Photoactivated In₂O₃-GaN Gas Sensors for Monitoring NO₂ with High Sensitivity and Ultralow Operating Power at Room Temperature

Abstract

Photoactivated gallium nitride (GaN) nanowire-based gas sensors, functionalized with either bare In₂O₃ or In₂O₃ coated with a nanolayer of evaporated Au (Au/In₂O₃), were designed and fabricated for high-sensitivity sensing of NO₂ and low-power operation. The sensors were tested at room temperature under 265 nm and 365 nm ultraviolet illumination at several power levels and in relative humidity ranging from over 20% to 80%. Under all conditions, photoconductivity was lower in the Au/In₂O₃-functionalized sensors compared to that of sensors functionalized with bare In₂O₃. However, when tested in the presence of NO₂, the Au/In₂O₃ sensors consistently outperformed In₂O₃ sensors, the measured sensitivity being greater at 265 nm compared to 365 nm. The results show significant power reduction (×12) when photoactivating at (265 nm, 5 mW) compared to (365 nm, 60 mW). Maximum sensitivities of 27% and 42% were demonstrated with the Au/In₂O₃ sensors under illumination at (265 nm, 5 mW) for 1 ppm and 10 ppm concentration, respectively.

1. Introduction

With the rise of interest in environmental protection, health concerns, air quality monitoring, toxic gasses and industrial waste detection and control, the development of gas sensors has been gaining increasing interest over the last several decades [1]. Among the various sensor technologies, those based on semiconducting metal oxides have important advantages, such as a simple principle of operation, low cost, and portability, and have been studied extensively [2]. Because their operation critically depends on the reaction of the analyte with the sensor surface, improvements have been achieved by forming bare metal oxide nanowires, nanoribbons, hollow porous structures, etc., which increase the specific surface area. Further improvements have been attained by combining bare metal oxide with suitable metal and semiconducting materials and by exploiting the catalytic and heterojunction properties therein. Functionalizing the sensing layer with noble metals, such as Au, Pt, Pd, or Ag, can improve sensitivity and tailor selectivity [3,4,5]. This route can be a promising solution to solve the inherent selectivity issues of semiconductor gas sensors. To achieve the required performance, however, these sensors need to operate at elevated temperatures, which increase their operating power. The recent rapid development of the Internet of Things (IoT) necessitates the availability of compact and portable gas sensors for integration into personal smart devices such as phones, watches, etc., which must, therefore, be designed for low power operation. The need for sensors which operate at room temperature has, therefore, become apparent and led to the emergence of “photoactivated” gas sensors—i.e., sensors where the energy needed to activate and promote analyte/sensor interaction is supplied by illumination with the light of appropriate wavelength and irradiance. Interest in photoactivated sensors is increasing rapidly, and several excellent reviews have recently been published [6,7,8]. According to these reviews and the references therein, most published works assume that the light-generated electron–hole pairs react with the gas molecules and facilitate their chemisorption and desorption. This simple view, however, cannot explain all the experimental results, where the best outcome is obtained, at times, with wavelengths close to the energy gap and other times, with wavelengths much shorter than that, and where, sometimes, higher irradiance leads to worse performance. These inconsistencies have since been resolved by Geng et al. [9], who proposed and proved that, as well as generating electron–hole pairs, light also interacts and excites the analyte molecules.

