Graphene nanonet for biological sensing applications

Ammonium meta-vanadate (Aldrich, 2 mg) and acidic ion-exchange resin (DOWEX 50WX8-100, Aldrich, 20 mg) were mixed in deionized water (40 ml), and the mixture was incubated at room temperature for 20 days. During this time period, V₂O₅ NWs continued to grow and became over 1–10 μm long and 10–20 nm wide (figure S1 in supporting information available at stacks.iop.org/Nano/24/375302/mmedia) [19].

A single-layer graphene was prepared by a chemical vapor deposition method, and it was transferred onto a SiO₂ substrate (1000 Å), as reported previously (figure 1(a)) [20, 21]. Photoresist (PR, AZ5214) was patterned on a single-layer graphene surface via conventional photolithography processes (MA6 aligner, Karl Süss Company, Germany) with a short baking time (<10 min at 95 °C). Then, the V₂O₅ NWs were adsorbed on the graphene surface with PR patterns from its solution (figure 1(b)). Subsequently, the PR was removed by rinsing it with acetone, resulting in the V₂O₅ NWs adsorbed only in the bare graphene region (figure 1(c)). We etched off the sample via the RIE process (South Bay Technology Inc., 60 W, 1 min) using the V₂O₅ NW networks as a shadow mask (figure 1(d)). Then, the V₂O₅ NWs were removed via HCl solution (5%), resulting in the GNN patterns on a SiO₂ surface (figure 1(e)). The HCl treatment removed only V₂O₅ NWs and left GNN structures (figures S2 and S3 in supporting information available at stacks.iop.org/Nano/24/375302/mmedia). The Au electrodes were prepared by photolithography followed by a thermal evaporation (ODIS, Korea) method (10 nm/30 nm of Pd/Au) under a high-vacuum condition (base pressure ∼5 × 10⁻⁶ Torr) (figure 1(f)) [22]. For biosensor devices, the metal electrodes were passivated with a PR layer. Note that this process allowed us to fabricate the patterns of continuous networks of GNRs, which can be ideal for practical device applications.

Zoom In Zoom Out Reset image size

GNN patterns were incubated for 1 h in 6 mg of 1-ethyl-3-(3-dimethylaminopropyl) carbodiimidehydrochloride (EDC) and N-hydroxysuccinimide (NHS) mixed solution. Then, the edges of the GNN were functionalized with NHS molecules. Subsequently, the NHS-functionalized GNN samples were immersed in a solution of amine-terminated DNA (10 μM) with 15 bases (5'-NH2-C6-CAT TCC GAG TGT CCA) and 100 μl of 1% sodium dodecylsulfate (SDS) in 890 μl of 2-(N-morpholino) ethanesulfonic acid (MES) buffer, then incubated for 1 h at room temperature [23]. Note that these processes enabled the amide bonding between the amine group of DNA and the carboxyl group of GNN edges. The physically adsorbed DNA molecules were removed by a 1% SDS wash for 60 min. To verify the tethering of this target DNA, the GNN sample was immersed into a probe DNA (5'-cy3-TGG ACA CTC GGA ATG) terminated with cy3 dye (570 nm emission) solution for 80 min at room temperature. The nonspecifically bound DNA was removed by a 1 h wash with 1% SDS solution. For the electrical sensing experiment, the GNN-FET was exposed to the solution of an amine-terminated probe DNA (10 μM) with 15 bases (5'-NH2-C6-CAT TCC GAG TGT CCA) for 1 h, so that the amine groups in the DNA anchored to the NHS-functionalized GNN surface. Then electrical measurements and DNA sensing experiments were carried out using a Semiconductor Characterization System 4200 (Keithley Instruments Inc.).

Figure 2(a) shows the optical microscope image of line-shaped patterns comprised of GNNs on a SiO₂ substrate. Note that the patterns had well-defined edges and covered a large surface area. It clearly shows that we could prepare uniform GNN patterns over a large surface area.

