Comparison of the Extended Gate Field-Effect Transistor with Direct Potentiometric Sensing for Super-Nernstian InN/InGaN Quantum Dots.

We systematically study the sensitivity and noise of an InN/InGaN quantum dot (QD) extended gate field-effect transistor (EGFET) with super-Nernstian sensitivity and directly compare the performance with potentiometric sensing. The QD sensor exhibits a sensitivity of -80 mV/decade with excellent linearity over a wide concentration range, assessed for chloride anion detection in 10-4 to 0.1 M KCl aqueous solutions. The sensitivity and linearity are reproduced for the EGFET and direct open-circuit potential (OCP) readout. The EGFET noise in the saturated regime is smaller than the OCP noise, while the EGFET noise in the linear regime is largest. This highlights EGFET operation in the saturated regime for most precise measurements and the lowest limit of detection and the lowest limit of quantification, which is attributed to the low-impedance current measurement at a relatively high bias and the large OCP for the InN/InGaN QDs.

Introduction

Electrochemical sensors are finding growing applications in a wide variety of fields, including medical diagnostics, environmental monitoring, food analysis, industrial process control, and biodefence. This is due to their robustness, easy and largely maintenance-free handling, and cost effectiveness. Among the two main implementations, amperometric and potentiometric sensors, potentiometric sensors have wider use.[1−4] They do not require (bio-)chemical reactions with charge transfer at the sensing electrode surface but solely rely on changes in the surface potential upon analyte attachment, measured relative to the constant potential of a reference electrode. This extends not only to the detection of ions as classical health indicators, such as chloride anions, but also to the sensitivity for macromolecular biomarkers up to the detection of virus infections.[5−7] The output voltage or open-circuit potential (OCP) can be read out directly using a high-impedance voltmeter or using a field-effect transistor (FET)-type device with a low-impedance amperemeter. The first implementation of a FET-type device was the ion-selective FET (ISFET), where the bare gate dielectric without metallization is the sensing layer.[8−10] Later, the extended gate FET (EGFET) was introduced with a separated sensing electrode, remotely connected to the gate of an untouched FET.[11−15] In addition to easy miniaturization, compatibility with complementary metal oxide semiconductor mass-production processes at low cost, and possibility of on-chip integration in multisensor arrays for parallel sensing, the EGFET has the advantages of easier fabrication, less exposure to illumination and temperature changes, easier packaging, and easy exchange of the sensing electrode because the FET is kept outside of the usually wet analyte environment.[16−19]

The FET offers two readout modes, in the linear and saturated regime of the drain–source current IDS versus drain–source voltage VDS curve. In the linear regime, the threshold voltage VT, which is equivalent to the OCP, in the IDS versus Vref curve, is determined for constant measured IDS at constant VDS. Vref is the voltage applied to the reference electrode, biasing the gate in ISFET and EGFET operation. VT can be extracted from the IDS versus Vref curves or directly followed by a constant voltage–constant current feedback circuit, such as a source–drain follower.[20,21] In the saturated regime, VT is proportional to the square root of IDS. The analyte concentration C is then given by the log C dependence on the OCP or VT, according to the Nernst equation with a maximum slope at room temperature of 2.3 kT/e = 59 mV/decade for a single elementary charge process. k is the Boltzmann constant, T is the absolute temperature, and e is the elementary charge. Explicitly, the FET current–voltage dependencies are as follows:

Linear regime

Saturated regime

W is the gate width, L is the gate length, μe is the electron mobility for an n-channel FET, and Cox is the areal gate capacitance. Although extensively discussed in dedicated studies, the OCP and EGFET readout modes including noise were barely compared directly and judged for their superiority.[22,23]