Due to the harmful effect of NO₂ on human health and the environment, there is a great deal of interest in NO₂ gas sensors, and their state-of-the-art characteristics have been thoroughly discussed in a comprehensive review by M. Setka et al. [10]. ZnO is among the most popular oxides for NO₂ gas sensors and has extensively been used to fabricate NO₂ gas sensors with promising results [11]. Excellent NO₂ gas sensors have also been made from bare In₂O₃, which is characterized by a wide band gap (direct/indirect 3.5 eV/2.8 eV), low resistivity, and high chemical stability [12]. Zhang et al. [13] developed a coral-like In₂O₃ sensor able to achieve sensitivity of 132% with an optimal operating temperature of 130 °C. Similarly, Xiao et al. [14] achieved a very high sensitivity of 350% to 1 ppm NO₂ with sensors made from In₂O₃ nanospheres, which, however, needed to be operated at a temperature of 120 °C. As mentioned above, an especially interesting family of gas sensors is based on nanostructures such as nanowires (NWs), nanotubes, and nanorods, which have seen significant progress in recent years [15,16,17,18,19,20] due to their large surface-to-volume ratio which considerably increases sensitivity and improves the time response of the analyte gases. Gallium nitride (GaN) nanostructure-based gas sensors have recently become very popular due to their unique physical features such as direct bandgap, excellent carrier mobility, high heat capacity, and high breakdown voltage [21,22]. These properties make GaN an excellent candidate for portable gas sensors. Furthermore, the surface engineering of GaN nanostructures significantly improves sensor performance. Recently, Shi et al. [23] developed a high-performance photoactivated GaN nanowire (GaN NW) sensor functionalized with TiO₂, which exhibits a 25% response to 500 ppm NO₂ concentration under 365 nm UV illumination. Khan et al. [24] developed multiple sensors based on oxide/GaN NW combinations with various metal oxides, such as ZnO, WO₃, and SnO₂, for NO₂ sensing. An optimum response of 7.1% for 10 ppm NO₂ was achieved using ZnO oxide at room temperature under 365 nm UV illumination.

The present study combines photoactivation with the advantages of In₂O₃ and GaN nanowires to design and fabricate NO₂ gas sensors operating at room temperature with excellent sensitivity and low-power operation. The oxide layer was either bare In₂O₃ or coated with an evaporated Au nanolayer. Two UV LEDs (265 nm and 365 nm) and several power levels were investigated to achieve the optimal solution for lowering the operating power of the finished sensor without sacrificing the analyte gas sensitivity. The sensors were tested at two different NO₂ concentrations (1 ppm and 10 ppm) and over a range of relative humidity (RH) levels, from 20% to 80% at room temperature.

2. Materials and Methods

2.1. Device Fabrication & Characterization

The sensors were fabricated following steps similar to those by Khan et al. [24]. An AlGaN buffer layer was first deposited on the Si substrate to minimize lattice mismatch and improve adhesion between Si and GaN NW. Then, the GaN NWs were patterned by stepper-lithography-assisted dry etching, followed by induced coupled plasma etching, using a metal hard mask to protect the GaN NW. The nanowire width target ranges between 200 and 400 nm. Subsequently, electrodes composed of a Ti/Al/Ti/Au metal stack were deposited on top of the GaN layer to form ohmic contacts. Following this, RF magnetron sputtering was used to deposit a thin layer of In₂O₃ (in the range of 5–10 nm) on top of the exposed GaN NW, which, for a subset of sensors, was followed by the deposition of 1 nm evaporated Au nanolayer on top of the In₂O₃ layer. For brevity, these two groups of sensors will be referred to as In₂O₃ and Au/In₂O₃, respectively, in the figure below. Finally, rapid thermal annealing (RTA) at 700 °C was performed to crystalize the receptor layers and improve ohmic contacts. The devices were characterized by Zeiss Auriga-Field Emission Scanning Electron Microscope (SEM) with energy dispersive spectroscopy (SEM w/EDS). Figure 1 shows schematics and SEM micrographs of the finished based sensors. Several sensors from each group were wire-bonded and mounted onto an array board chamber for testing.

EDS analysis was performed to characterize the composition of In₂O₃ and Au/In₂O₃ sensors, as observed in Figure 2a,b.

Figure 2a,b both reveal the presence of In, N, O, Ga, Al, and Si with peaks located at 0.365 kV, 0.392 kV, 0.525 kV, 1.098 kV, and 1.486 kV, respectively. The presence of a small peak of Au at 2.1 kV is observed in the inset of Figure 2b. A corresponding quantity of 0.11 measured in weight % confirms the deposition of a small layer of Au on the Au/In₂O₃ sensors. Details of EDS spectra values quantified in atomic % and weight % for Ga, N, and Au elements are listed in

Table 1. EDS weight and atomic ratio analysis for In₂O₃ and Au/In₂O₃ sensors.

