Visible-Light-Driven Room Temperature NO 2 Gas Sensor Based on Localized Surface Plasmon Resonance: The Case of Gold Nanoparticle Decorated Zinc Oxide Nanorods (ZnO NRs)

: In this work, nitrogen dioxide (NO 2 ) gas sensors based on zinc oxide nanorods (ZnO NRs) decorated with gold nanoparticles (Au NPs) working under visible-light illumination with different wavelengths at room temperature are presented. The contribution of localized surface plasmon resonant (LSPR) by Au NPs attached to the ZnO NRs is demonstrated. According to our results, the presence of LSPR not only extends the functionality of ZnO NRs towards longer wavelengths (green light) but also increases the response at shorter wavelengths (blue light) by providing new inter-band gap energetic states. Finally, the sensing mechanism based on LSPR Au NPs is proposed.


Introduction
Gas pollutants have become, in recent years, one of the main concerns of people living in densely populated areas. The increasing awareness of their inhabitants about the potential adverse health effects of hazardous gases has encouraged administrations to establish new regulations about air quality in indoor and outdoor spaces. Among them, nitrogen dioxide (NO 2 ) is a pollutant gas produced in many of the combustion processes related to heating, industry, and transportation [1]. Traditionally, metal oxide semiconductor (MOX)based chemi-resistive gas sensors are superior to other signal transduction solutions that are usually based on measuring optical or capacitive properties. This is due to the simplicity of setting up the resistance measurement circuit. The operation of resistive gas sensors is usually based on changes in resistivity or conductivity in semiconductor materials, which are due to a series of chemical steps on the surface, including the adsorption, reaction, and desorption of gas molecules [2]. Many efforts have been invested to make such sensors more sensitive, selective, stable, durable, energy efficient, and safe.
In particular, the introduction of metal nanoparticles into the matrix of a MOX to induce the absorption of light due to the resonance effect on the nanoparticles stands out as one of the most promising approaches because of our ability to tune the absorption band by modifying the surface and shape of the nanoparticles [35]. Moreover, the role of these nanoparticles in the performance of the gas sensors is not limited solely to the absorption of light, but also includes other effects related to the interaction between the metal nanoparticles and the MOX that influence and modify the gas sensor characteristics [36].
In this paper, we present a NO 2 gas sensor based on ZnO nanorods (NRs) decorated with gold nanoparticles (Au NPs) able to work at room temperature. The LSPR properties of Au NPs enhanced the absorption in the visible range (around 600 nm) providing sufficient energy to promote the reaction mechanisms and improving the sensor response compared to the bare ZnO NRs. Even though a similar work has been conducted by Chunxu Chen, et al. [37], our work offers a more in depth description of the mechanisms governing the gas sensor reactions and their performance. Furthermore, a detailed description of the influence of the Au NPs on the characteristics of our sensor is provided. In addition, we also show that the presence of Au NPs not only extends the absorption to the visible band but also improves the response and recovery of gas sensors.

