A chemical sensor is a transducer that can convert chemical information, ranging from the concentration of a specific sample component to total composition analysis (like gas or ion concentrations), into signals that can be read out by an observer or by an (mostly electronic) instrument. Chemical sensors may be classified according to the operating principle of the transducer, including optical, electrochemical, electrical, mass sensitive, magnetic, and thermometric devices. Most of the SGGT-based chemical sensors are electrochemical or electrical ones, in which the interaction between the analyte and the sensitive layer induces a potential change that is then transformed into a change of the channel current. SGGTs have been successfully used as pH sensors and ion (i.e., K+, Na+, Ca2+, Hg2+,Mg2+, Pb2+, etc.) sensors that are described as follows.
3.1 pH Sensor
pH sensors have many applications because pH is an important parameter for lots of chemical and biological reactions. The first transistor-based pH sensor is the ion-sensitive field-effect transistors (ISFET) fabricated on silicon chips in 1970 by P. Bergveld. An ISFET normally shows a gate voltage shift ΔVG as a function of pH value of the aqueous solution given by:
where k is the Boltzmann constant, T is the temperature of system, q is the electronic charge, α is a constant less than 1, ΔpH is the pH relative change of solutions. Only in ideal conditions, α is equal to 1 and in this case Equation (4) is called Nernstian relationship. At room temperature (T = 300 K), the gate voltage shift is about 59 mV pH−1 when α = 1.
Then many different types of transistors, including polysilicon thin film transistors, carbon nanotube transistors, oxide nanowire transistors, and organic thin film transistors were successfully used in pH sensors. SGGTs has been used and studied as pH sensors since 2008. Although the sensing mechanism in some devices remains unclear, large variation of pH sensitivities ranging from zero to a value of 99 mV pH−1 (larger than the ideal value given by Nernstian relation) has been reported. Recent research indicated that this variation could be attributed to different graphene quality and a clean graphene device actually had minor sensitivity toward pH.
In 2008, Ang et al. firstly fabricated SGGTs using epitaxial graphene grown on SiC substrates by thermal deposition and patterned by photolithography. Because the insulating substrate is very thick, only top gate transistor can be fabricated on the epitaxial graphene via “solution gating.” The transfer curve exhibited ambipolar behavior with the maximum hole and electron mobilities of about 3600 and 2100 cm2 V s−1, respectively. By applying a gate potential from an Ag/AgCl reference electrode placed on top of the channel, the channel conductance was modulated for only about 30%, which is much lower than those of the SGGTs reported later.[36, 74, 75] So the graphene used in the device may have high density of traps. The transfer curves of the devices shifted to more positive gate voltage with the increase of pH and showed a supra-Nernstian response of 99 mV pH−1, indicating great potential for pH sensing applications. The mechanism may be attributed to the interplay between surface potential modulation by ion adsorption and the attached amphoteric OH– groups on the graphene surface. They also investigated the electrochemical properties of the EDL on graphene using cyclic voltammetry (CV) and frequency dependent impedance methods and found that the graphene/electrolyte interface was very sensitive to pH, which further confirmed that the pH-sensitive behavior of the SGGTs is due to the graphene/electrolyte interface.
Ohno et al. fabricated SGGTs with mechanically exfoliated pristine graphene for electrical detecting pH as well as protein adsorption.[47, 76] The transfer characteristics were measured in buffer solutions with different pH ranged from 4.0 to 8.2. The dependence of the channel conductance at fixed gate and drain voltages on the pH value was also characterized. It was reasonable to find that the Dirac point of the SGGT shifted to a positive direction with increasing pH, similar to the previous report. However, the transfer curve only shifted for about 25 mV pH−1, which is much lower than the above result reported by Ang et al.