	In2O3			Au/In2O3
Element	Weight %	Atomic %	Error %	Weight %	Atomic %	Error %
N	31.84	69.93	8.27	33.73	71.73	8.28
Ga	68.16	30.07	2.47	66.15	28.26	2.52
Au	0	0	3.4	0.11	0.02	3.52

.

2.2. Gas Sensor Measurement

During testing, the devices were biased at 5V DC. UV LEDs were mounted within the chamber 10 mm above the sensors, providing constant illumination for the duration of testing. Two UV light wavelengths were consecutively used: 265 nm and 365 nm. Both 265 nm and 365 nm were selected as their photon energies (4.6 eV and 3.4 eV, respectively) are both larger than the GaN bandgap (3.4 eV), a factor that is necessary for electron–hole pair generation. For each wavelength, testing was performed at three power levels: 5 mW, 30 mW, and 60 mW. Optical power was measured using a Network-Power Meter-Model 1928-C. Photoconductivity measurements, humidity testing, and NO₂ gas testing (1 ppm and 10 ppm concentration) under various humidity conditions were also conducted. Resistance responses were collected using an Arduino Mega 2560 controlled by NI-LabView software. A gaseous mixture of analyte and carrier gas (breathing air), controlled by mass flow controllers (MFC), was first introduced into a Bronkhorst CEM Evaporator W-101A humidifier prior to flowing into the chamber containing the sensors. The gas mixture was maintained at a constant flow rate of 0.6 slpm and a constant pressure of 5 psi. Figure 3 illustrates the experiment setup for sensor testing.

3. Results and Discussions

3.1. Photoconductivity

Photoconductivity and humidity testing of the sensors were performed under two illumination conditions: 265 nm at 5 mW (3.3 µW/cm² irradiance) and 365 nm at 60 mW (1462 µW/cm² irradiance). Figure 4 shows the I–V characteristics of both In₂O₃ and Au/In₂O₃ sensors in the dark and under illumination.

In the dark condition, current levels are rather low for both sensor groups, similar to those observed by Shi et al. [23]. Electron–hole pairs are generated under UV illumination. The In₂O₃ layer absorbs a small part of the UV light to generate photocarriers, as it is quite thin. Consequently, the density of electrons in the GaN conducting channel increases, thereby explaining the improvement of conductivity between the dark and illuminated conditions observed in Figure 4. Chemisorption of environmental O₂ at the sensors surface can also explain the difference in conductivity seen in Figure 4, as described by the following equation [10]:

$${\mathrm{O}}_2+e^{-}={\mathrm{O}}_2^{-}\left(ads\right)$$

(1)

Oxygen anions will deplete the electrons from the conduction channel of GaN. The amount of generated oxygen anions is small, while the amount of photogenerated electrons in the conducting channel of GaN is large and dominates the conductivity. It should be noted that there are limited active sites for O₂ adsorption. Functionalizing In₂O₃ with Au provides additional active sites for gas adsorption [25]. More O₂ molecules can be adsorbed at the Au interface, thus further depleting electrons from the channel. Consequently, the conductivity of Au/In₂O₃ sensors is lower than that of bare In₂O₃, as observed in Figure 4. The highest current levels were observed in bare In₂O₃ sensors, 11.2 µA and 9.6 µA for 265 nm and 365 nm, respectively, while the current of Au/In₂O₃ sensors diminished (6.9 µA and 7.2 µA). The inset in Figure 4 shows the UV response gain, defined as:

$$U{V}_{Gain}=\frac{I_{UV}-I_{dark}}{I_{dark}}$$

(2)

where I_UV represents the sensor current under UV illumination and I_dark the current under the dark condition. It is interesting to note that the obtained UV_Gain (250%) under low power illumination at (265 nm, 5 mW) is comparable to that obtained under much higher power illumination at (365 nm, 60 mW). Photoconductivity results indicate that using the 265 nm LED makes it possible for the sensor to be operated at 12 times less power than when using the 365 nm LED, without sacrificing any UV_Gain. Power reduction is an essential aspect when designing portable devices that rely on batteries.