Materials and Sensor Preparation
Gold-interdigitated electrodes (Au-IDE) on glass (MicruX Technology, Asturias, Spain) were used as an electronic platform to measure the electrical characteristics of the ZnO NRs. The IDE size was (10 × 6 × 0.75) mm, with 90 pairs of electrodes separated by 10 µm and 10 µm width. ZnO NRs were synthesized by hydrothermal method as follows: a zinc acetate dehydrate (Zn(OOCH 3 ) 2 ·2H 2 O, Sigma Aldrich) (25 mM) without additional purification dissolved in 10 mL of ethylene glycol was used as a seed layer. The seed layer was deposited onto IDE by a spin coating process. A total of 5 µL of the seed layer solution was drop casted onto the IDE followed by a spin coating process (500 rpm for 10 s and 5500 rpm for 20 s). In order to evaporate the solvent, the samples were heated on a hotplate at 90 • C for a few minutes. This process was repeated 3 times to ensure full coverage of the IDEs. Afterwards, an annealing process at 320 • C for 20 min (5 • C/min) was applied to fix the seed on the IDE and remove all the organic compounds. Zinc oxide nanorods (ZnO NRs) were grown by a hydrothermal method: 30 mM of Zn acetate dehydrate (Zn(OOCH 3 ) 2 ·2H 2 O) and 30 mM of hexamethylenetetramine (HTM) (C 6 H 12 N 4 ) was dissolved in 25 mL of Millipore water as a growth solution. The sample was immersed in the growth solution for 3-5 h at 90 • C. Then, thermal treatment at 350 • C in the air for 1 h was applied to remove the residual organic compounds and crystallize the sample. This procedure is shown in Figure 1a.
Gold colloidal (20 nm, concentration:~0.01% HAuCl 4 ,~1 A 520 units/mL) were purchased from Sigma-Aldrich and used without any additional treatment. An electrophoresis method was applied to attach the Au NPs onto the surface of the ZnO NRs. A previously fabricated ZnO NR sample (ZnO NRs on IDE) was immersed in the Au NP gold suspension for 300 s, in which a constant voltage (3V) between the IDE and electrode was simultaneously applied. The ZnO NRs that covered half of the IDE is shown in Figure 1b to simplify the view of IDE, ZnO NRs and the expected result of the Au NP decoration. electrophoresis method was applied to attach the Au NPs onto the surface of the ZnO NRs. A previously fabricated ZnO NR sample (ZnO NRs on IDE) was immersed in the Au NP gold suspension for 300 s, in which a constant voltage (3V) between the IDE and electrode was simultaneously applied. The ZnO NRs that covered half of the IDE is shown in Figure 1b to simplify the view of IDE, ZnO NRs and the expected result of the Au NP decoration.

Materials Characterization
The morphology and elemental composition of as-grown bare ZnO NRs and decorated Au NPs were studied using JSM-7100F thermal field emission scanning electron microscope (FE-SEM) and energy dispersive spectroscopy (EDS). To know the 3D surface profile, an atomic force microscope (AFM) analysis was conducted by Veeco Dimension 3100. The optical properties were measured using Specord ® PC 205 ultraviolet-visible (UV-vis) spectrophotometers, Analytik Jena AG, Jena, Germany and photoluminescence (PL) spectra at room temperature. Figure 2 is a schematic of the measurement system for light-activated gas sensing response. The experiments of gas measurement were performed in a customized chamber of 200 mL in volume. The gas flow was maintained steady at 200 mL/min during all the experiments. The gas chamber was equipped with sealed feedthrough electrical connections to drive the light-emitting diodes (LEDs) and to measure the electrical resistance of the ZnO NRs, using a Keithley 2400 source measure unit (SMU). The LED driving and resistance measurements were carried out under constant current conditions. The desired atmospheric conditions were provided by independent mass flow controllers blending between synthetic air (SA), carbon monoxide (100 ppm of CO in SA), carbon dioxide (500 ppm of CO2 in SA), and nitrogen dioxide (10 ppm of NO2 in SA) at convenience. To investigate the optoelectronic properties of ZnO NRs and ZnO NRs/Au NPs, conductance

Materials Characterization
The morphology and elemental composition of as-grown bare ZnO NRs and decorated Au NPs were studied using JSM-7100F thermal field emission scanning electron microscope (FE-SEM) and energy dispersive spectroscopy (EDS). To know the 3D surface profile, an atomic force microscope (AFM) analysis was conducted by Veeco Dimension 3100. The optical properties were measured using Specord ® PC 205 ultraviolet-visible (UV-vis) spectrophotometers, Analytik Jena AG, Jena, Germany and photoluminescence (PL) spectra at room temperature. Figure 2 is a schematic of the measurement system for light-activated gas sensing response. The experiments of gas measurement were performed in a customized chamber of 200 mL in volume. The gas flow was maintained steady at 200 mL/min during all the experiments. The gas chamber was equipped with sealed feedthrough electrical connections to drive the light-emitting diodes (LEDs) and to measure the electrical resistance of the ZnO NRs, using a Keithley 2400 source measure unit (SMU). The LED driving and resistance measurements were carried out under constant current conditions. The desired atmospheric conditions were provided by independent mass flow controllers blending between synthetic air (SA), carbon monoxide (100 ppm of CO in SA), carbon dioxide (500 ppm of CO 2 in SA), and nitrogen dioxide (10 ppm of NO 2 in SA) at convenience. To investigate the optoelectronic properties of ZnO NRs and ZnO NRs/Au NPs, conductance measurements were performed under synthetic airflow with different LED wavelengths of visible light irradiation (i.e., blue (465 nm), green (520 nm), yellow (590 nm) and red (640 nm)) and in dark conditions. In this regard, during the experiment, the intensity of LEDs applied was 10 mW/cm 2 . In that regard, our group has recently developed a micro light platform where an LED is embedded beneath an IDE with power consumptions in the microwatt (µW) range (i.e., 30 µW to detect 250 ppb of NO 2 with ZnO NPs) [13]. In order to determine the sensitivity of the ZnO NR sensors towards particular gases, different concentrations of NO 2 from the low to the high concentrations (i.e., 0.5, 1, 2, 3, 5, and 10 ppm) were then exposed to the chamber under LED irradiation. The response (sensitivity) of NO 2 was calculated from the relation as follows:

Gas Measurement
where R NO 2 and R air are the electrical resistance of the sensor when exposed to NO 2 gas and in air, respectively. The definition of the response time and recovery time is the time needed to reach 90% of the total resistance change during the chemisorption (adsorption and desorption) process. Furthermore, all experiments were conducted at room temperature.
of visible light irradiation (i.e., blue (465 nm), green (520 nm), yellow (590 nm) and red (640 nm)) and in dark conditions. In this regard, during the experiment, the intensity of LEDs applied was 10 mW/cm 2 . In that regard, our group has recently developed a micro light platform where an LED is embedded beneath an IDE with power consumptions in the microwatt (µW) range (i.e., 30 µW to detect 250 ppb of NO2 with ZnO NPs) [13]. In order to determine the sensitivity of the ZnO NR sensors towards particular gases, different concentrations of NO2 from the low to the high concentrations (i.e., 0.5, 1, 2, 3, 5, and 10 ppm) were then exposed to the chamber under LED irradiation. The response (sensitivity) of NO2 was calculated from the relation as follows: where and are the electrical resistance of the sensor when exposed to NO2 gas and in air, respectively. The definition of the response time and recovery time is the time needed to reach 90% of the total resistance change during the chemisorption (adsorption and desorption) process. Furthermore, all experiments were conducted at room temperature.

Active Material Characteristic
Energy dispersive X-ray spectroscopy (EDS) was applied to measure the chemical composition of the produced active materials. A typical EDS spectrum in Figure 3a confirms the existence of zinc and oxygen, revealing the successful formation of ZnO. In addition, the successful formation of ZnO decorated Au NPs is confirmed by the other peak of gold, which is observed in Figure 3b. Figure 4a,b shows the field emission scanning electron microscopy (FE-SEM) image of bare ZnO NRs and ZnO NRs/Au NPs with different magnifications. ZnO NRs presented a hexagonal shape due to the preferential growth along the <0001> direction [38] of the wurtzite crystal structure (JCPDS 01-075-0576). All samples displayed a uniform NRs distribution and density throughout the sample with diameters ranging from 150-180 nm. After the electrophoresis process, Au NPs were uniformly and completely attached to the surface of the ZnO NRs. The length of the NRs was measured by atomic