Super-Nernstian InN/InGaN quantum dot (QD) potentiometric biosensors, functionalized for the selective detection of glucose and cholesterol, and anion-selective potentiometric sensors have been reported before.[24−26] Here, we study the EGFET sensor readout in the linear and saturated regime of an InN/InGaN QD electrode with super-Nernstian response grown on a Si(111) substrate and compare it with the direct OCP readout to elucidate the optimum performance. As a “model” analyte for comparison and demonstration, chloride anions are chosen for their physiological importance, easy handling, and absence of complex chemistry. The super-Nernstian response discovered for InN/InGaN QDs has been explained by the high density of intrinsic positively charged surface donors and the zero-dimensional quantum confinement of the InN QDs.[27] Because of the Pauli exclusion principle, not all surface donors are compensated by electrons allowed to enter the QDs. Electrons are partly expelled to the InGaN layer underneath the InN QDs. This electron partition together with the positively charged surface donors creates a strong electric dipole, which has been measured directly by Kelvin probe force microscopy[28] and is in origin of the super-Nernstian response.[29] The high sensitivity is reproduced for the direct OCP and EGFET readout with excellent linearity over a wide concentration range for 10–4 to 0.1 M KCl aqueous solutions, covering the physiologically relevant range in human blood and sweat. The wide linear range emphasizes the importance of planar c-plane InGaN layers underneath the InN QDs. Such planar InGaN layers have been previously achieved only for growth on GaN/sapphire substrates[24−26,30] and now also on the technologically most relevant Si substrates for optimized growth at high rates. For judging and discriminating the different readout modes, special emphasis is on the noise in measurements over extended time. Therefore, the IDS noise δIDS in the EGFET readout is translated to the VT noise δVT in the linear and saturated regime, according to the following:

Linear regime

Saturated regime

This allows the direct comparison with the OCP noise δOCP in the direct OCP readout, considering that the EGFET threshold voltage VT/EGFET is given by

VT/FET is the FET threshold voltage, Eref is the reference electrode potential, ϕ is the sensing electrode surface potential, χSol is the buffer solution surface dipole potential, and Φ/q is the metal gate work function.[9,19,31,32] For the expressions of the VT noise, VDS and Vref, being the FET settings, are assumed noise-free. k is determined from the IDS versus the gate-source voltage VGS curve and IDS versus the VDS curve in the linear and saturated regime of the FET alone without the connected sensing electrode. k = 6.5 10–4 A/V2 with VT/FET = 1.35 V, determined from the IDS versus the VGS curve. The VT noise in the saturated regime is found to be slightly lower than the OCP noise, which is much less than the VT noise in the linear regime. Therefore, the EGFET readout in the saturated regime is highlighted to be the best with the lowest noise, the lowest limit of detection (LOD), and the lowest limit of quantification (LOQ), being important analytical figures of merit.

Results and Discussion

Figure a shows the atomic force microscopy (AFM) height image of the InN/InGaN QDs. In (b, c), the corresponding scanning electron microscopy (SEM) top-view and cross-sectional view of the 1-monolayer (ML)-InN/In0.36Ga0.64N QDs are shown. A three-dimensional AFM image is presented in the Supporting Information Figure S1. The InN QDs exhibit an average diameter of 21 nm, an average height of 6 nm, and a very high area density of ∼1 × 1011 cm–2. The underlying planar InGaN layer is very smooth. The brighter, larger features in (a, b) are due to In droplets due to the slightly metal-rich growth conditions, favoring the planar InGaN layer growth at high growth rates. The repeated shape is visible in AFM and SEM, probably because of the solidification process of the In droplets upon substrate cooling.

Figure 1

(a) AFM height image and SEM (b) top-view and (c) cross-sectional view of the 1-ML-InN/In0.36Ga0.64N QD structure grown on Si(111). The brighter larger features are In droplets. The full height contrast in the AFM image is 52 nm.

The OCP measurements as a function of time during successive immersion of the InN/InGaN QD sensing electrode and Ag/AgCl reference electrode in KCl aqueous solutions with increasing concentrations from 10–7 to 0.1 M are shown in Figure a with a plot of the OCP versus logarithm of the KCl concentration in (b). The inset in (a) demonstrates the reversibility. Most accurate would be a plot versus the logarithm of the chloride anion activity, which, however, approaches the concentration for the values under investigation. The data points are taken 150 s after immersing the electrodes for stabilization. Very good linearity is observed from 0.1 down to 10–4 M KCl concentration with a super-Nernstian slope of −80 mV/decade. The red solid line is a linear fit to the data points. For lower concentrations, the OCP saturates. As mentioned, the wide linear range is attained by the smooth c-plane InGaN layer, now realized on the technologically important Si substrate, being crucial for the formation of InN QDs with the most stable super-Nernstian response.