Zoom In Zoom Out Reset image size

Figure 2(b) is the atomic force microscopy (AFM) topography image of the patterned GNN on a SiO₂ substrate. It shows continuous networks of GNRs with a rather uniform width. In previous works about networks of one-dimensional nanostructures such as carbon nanotubes (CNTs), the contact resistance between individual nanostructures often caused significant problems, such as low mobility and high noise in the device channels. In our case, GNN is comprised of continuous networks of GNRs without contact resistances, which can be a significant advantage in achieving high-performance devices.

Figures 2(c) and (d) show the x-ray photoelectron spectroscopy (XPS) spectra of the bulk graphene layer and GNN, respectively. The bulk graphene layer exhibited only a single peak associated with the sp² carbon bonds (red curve at 284.8 eV). On the other hand, the GNN had peaks associated with oxygen, hydroxyl groups and carboxylic groups as well as C–C bonds (combination of multiple fitting curves). Note that the oxygen-containing groups were attached to the broken carbon bonds on the GNN, which was caused by the RIE process. This implies that the GNN had many functional groups such as OH and COOH at its edge regions [14], which should allow us to chemically functionalize the GNN for various applications [24].

Figures 2(e) and (f) show the Raman spectra of the bulk graphene layer and GNN, respectively. The bulk graphene layer had only G and 2D bands (figure 2(e)), while the GNN had an additional D band (figure 2(f)). Here, the presence of the defect-related D band was attributed to the broken carbon bonds of the GNN [25]. This result confirms that a number of chemically reactive groups existed in our GNNs.

Figure 3(a) shows schematic diagrams depicting the procedure for the functionalization of a DNA on the GNN for optical biochip applications. GNN patterns were expected to have a number of carboxylic groups due to the edges of GNRs it contained (figure 3(a)(i)). These carboxylic groups could be utilized to link with amine-terminated probe DNA via amide bonding after an activation process with NHS/EDC (figure 3(a)(ii)) [26]. Then, the fluorescent DNA probe terminated with cy3 dye was utilized to verify the DNA functionalization of the GNN (figure 3(a)(iii)).

Zoom In Zoom Out Reset image size

Figure 3(b) shows the optical microscopy image of GNN patterns over a large surface area on a SiO₂ substrate. Note that it shows uniform patterns with well-defined edges. Figures 3(c) and (d) show the fluorescence image and its intensity profile of the cy3-DNA-hybridized GNN patterns, respectively. The fluorescent image clearly shows that the cy3-DNA was selectively attached only on the GNN which was functionalized with amine-terminated DNA. The intensity profile of figure 3(d) confirms that a uniform layer of cy3-DNA was formed after the hybridization. These results clearly show that our GNN can be selectively functionalized with biomolecules, and it can be labeled with fluorescence molecules for fluorescence-based biochip applications.

For the control experiment, the amine-terminated DNA attached on GNN was hybridized with non-complementary DNA (5'-cy3-GAT CTG AGT ATC CGT). In contrast to figure 3(c), no significant fluorescence signal was observed (figure S4(a) in supporting information available at stacks.iop.org/Nano/24/375302/mmedia). Furthermore, the nonspecific binding of the cy3-DNAs onto the GNN pattern was not observed either (figure S4(b) in supporting information available at stacks.iop.org/Nano/24/375302/mmedia).

Figure 4(a) shows the I_ds–V_g curves of the bulk graphene layer-based FET (a blue line) and the GNN-based FET (a red line). The length and width of the GNN channel were 10 μm and 2 μm, respectively (figure S5 in supporting information available at stacks.iop.org/Nano/24/375302/mmedia). For the current measurement, the bias voltage of 1 V was applied to the GNN-FETs. The bulk graphene-FET exhibited typical ambipolar characteristics, while the GNN-FET exhibited p-type characteristics (figure 4(a) and figure S6 in supporting information available at stacks.iop.org/Nano/24/375302/mmedia). The p-type characteristics of GNN-FET were attributed to oxygen edge groups which were generated by the RIE process [27]. It should be noted that the FET device based on our networks of GNRs exhibited a rather high mobility (∼100 cm² V⁻¹ s⁻¹) compared to that based on CNT networks [28]. Presumably, the nanoribbons in our GNN were connected continuously without any high-resistance contact points between individual nanoribbons, which should have helped in reducing the scattering of charge carriers and enhancing its mobility. Such high mobility should be advantageous in building high-performance devices based on GNN, such as FETs and sensors.