Cheng et al. reported performance improvement of SGGTs by suspending them in aqueous solution through a novel in situ etching technique. The transconductance of the device was increased for about two times after the suspension of graphene from the substrate, whereas the low-frequency noise was decreased for about one order of magnitude. Therefore, the sensitivity of a sensor based on the suspended device can be improved. The devices were demonstrated as real-time and sensitive pH sensors in testing solutions with pH values varied from 6 to 9. The Dirac point voltage of the transfer curve shifted positively with the increase of pH while the shift was only about 20 mV pH−1.
Fu et al. found that SGGTs with high-quality graphene were insensitive to the pH values of solutions. As shown in Figure 3a–c, they fabricated devices with CVD-grown graphene and observed little gate voltage shift (6 ± 1 mV pH−1) in the transfer characteristic of a SGGT when the pH of the solutions were varied from 4 to 10. The voltage-shift of the SGGT can be further reduced when the device was covered by a hydrophobic fluorobenzene layer on the graphene. But the voltage shift was increased to 17 ± 2 mV pH−1 when a thin Al2O3 layer was coated on the graphene, as shown in Figure 3d. It is notable that Al2O3 is a pH-sensitive material that has been used in ion-sensitive field-effect transistors before. So, the results suggested that clean graphene was not sensitive to the concentration of proton or pH in solutions, which is a consequence of its ideal hydrophobic surface with few dangling bonds. They believed that the gate voltage shifts induced by pH change reported in the previous literature reflected the quality of graphene. So, it is reasonable to conclude that defective graphene, where free bonds exist on the surface, is sensitive to the pH value of the solution, whereas high-quality graphene with no dangling bonds is not.
Figure 3. a) Optical image of a graphene transistor beneath a liquid channel. b) Schematics of the experimental setup and the electrical circuitry of the SGGT. c) Source-drain conductance (Gsd) as a function of gate voltage applied by the reference electrode (Vref) measured in different pH buffer solutions for an as-prepared SGGT and d) a SGGT with a thin Al2O3 film coated on graphene channel. Reproduced with permission. Copyright 2011, American Chemical Society.
Download figure to PowerPoint
These studies indicated that pH sensors can be realized by using SGGTs only in some special conditions. Compared with typical silicon-based ISFETs that showed pH-dependent gate voltage shifts close to Nernstian relationship,[68, 69] the pH sensors based on graphene transistors exhibited responses diverged greatly from this relationship. Supra-Nernstian response was even observed in the devices with high-density defects in the graphene layers. Therefore, to realize the high-performance pH sensors, it is necessary to know the underline mechanism that can induce the big shift of Dirac point voltage, which is unclear until now. Further work is needed to better understand this effect.
3.2 Ion Sensors
Ion sensors have many important applications, such as environment monitoring, food safety inspections, and so on. On the other hand, the interactions between ions and graphene are important to all types of SGGT-based chemical or biological sensors since the devices operate in aqueous solutions. Besides the pH sensors, various ion sensors based on SGGTs, including Na+, K+, Ca2+, Mg2+, Hg2+, and Pb2+ sensors, have also aroused great attentions in recent years.
Chen et al. prepared SGGTs with mechanically exfoliated graphene on SiO2 substrates. Ag wires or Ag/AgCl reference electrodes were used as gate electrodes. They studied the influence of ionic concentrations in NaF solutions on the device performance and observed the shift of the transfer curve to lower gate voltage with the increase of ionic concentration. Assuming that the potential drop across the electrolyte/SiO2 interface was induced by the impurity charges and partially counteracted by the ions in solution, the influence of ionic concentrations on the shift was then simulated with an analytical model successfully. In addition, they found that the charged impurity on the SiO2 substrates was another important factor that can influence charge transport in graphene layer. By fitting the device performance, the concentrations of charged impurities in different devices were extracted. A clear relationship between the minimum conductivity at the Dirac point and the impurity concentration was obtained. The minimum conductivity decreased exponentially with the impurity density, which was attributed to the impurity scattering of the carriers.
