1. Fabrication

Fig. 2. Sensing mechanism of NO$_{2}$ gas sensor based on AlGaN/GaN HEMT.

Fig. 1 shows cross-sectional diagram (a) and microscope image of fabricated sensors (b). AlGaN/GaN HEMT-type sensors were fabricated at the Inter-University Semiconductor Research Center (ISRC), Seoul, Korea. AlGaN/GaN-on-Si substrate consists of a 4 nm GaN cap layer, a 10-nm Al$_{\mathrm{0.2}}$Ga$_{\mathrm{0.8}}$N barrier layer, a 5.2 μm i-GaN layer, and AlGaN/AlN buffer layers. The source and drain contacts with Ti/Al/Ni/Au (200/1200/ 250/500\AA{}) were formed by e-beam evaporation with lift-off process and followed by rapid thermal annealing (RTA) at 833 $^{\mathrm{o}}$C for 32 s in N$_{2}$ ambient. 315 nm-depth mesa isolation was formed by inductively coupled plasma (ICP) etching with BCl$_{3}$/Cl$_{2}$ to define the active region. The sensing area as a gate electrode was then formed by e-beam evaporation and lift-off process. We fabricated 2 samples with the same thickness (30 nm) of Pt or Pd layers. Finally, the probing pads were formed with Ti/Au (20/250 nm) evaporation.

2. Measurement

Gas sources consist of synthesized dry air as the reference gas, and 100 ppm NO$_{2}$ as target gas. For different concentrations of NO$_{2}$, we mixed target gas and dry air using mass flow controllers (MFCs). The sensors were loaded in a chamber containing a hot chuck to control the operating temperature. I-V and transient characteristics were measured using Agilent 4155A parameter analyzer. We maintained a constant flow rate of 200 sccm in all measurements.

3. Sensing Mechanism

The nitrogen dioxide sensing mechanism of sensors based on AlGaN/GaN HEMTs has been investigated and reported in ^(8,9,11,12). When nitrogen dioxide gas is introduced, nitrogen dioxide molecules are dissociated on the catalytic layer (Pt or Pd) and result in negatively charged oxygen ions that diffuse through the gate via pores or grain boundary and reach the surface of AlGaN layer. Here they affect the number of mobile carriers in two-dimensional electron gas (2DEG) of the HEMT structure, it leads to a reduction of drain current (Fig. 2). While the adsorption of oxygen ions can take place at room temperature, the desorption occurs only at high temperatures ⁽¹²⁾, which directly affects the recovery of the sensor. %Su su NguyenSsN2020-04-05T12:20:00Z Su su NguyenSsN Su su NguyenSsN Su su Nguyen Su su Nguyen SsN SsN 2020-04-05T12:20:00Z 2020-04-05T12:20:00Z Reviewer 2, question 1: explanation of how the oxygen ions are removed from the channel Reviewer 2, question 1: explanation of how the oxygen ions are removed from the channel

Fig. 3. The transient characteristics of (a) Pt-AlGaN/GaN, (b) Pd-AlGaN/GaN sensors at 300 $^{0}$C under 100 ppm of NO$_{2}$.

The sensitivity is defined as the ratio between the change of drain current and initial current,

S = (I$_{\mathrm{dry air}}$ - I$_{\mathrm{NO2}}$) / I$_{\mathrm{dry air}}$

where I$_{\mathrm{dry air}}$ and I$_{\mathrm{NO2}}$ are drain currents under the flow of dry air and NO$_{2}$ gas, respectively. The response and recovery times were calculated from transient characteristics, where they show 10% to 90% of the total change in drain current.

1. Performance Comparison Between Pt-sensor and Pd-sensor

Fig. 4. The transfer characteristic (a), response of Pd-sensor at different gate voltages V$_{\mathrm{G}}$ = -0.1; 0; 0.1; 0.3 V (b, c, d, e respectively) at 300 $^{\mathrm{o}}$C under 100 ppm NO$_{2}$.

Fig. 3 shows the transient characteristics of Pt-AlGaN/GaN and Pd-AlGaN/GaN sensors with alternately exposed to dry air and 100 ppm NO$_{2}$. At 300~$^{\mathrm{o}}$C, the sensitivity of Pt-sensor is 15%, and it took about 15 minutes to recover. The Pd-sensor demonstrated much better performance: 53 % of sensitivity and recovery time is only 196 s. These results show the advantage of Pd catalyst over Pt catalyst in NO$_{2}$ detection.

2. High Performance NO2 Gas Sensor Based on Pd-AlGaN/GaN HEMT

A. Modulation of Gate Bias

One of the advantages of sensors based on HEMT compared to Schottky diode is the ability to optimize sensitivity by adjusting gate bias ⁽¹¹⁾. Fig. 4 demonstrates the performance of Pd-sensor at different gate biases, from -0.1 to 0.3 V. The sensor exhibits high sensitivity when it is biased near the threshold voltage (Fig. 5). At negative gate bias V$_{\mathrm{G}}$=-0.1 V, the sensor shows high sensitivity, but the current level was too low and unstable. When increasing V$_{\mathrm{G}}$, initial current, the change of drain current ΔI and leakage current I$_{\mathrm{G}}$ also increased, leading to the reduction of sensitivity. This is in agreement with the explanation of Bishop et al ⁽¹¹⁾. Based on this result, we choose V$_{\mathrm{G}}$ = 0 V as the operating point in the next measurements.

Fig. 5. The sensitivity of Pd-sensor at different gate voltages at 300 $^{\mathrm{o}}$C under 100 ppm NO$_{2}$.

Fig. 6. Pd-sensor’s response at high temperatures under 100 ppm of NO$_{2}$ in 30 s exposure time and 90 s recovery time.

C. Performance of Pd-sensor at Different Concen-trations of NO₂ NO²

Fig. 7 shows our Pd-sensor can detect NO$_{2}$ in a range from 10 to 100 ppm at 300 $^{\mathrm{o}}$C. Even at a low concentration of 10 ppm, the sensitivity is 18 %, which is much improved when compared to other studies using Pt-AlGaN/GaN HEMTs (Table 2). Our sensor consumes about 0.23 mW in this regime, this is a great advantage when integrating sensors into circuits.

Table 1. The performance of sensor at high temperatures

Temperature (℃)	Sensitivity (%)	Response time (s)	Recovery time (s)
200 ℃	26.4	24	too long
300 ℃	40.3	22	11
350 ℃	35.2	14	9
400 ℃	19.8	3	3

Fig. 7. Pd-sensor’s response under 10 ppm, 50 ppm and 100 ppm of NO$_{2}$.

In real working conditions, the sensors should respond quickly to different gas concentrations with small exposure and recovery times. Fig. 8 shows significant changes of drain current when the Pd-sensor was exposed in NO$_{2}$ in 30 s under 10 ppm to 100 ppm of NO$_{2}$. This measurement suggested that our sensor can work in real time conditions.

Fig. 8. Pd-sensor’s response under different concentrations in 30 s exposure time and 90 s recovery time.