3.2. Humidity Response

The performance of the sensors under the illuminated condition was evaluated for relative humidity (RH) in the range from 20% RH to 80% RH at room temperature (20 °C). The “Humidity Response” HR was defined by the relationship:

$$HR=\frac{R_{actual}-R_{20\%}}{R_{20\%}}$$

(3)

where R_actual is the resistance at the measured humidity, and R_20% represents the baseline resistance at 20% RH. Each sensor was sequentially tested for 900 s at RH of 20%, 40%, 60%, and 80%. Figure 5a,b show the results of the HR experiments.

At low humidity (<40% RH), all devices displayed a relatively low response to water adsorption under both illumination conditions, with the response deviating less than 5% from the baseline. Mostly, this small change in resistivity is due to chemosorbed water molecules with the metal oxide layer and photogenerated carriers [26]. At medium-to-high humidity (40–80% RH), resistance levels started to increase. Water layers accumulated on top of the chemosorbed layer, promoting proton (H⁺) hopping [26]. Protons transfer from one molecule to the next by the exchange of a hydrogen bond which generates a change in conductivity. At elevated humidity (80% RH), more water layers accrued, providing additional paths for proton hopping and causing a drastic increase in device resistance. Moreover, water can condensate due to the capillary effect and infiltrate the nanopores, offering further conduction channel paths. This is known as the Grotthuss mechanism [27]. In addition, OH⁻ can be generated from H₂O under UV irradiation. The OH⁻ will form hydroxyl bond OH on the surface and deplete the electrons of n-type GaN, leading to an increase in resistance [28]. Au/In₂O₃ sensors produced a more pronounced response at elevated humidity compared to bare In₂O₃ sensors. As seen in Figure 5b, the humidity response of Au/In₂O₃ sensors sharply increased to 34%, whereas a relatively low response of 7% is observed for In₂O₃ under 265-nm illumination. Owing to its high catalytic proprieties, Au facilitates the dissociation of water molecules, and these separated species contribute to the increase in resistance. Interestingly, bare In₂O₃ devices illuminated under 365 nm did not respond to the change in humidity. An important observation from Figure 5 is that sensors illuminated at 265 nm showed a greater response to humidity increase compared to those illuminated at 365 nm. The LED power level may be playing an important role here on the adsorption/desorption rate, as well as on water molecules dissociation. Since (365 nm, 60 mW) irradiates at higher power, higher numbers of photogenerated carriers are available to react with dissociated water, accelerating its desorption rate and, hence, leading to stable response compared to (265 nm, 5 mW). Robust and stable sensors are more desired for environmental monitoring, as outdoor humidity often fluctuates.

3.3. NO₂ Response under Various LED Power and Wavelengths

After photoconductivity and humidity testing, the evaluation of NO₂ detection was performed. In the presence of NO₂, three possible surface reactions can occur, leading to a change in resistivity, measured by the sensitivity S:

$$S=\frac{R_{N{O}_2}-R_{air}}{R_{air}}$$

(4)

where $R_{N{O}_2}$ is the sensor’s resistance in the presence of NO₂, and R_air represents the baseline resistance under ambient air environment. In the first reaction, NO₂ interacts with the pre-adsorbed oxygen ion forming a NO₂⁻ ion and an oxygen gas molecule. Secondly, two molecules of NO₂ interact with the pre-adsorbed oxygen ion and photogenerated electron at the surface, producing two NO₃⁻ ions. Finally, in the third reaction, NO₂ directly reacts with a photogenerated electron at the surface, producing NO₂⁻ [6,7,8,9,10]. Figure 6a,b show the transient response of the Au/In₂O₃ sensors for 1 ppm NO₂ detection at the fixed humidity level of 40% and under three LED power levels (5 mW, 30 mW, 60 mW).