Heller et al. investigated SGGTs composed of single-layer graphene flakes and Cr/Au source and drain electrodes on SiO2 substrates. An Ag/AgCl reference electrode was used as the gate. The device performance was sensitive to the ionic concentrations and pH values of the electrolytes. The transfer curve shifted to positive gate voltage with the increase of pH, similar to the pH-sensitive SGGTs reported before. At the same pH condition, the transfer curve shifted with the change of ionic concentrations. When pH is 7, the shift is −42.7 mV decade−1 for both LiCl and KCl solutions, whereas for pH of 3, the shift is +18.9 mV decade−1. So the shift of the transfer curve was sensitive to different ions without selectivity. They believed that the response to ions in electrolytes could be affected by a high density of ionizable groups on both the underlying substrate and the graphene surfaces. These effects should be considered in many other sensing applications of SGGTs, in which the electrical signal can be affected by charged target molecules as well as electrolyte ions. Therefore, careful control of electrolyte properties is needed in some experiments.
Sofue et al. used mechanically exfoliated graphene to fabricate SGGTs for ion sensors and demonstrated sensitive electrical detection of NaCl of various concentrations in Tris–HCl buffer solution. Sodium ion in electrolytes was found to affect the electrical potential of graphene channels. As a result, the transfer curves shifted toward negative voltages with increasing Na+ concentration. The voltage shift of the SGGT can be used to accurately detect the concentrations from 1.0 × 10−9 m to 1.0 × 10−3 m. The device was also demonstrated for real-time detection of Na+ concentrations with high sensitivity.
In practical applications, device packaging of SGGTs is important to achieving stable performance. SGGTs can be integrated in microfluidic systems with only small active areas exposed to electrolyte while all contacts are well packaged. More importantly, microfluidic chips can be easily fabricated in clean room to achieve low cost, highly sensitivity, and high-throughput detections. Our group fabricated and integrated SGGTs into microfluidic channels on both glass and flexible substrates (PET), as shown in the Figure 4a. The transfer curve of a SGGT with Ag/AgCl gate electrode shifted toward the negative voltage direction with the increase of KCl concentration in the microfluidic channel, as shown in Figure 4b. But the device was insensitive to ionic concentration when the Ag/AgCl gate electrode was replaced with an Au wire, indicating that the gate electrode was responsible for the ion-sensitive performance. As shown in Figure 4c, the voltage shift in the former case was about 61.9 mV decade−1 close to the ideal value (59 mV decade−1) given by Nernst equation at room temperature. So the response was caused by the variation of the potential drop at the Ag/AgCl gate electrode.
Figure 4. a) Schematic diagram of a SGGT with an Ag/AgCl gate electrode integrated in microfluidic channel; b) transfer curves of a SGGT measured in KCl aqueous solutions with different concentrations; c) the shift of Dirac point voltage as a function of KCl concentration. d) Time-dependent channel current of the SGGT characterized in flowing KCl solution with different velocities; e) transfer curves of the SGGT characterized at different flow velocities; f) the shift of Dirac point voltage (ΔVG) of SGGTs on glass or plastic substrates at different flow velocities. Reproduced with permission. Copyright 2012, American Chemical Society.
Download figure to PowerPoint
The SGGT could also be used to detect flow velocities in the microfluidic channel. As shown in Figure 4d, the channel current IDS changes with the variation of flow velocity. We also observed that the transfer curve of the device shifted horizontally with the change of flow velocity shown in Figure 4e. The voltage shift can be fitted with the following equation for the streaming potential (Vstr) generated by the moving counterions inside the EDL:
where ζ is the zeta potential on the surface of the microchip channel, ε0 is the vacuum permittivity, εr is the relative dielectric constant of the electrolyte solution, w, h, and R are the width, height, and flow resistance of the microchannel, respectively; η is the dynamic viscosity of the electrolyte solution, e is the electron charge, C is the ionic concentration, λ is an offset concentration that arises from the background concentration of ions, μ is the effective ionic mobility, and ν is the flow velocity in the microchannel.
