP solution at V_sd of 50 mV. A single downward current step results in a resistance change of 9.7 kΩ to the nanotube after V_lg reaches −100 mV. No further current steps are observed after setting V_lg to −500 mV. Data in the I-t trace graphs are sampled at 10-ms intervals from 1 kHz raw-data.

Figure 5.

Verifying the application of point-functionalized CNTFETs by sensing DNA melting dynamics. (a) Probe DNA (green) shown tethered to a two-point-functionalized device (top, ID: Device B3) and a single-point-functionalized device (bottom, ID: Device B5). After tethering probe DNA to the defect site, the solution containing fully complementary target DNA (colored in blue, 100 nM) is introduced to the devices at a fixed temperature (40°C) while holding V_lg at 400 mV. (b) I-t measurements (V_sd = 100 mV) for each device. The total recording time is 180 s, and a representative two-second-length trace of the raw source-drain current (colored in black) is overlaid with an idealized fit (colored in red). Corresponding histograms of the two devices are plotted by counting the current in 60-second-length traces (colored in black) which are fit by two Gaussian distributions. Counts of the histogram are normalized by dividing by sum of the total counts. (c) Plots of the dwell time of the lower conductance state (green) and the high conductance state (blue) are constructed from the idealized trajectories for the total recorded length (180 s) and fit with a double exponential function (red), yielding the rates of hybridization (k_hyb) and melting (k_melt). For the analysis of Device B3, the r-squared value of k_hyb and k_melt are 0.998 and 0.999, respectively. In case of Device B5, the r-squared value of k_hyb and k_melt are 0.999 and 0.996, respectively.

Scheme 1.

Illustration of V_lg-induced diazonium reaction on the CNTFET. 4-Formylbenzene diazonium hexafluorophosphate (FBDP) is employed as the substituent. The hexafluorophosphate (PF₆) counterion provides stability to the FBDP molecule. An aldehyde group opposite to the diazonium cation is used for the DNA conjugation. (a) When V_lg is positive relative to the CNT surface, the Fermi level (E_F) of the m-SWCNT is shifted up, promoting sp³ defect generation by donating the electrons to the positively charged N₂ group of the FBDP to form an aryl-radical. (b) When V_lg is negative, E_F is shifted down, halting the reaction as the population of electrons near the Fermi level is reduced.