Figure 2

KCl concentration dependence of (a) OCP as a function of time for the InN/InGaN QD sensing electrode vs Ag/AgCl and (b) OCP vs logarithmic KCl concentration. The error bars refer to three consecutive measurements.

Figure shows the EGFET KCl concentration-dependent output, with (a, b) for evaluation of the linear regime and (c, d) for evaluation of the saturated regime. The IDS versus Vref curves in the linear regime in (a) are measured for a VDS of 0.2 V. The Vref versus logarithm of the KCl concentration plot in (b) is deduced for a constant current–constant voltage setting of an IDS of 0.1 mA and VDS of 0.2 V. The high sensitivity of 82 mV/decade and linearity between 10–4 and 0.1 M, obtained from the OCP measurements, are reproduced. The IDS vs VDS curves in (c) are measured for a Vref of 3 V. In the saturated regime, the plot of the square root of IDS versus logarithm of the KCl concentration in (d) is deduced for both Vref and VDS of 3 V. High sensitivity and good linearity are also confirmed here with a current sensitivity of 0.0487 (mA)1/2/decade between 10–4 and 0.1 M KCl concentration.

Figure 3

KCl concentration dependence of (a) IDS vs Vref curves for VDS = 0.2 V, (b) Vref vs logarithmic KCl concentration for IDS = 0.1 mA and VDS = 0.2 V, (c) IDS vs VDS curves for Vref = 3 V, and (d) square root of IDS vs logarithmic KCl concentration for Vref = 3 V and VDS = 3 V for the InN/InGaN QD EGFET sensor. The error bars refer to three consecutive measurements.

For the analysis of the noise, the OCP and EGFET IDS in the linear and saturated regime are measured over 1 h for various voltage settings with the InN/InGaN QD sensing electrode and Ag/AgCl reference electrode immersed in 0.1 M KCl aqueous solution. The noise measurements are performed after an initial stabilization phase until long-time drift is negligible. Representative measurements are shown in Figure for (a) the OCP, (b) the EGFET IDS in the linear regime, and (c) the EGFET IDS in the saturated regime. In the linear regime, VDS is set to 0.1, 0.2, and 0.4 V and Vref is adjusted such that the measured IDS is around 0.1 mA. In the saturated regime, IDS is measured for a Vref of 2, 3, and 4 V with a constant VDS of 3 V. For all voltage settings, the current noise is comparable in the mid 10–3 mA range. The VT noise or standard deviation δVT is calculated from the measured IDS noise or standard deviation and compared with the OCP noise δOCP. For repeated measurements, the VT noise in the linear regime δVT/lin is larger than the OCP noise, while the VT noise in the saturated regime δVT/sat is slightly smaller. Average values for various current and voltage settings are δVT/sat = 3 mV, δVOCP = 4 mV, and δVT/lin = 20 mV. Therefore, the most precise measurements are obtained for the EGFET output in the saturated regime. A three-standard deviation limit of detection, 3σ LOD, between 10–7 and 10–6 M and a 10σ limit of quantification, 10σ LOQ, around 10–5 M are obtained from the saturation behavior of the output.[33] As the current noise is quite independent of the voltage settings, this is assigned to the larger IDS at larger VDS and Vref in the saturated regime measured with a low-impedance amperemeter, compared to the small IDS in the linear regime where VDS has to be kept small and the rather high OCP. An expensive high-impedance voltmeter is not needed. Our conclusions are made for a sensor setup without special precautions such as a Faraday cage or on-chip integration of the InN/InGaN QD sensing electrode with the FET, which are inconvenient and often not possible to implement.