Because the streaming potential in Equation (5) is dependent on three physical quantities, including the flow velocity v, the ionic strength of the fluid C, and the zeta potential of the substrate ζ, the device could be adopted for sensing any one of the three quantities when the other two were known. As shown in Figure 4f, the SGGTs on different substrates (glass or plastic) exhibited different sensing behavior, which further confirmed the sensing mechanism of the devices. This flexible, multifunctional, and miniaturized SGGT-based sensor might have great potential for applications in lab-on-a-chip platforms, biological systems, or medical devices.
Recently, Newaz et al. investigated the influence of fluid flow on the performance of a SGGT in a microfluidic channel. Because the change of flow velocity or the concentration of ions (NaCl) would induce transfer characteristic shift due to the change of streaming potential given by Equation (5), they developed a graphene-based mass flow and ionic strength sensors. The flow sensitivity of SGGTs reached about 70 nL min−1, which was about 300 times higher than the reported flow sensitivity of a carbon nanotubes device, and about four times higher compared with a device based on Si nanowire. Their SGGTs could also detect changes in the ionic strength of a moving liquid with the sensitivity of about 40 × 10−9 m. So the devices sensitive to liquid flow and ionic strength may find some applications as mentioned above.
After surface modification on graphene, SGGTs could achieve high sensitivity and specificity to certain kind of ions. Wen et al. reported Pb2+ ion sensors using gold nanoparticle and DNAzyme-functionalized SGGTs. As shown in Figure 5, CVD-grown graphene was decorated with gold nanoparticles that serve as the anchoring sites to covalently immobilize thiolated Pb2+-dependent DNAzyme molecules. Upon binding with Pb2+ ion, the enzymatic strand cleaved the substrate strand and induced the diffusion of enzymatic strand and the unthiolated portion of the substrate strand from the graphene active layer, which altered the original electronic coupling between the charged DNAzyme complex and the graphene. They found that the transfer curve shifted to positive gate voltage after adding Pb2+ ion, indicating the alleviation of n-doping by DNA molecules. So the Dirac point shift was caused by the interaction between DNAzyme molecules and graphene surface rather than between Pb2+ ions and graphene. The detection limit of the devices to Pb2+ was about 20 × 10−12 m, which was several orders of magnitude lower than that of other approaches, such as optical methods. Moreover, the selectivity of the devices was very high because Pb2+-dependent DNAzyme was used as the recognition element.
Figure 5. a) Schematic illustration of the SGGT functionalized with Au nanoparticles (AuNPs) and DNAzyme molecules. b) Schematic of Pb2+-induced self-cleavage of the DNAzyme. c) The Dirac point shifts of SGGTs based on bare graphene, AuNP-decorated graphene, nonspecific DNA–AuNP complex decorated graphene, and DNAzyme–AuNP complex decorated graphene, to various Pb2+ concentrations. d) The averaged Dirac point shift of DNAzyme–AuNP complex decorated SGGTs in the presence of 20 × 10−12 m Pb2+ ion for different incubation time. Reproduced with permission.
Download figure to PowerPoint
Sudibya et al. used micropatterned rGO films to fabricate SGGTs as ion sensors. rGO is the desirable alternative to the pristine graphene due to its low cost, solution processable, and scalable production.[84, 85] After the modification of specific proteins, the devices can be used to effectively detect Ca2+, Mg2+, Hg2+, and Cd2+ ions with high specificity and the detection limit down to 1 × 10−9 m, which is comparable to those of conventional methods. By applying different gate voltages, both the p-type and n-type detection can be easily realized in the same rGO-based SGGT. The simple rGO devices could be readily patterned and fabricated by solution process on various substrates and utilized to detect various metal ions in solutions rapidly and label-freely with high sensitivity and specificity. The devices could be easily integrated in lab-on-a-chip systems as microarrays for parallel recording multiple targets.