An increase in resistance upon exposure to NO₂ is observed, confirming the oxidation type of NO₂ gas. NO₂ and the pre-adsorbed oxygen attract electrons at the surface, leading to an increase in the depletion layer width, causing a large increase in the sensor resistance. Figure 6c,d show NO₂ sensitivity at different LED power levels, along with the respective irradiances measured at the sensor surface. For the same power level, sensitivity is lower at a higher wavelength. For example, at 30 mW, the sensitivity was 23% under 265 nm and 15% at 365 nm. The reason behind this remains a complicated question which, despite many previous efforts to answer it, still remains open [6,29]. A notable observation is that the highest sensitivity (29%) was obtained at the lowest power (5 mW) for both wavelengths and that sensitivity degrades with increasing power. For example, sensitivity is reduced from 29% to 20% when the 265 nm LED power is raised from 5 mW (corresponding to an irradiance of 1.1 µW/cm²) to 60 mW (600 µW/cm²). A similar trend is observed for 365 nm illumination. This behavior agrees with Zhao’s work [30].

Although sensitivity increases with the decreasing power of UV LED, Figure 7c,d show that response and recovery times decrease upon exposure to 1 ppm NO₂ with increasing power for both wavelengths. For instance, response time under 365 nm illumination decreased from 233.3 s at 5 mW to 159.6 s at 60 mW power. Interestingly, recovery time drastically decreases at higher LED power for 265 nm: from 572.9 s to 257.4 s. However, sensors under 365 nm illumination have a slow recovery time (i.e., more than 600 s). It is not completely understood why this is, but it can perhaps be explained by considering the newly proposed role of light-activated NO₂ [9] and absorption depth of light where the electrons/holes are generated in the semiconductor. Our devices Au/In₂O₃ demonstrate good short-term repeatability over 6 cycles of exposure to 1 ppm NO₂, as observed in Figure 7a,b. However, a slight degradation in sensor sensitivity is observable after 2 months, as evidenced by Figure S1a. This could be caused by the oxidation of our sensors. Finally, our Au/In₂O₃ sensors have excellent selectivity toward NO₂ compared to other gases, such as SO₂, H₂S, or HCN. Note that the Au/In₂O₃ did not respond to CO nor NH₃, as seen in Figure S1b. Our results indicate that LED power can be optimized for different applications needs.

3.4. NO₂ Response at Various Relative Humidity Levels

Figure 8 shows the sensitivity to 1 ppm and 10 ppm NO₂ detection at various relative humidity levels. In this experiment, two LED conditions were used: (265 nm, 5 mW) and (365 nm, 60 mW).

It was consistently observed that the Au/In₂O₃ sensors outperformed the bare In₂O₃ sensors. At 1 ppm, bare In₂O₃ sensors suffered from a low response rate (<5%). The gas sensing enhancement of Au/In₂O₃ sensors can be explained by the following two mechanisms. Firstly, the depletion layer modification caused by the presence of Au (work function 5.1 eV compared to 4.3 eV for GaN) causes electrons from the In₂O₃ to migrate toward the Au nanolayer and react with NO₂ [31,32,33,34,35,36,37]. Secondly, due to its high catalytic proprieties, Au allows for an easier dissociation of NO₂ and water molecules, known as spillover effect [38]. The separated species transfer toward the In₂O₃ layer, thereby altering the sensor resistance due to the concentration increase in the Au-induced active adsorption sites. For instance, at (1 ppm, 40% RH, 265 nm), Au/In₂O₃ exhibited a sensitivity of 27%, while the In₂O₃ sensors only showed 2% sensitivity. At (10 ppm, 40% RH, 265 nm), response of the In₂O₃ sensors was 16%, which increased to 33% for the Au/In₂O₃ sensors. Maximum sensitivities of 32% and 42% were achieved with the Au/In₂O₃ sensors under illumination at (265 nm, 5 mW) for 1 ppm and 10 ppm concentration, respectively. Remarkably, the NO₂ response signals did not degrade with increasing humidity. At higher concentrations, sensitivity increased linearly with increasing humidity. One possible explanation might be that water molecules introduce an additional electron cloud that is favorable to NO₂ adsorption due to its electrophilic nature.