Zoom In Zoom Out Reset image size

Figure 4(b) shows the current-normalized power spectral density (PSD) S/I² as a function of the frequency f. The noise measurement was performed using a FFT network analyzer (SR770, Stanford Research Systems) with a source–drain bias voltage of 1 V on the GNN-FET. The PSD was found to be inversely proportional to the frequency f, which is a typical low-frequency noise behavior of graphene-based devices [29, 30]. Note that the noise level of the GNN-FET was similar to that of bulk graphene-based devices (S/I² < 10⁻⁸ Hz⁻¹) [29, 30] and lower than that of CNT network-based devices (S/I² < 10⁻⁵ Hz⁻¹) [31]. Furthermore, the noise amplitude normalized by a resistance value (A/R) of the GNN-FET was ∼6.46 × 10⁻¹² Ω⁻¹, which was lower than that of CNT network-based devices [31] and was similar to that of the pristine single-layer graphene-based devices [29, 30]. In previous works, the network-based devices were usually comprised of separate nanostructures such as CNTs or NWs, where the contact resistance between individual nanostructures can cause a significant noise and resistance in the device [32]. In contrast, our GNN is comprised of continuous networks of GNRs, which should have reduced possible noise sources in the channel. Such low noise characteristics of our GNN should be an important advantage for high-precision devices such as highly sensitive sensors.

As a proof of concept, we functionalized GNN with DNA and built DNA sensors. Figure 4(c) shows the real-time detection of complementary DNA (5'-TGG ACA CTC GGA ATG-3') at various concentrations from 10 pM to 1 μM. In this experiment, the GNN channel in a GNN-FET was first functionalized with DNA. Then, a 100 mV bias voltage was applied between the source and drain electrodes, and the source–drain current was monitored after the addition of the target DNA with different concentrations. The red, blue and green lines are the responses of our sensor to the addition of the complementary, non-complementary DNA 1 (5'-TGG AGA CTC GGA ATG-3') and non-complementary DNA 2 (5'-GAT CTG AGT ATC CGT-3'), respectively. Note that the non-complementary DNA 1 had only a single mismatched base compared with the complementary DNA. The addition of the complementary DNA with its concentration over 1 nM resulted in significant current changes, while the sensor did not respond to the non-complementary DNA. This conductance decrease by the complementary DNA could be explained by the electron-doping of the negatively charged DNA on the GNN channel [33]. Our results show that the GNN structures can be functionalized with biomolecules and utilized to build biosensors for the selective detection of biomolecules such as DNA. Also, the limit of detection is comparable or even lower than previous graphene-based electrical sensors for DNA detection [6, 9, 34].

Figure 4(d) shows the response curve of our DNA sensor based on GNN. This sensor response could be fitted by the typical response equation of FET-type sensors, as reported previously [35–38]:

$$\begin{eqnarray*} &&\Delta G=\frac{\Delta {G}_{\mathrm{max}}{C}_{\mathrm{DNA}}}{{K}_{\mathrm{d}}+{C}_{\mathrm{DNA}}} \end{eqnarray*}$$

where ΔG_max and C_DNA are the saturated net change in conductance and the concentration of complimentary DNA in solution, respectively. The term K_d represents the dissociation constant between target and probe DNA molecules. Note that the sensor response curve is fitted well to this theoretical fitting curve (red line). The fitting allowed us to estimate the dissociation constant K_d as 1.4 × 10⁻¹⁰ M, which is similar to the previous reported value measured by other methods [39, 40].