Figure 4

Time dependence of (a) OCP for the InN/InGaN QD sensing electrode vs Ag/AgCl, (b) IDS around 0.1 mA adjusted by Vref in the linear regime for VDS = 0.1, 0.2, and 0.4 V, and (c) IDS in the saturated regime for Vref = 2, 3, and 4 V and VDS = 3 V for the InN/InGaN QD EGFET sensor for 0.1 M KCl.

Conclusions

In conclusion, we have fabricated an InN/InGaN QD EGFET sensor and compared the readout in the linear and saturated regime with the direct potentiometric OCP readout for the detection of chloride anions in KCl aqueous solutions. The sensor exhibited a super-Nernstian response of −80 mV/decade for the OCP and EGFET readout in the linear regime and 0.0487 (mA)1/2/decade for the EGFET readout in the saturated regime with good linearity for KCl concentrations between 10–4 and 0.1 M. The current noise for the EGFET readout has been translated to the threshold voltage noise δVT for direct comparison with the OCP noise δOCP. For the respective noise, δVT in the saturated regime < δOCP < δVT in the linear regime was obtained. This puts forward EGFET operation in the saturated regime for most precise and convenient measurements.

Experimental Details

The self-assembled InN/InGaN QD structure was grown using the plasma-assisted molecular beam epitaxy (PA-MBE) on a Si(111) substrate.[34] In and Ga were provided by standard Knudsen effusion cells and the active nitrogen was supplied using a radio frequency (RF) plasma source (Oxford H25). The native surface oxide layer on the Si(111) substrate was removed by dipping into 10 wt % HF aqueous solution for one min and rinsing with deionized water. After loading, the Si(111) substrate was degassed at 300 °C for 1 h in the MBE buffer chamber, transferred to the growth chamber, annealed at 900 °C for 10 min for residual surface oxide removal, and exposed to active N flux for 10 min to form a thin SiN layer. InGaN growth was at 550 °C (thermocouple reading) close to the InGaN decomposition temperature under slightly metal-rich conditions with the In and Ga beam equivalent pressure (BEP) of 1.80 × 10–7 and 1.74 × 10–7 Torr, respectively. The active N plasma source settings were 580 W forward power and 0.9 standard cubic centimeters per minute (sccm) molecular N2 flow. The InGaN growth rate was 510 nm/h. For these growth conditions, a compact, planar c-plane InGaN layer forms. The layer thickness was 170 nm. On top of the InGaN layer, one-monolayer InN QDs were deposited without growth interruption. The In content of the InGaN layer of 36% was determined using an X-ray diffractometer (Bruker D8), as shown in the Supporting Information Figure S2. AFM (Bruker Multimode) and top-view and cross-sectional SEM (ZEISS Gemini 500) were performed to observe the surface morphology and layer thickness. High-resolution transmission electron microscopy and detailed optical and electrochemical characterizations of the InN/InGaN QDs have been presented before.[34−36]

To fabricate the sensing electrode, the InN/InGaN QD sample was cleaved into 0.5 × 0.5 cm2 square pieces, attached on a copper foil using In-Ga eutectic on the Si substrate back side to form an ohmic contact and glued on a glass plate. The whole structure was sealed with epoxy resin, leaving an opening on the sample surface for contact with the electrolyte. The OCP was measured using an electrochemical workstation (CHI660E) with a connected InN/InGaN QD working electrode and Ag/AgCl (saturated KCl) reference electrode. For EGFET measurements, the InN/InGaN QD sensing electrode was connected to the gate of a commercial n-channel MOSFET (Texas Instruments CD4007UBE). Vref was applied using the electrochemical workstation. VDS and IDS were applied and recorded using a computer-controlled Keithley 2400 source meter. For both OCP and EGFET measurements, the InN/InGaN QD sensing electrode and Ag/AgCl reference electrode were successively immersed in KCl aqueous electrolyte solutions prepared in separate beakers from low to high concentrations ranging from 10–7 to 0.1 M. Before the measurement, the InN/InGaN QD sensing electrode was conditioned by rinsing in deionized water for 10 min. The schematic drawing of the EGFET measurement setup and circuit is presented in Figure .

Figure 5

Schematic drawing of the EGFET measurement setup and circuit.