Table 2. Comparison of the NO₂ sensitivity of the sensors presented in this study with those reported in the literature.

Sensing Material	Concentrations (ppm)	λ (nm)	Operating Temp. (°C)	Power Consump. (mW)	Sensitivity (%)	Reference
DNA-CTMA/GaN	10	264, 364	RT	-	10	[39]
g-C3N4/GaN NR	1	392	RT	-	10	[40]
AlGaN/GaN HMET	1	-	300	200	1.1	[41]
G/GaN	10	265–350	RT	-	8	[42]
In2O3/Zn nanofibers	1, 10	-	50	-	2.38, 130	[43]
Au-loaded Porous In2O3	1	-	100	-	900	[34]
Au/In2O3 GaN NW	1, 10	265	RT	5	27, 42	Present Study

provides a comparison between the present work and the state of the art for NO₂ sensing. For instance, Redeppa et al. investigated the functionalizing of a GaN thin film layer with deoxyribonucleic acid (DNA-CTMA/GaN). The sensor’s response was enhanced with UV illumination at 254 nm and 364 nm, similar to our work. At 10 ppm, the resulting sensitivity was around 10% for both wavelengths, a percentage which is lower compared to our sensor’s response of 42% [39]. Recently, NO₂ sensing has also been demonstrated using a heterostructure of graphitic carbon nitride on GaN nanorods (g-C₃N₄/GaN NR) illuminated at 392 nm at room temperature. Despite having a faster response/recovery time and lower limit of detection (500 ppb) than our sensors, the g-C₃N₄/GaN NR sensor’s response at 1 ppm was 10% lower than that of our Au/In₂O₃ sensor, which had a 27% response rate [40]. Sun et al. reported a response of 1.1% to a 1 ppm concentration of an AlGaN/GaN HMET sensor with a built-in microheater [41]. The microheater produces 200 mW, which is about 40 times more than the minimum LED power required for the present sensors. Shin et al. [42] demonstrated that a GaN NW sensor functionalized with graphene (G/GaN NW) could reach a response of 8% to 10 ppm NO₂ under 40% RH using LED with wavelengths ranging from 265–350 nm, whereas, under the same conditions, the present sensors exhibited a response of 37%. Recently, several studies have reported excellent sensitivity to NO₂ using In₂O₃ gas sensors with different morphologies. Ueda [34] reported an Au-loaded porous In₂O₃ sensitivity of 900% to 1 ppm NO₂, which is far greater than our sensor’s response. Chen et al. [43] demonstrated that an In₂O₃/Zn nanofiber sensor could achieve a 130% sensitivity to 10 ppm. However, those In₂O₃-based sensors operated at a relatively high temperature (100 °C), which increases power consumption. Our sensors achieved reasonable sensitivity with a low power consumption.

4. Conclusions

In summary, GaN NW-based sensors functionalized with In₂O₃ and Au/In₂O₃ sensors were presented for NO₂ sensing. Photoconductivity results exhibited lower current levels in Au/In₂O₃-functionalized GaN NW sensors compared to In₂O₃-functionalized ones. However, the Au/In₂O₃ sensors consistently demonstrated a superior sensitivity to NO₂ than the In₂O₃ sensors at all tested humidity conditions. The highest sensitivities of 27% and 42% were achieved using the Au/In₂O₃ sensors under low-power illumination at (265 nm, 5 mW) for 1 ppm and 10 ppm concentrations, respectively. More importantly, the results show the achievement of significant power reduction (×12) when using (265 nm, 5 mW) UV illumination rather than (365 nm, 60 mW), without sacrificing sensitivity. This combination of power reduction and high NO₂ sensitivity makes our Au/In₂O₃-GaN NW sensors an excellent candidate for portable and smart sensor integration.