Active Material Characteristic
Energy dispersive X-ray spectroscopy (EDS) was applied to measure the chemical composition of the produced active materials. A typical EDS spectrum in Figure 3a confirms the existence of zinc and oxygen, revealing the successful formation of ZnO. In addition, the successful formation of ZnO decorated Au NPs is confirmed by the other peak of gold, which is observed in Figure 3b. Figure 4a,b shows the field emission scanning electron microscopy (FE-SEM) image of bare ZnO NRs and ZnO NRs/Au NPs with different magnifications. ZnO NRs presented a hexagonal shape due to the preferential growth along the <0001> direction [38] of the wurtzite crystal structure (JCPDS 01-075-0576). All samples displayed a uniform NRs distribution and density throughout the sample with diameters ranging from 150-180 nm. After the electrophoresis process, Au NPs were uniformly and completely attached to the surface of the ZnO NRs. The length of the NRs was measured by atomic force microscopy (AFM). A region with faulty growth was used to measure a profile line resulting in a height of around 1 µm (Figure 4d).    Figure 5a shows the UV-vis diffuse reflectance spectra (UV-vis DRS) of ZnO NRs and ZnO NRs/Au NPs where the contribution of Au NPs is clearly evidenced by an absorption band of around 550 nm corresponding to the LSPR gold band [19,25]. According to the result, the absorption spectrum broadened to the visible light range around 500-600 nm. In this particular case, there is a chance to generate the energy of resonance when the    Figure 5a shows the UV-vis diffuse reflectance spectra (UV-vis DRS) of ZnO NRs and ZnO NRs/Au NPs where the contribution of Au NPs is clearly evidenced by an absorption band of around 550 nm corresponding to the LSPR gold band [19,25]. According to the result, the absorption spectrum broadened to the visible light range around 500-600 nm. In this particular case, there is a chance to generate the energy of resonance when the  Figure 5a shows the UV-vis diffuse reflectance spectra (UV-vis DRS) of ZnO NRs and ZnO NRs/Au NPs where the contribution of Au NPs is clearly evidenced by an absorption band of around 550 nm corresponding to the LSPR gold band [19,25]. According to the result, the absorption spectrum broadened to the visible light range around 500-600 nm. In this particular case, there is a chance to generate the energy of resonance when the frequency of the photon of incident light fits the coherent oscillation frequency of conduction electrons confined at the Au NPs surface [6].
NRs because these photons are absorbed by the Au NPs throughout the surface plasmon resonance whose absorption wavelength is typically around 550-600 nm. The intensity and the ratio of the NBE to DL emission vary obviously with the different deposition times of Au NPs [34]. Furthermore, these ZnO NRs were decorated with Au NPs for 300 s. The enhancement of the SPR achievement is affected by the light emission and surface plasmon resonance of the Au NPs energy. Moreover, the shapes and sizes of nanoparticles closely affect the frequency of the SPR. [35].

Sensor Measurement
In order to know the response of the sensor to photo-activation, an ON/OFF mode was then conducted without gas under LED illumination with different wavelengths. To investigate the sensitivity of the ZnO NRs sensors towards oxidizing gases (i.e., NO2), different concentrations of nitrogen dioxide (i.e., 0.5, 1, 2, 3, 5, and 10 ppm) were then applied to the chamber under LED illumination with different wavelengths. Before introducing the target gases, a 2 h stabilization step with SA in the chamber was conducted to achieve a reference baseline. Then, 10 ppm of NO2 was introduced to the chamber for 30 min and then removed by introducing again SA until the next NO2 level concentration. The same procedure was conducted for different wavelengths of LED illumination during the experiment. Figure 7, by referring to Figure 6b, shows the dynamic responses of ZnO NRs/Au NPs to different concentrations of NO2 under blue, green, yellow, and red LED illumination. As expected, exposing ZnO (n-type) to an oxidizing gas (NO2)   [24,39]. The sensors of ZnO NRs/Au NPs show an increment in emission compared to that of bare ZnO NRs, which is likely due to the enhanced light trapping within the structures. The DL emission from ZnO NRs/Au NPs is significantly higher than the bare ZnO NRs. According to the result, this effect might be due to new deep-level traps states because of the presence of the Au NPs. However, other groups [25,35] found a completely opposite trend when decorating ZnO NRs with gold nanoparticles, i.e., the emission band associated with the trap states are quenched compared to the bare ZnO NRs because these photons are absorbed by the Au NPs throughout the surface plasmon resonance whose absorption wavelength is typically around 550-600 nm. The intensity and the ratio of the NBE to DL emission vary obviously with the different deposition times of Au NPs [34]. Furthermore, these ZnO NRs were decorated with Au NPs for 300 s. The enhancement of the SPR achievement is affected by the light emission and surface plasmon resonance of the Au NPs energy. Moreover, the shapes and sizes of nanoparticles closely affect the frequency of the SPR. [35].

Sensor Measurement
In order to know the response of the sensor to photo-activation, an ON/OFF mode was then conducted without gas under LED illumination with different wavelengths. Figure 6a,b shows the responses for the bare ZnO NRs and ZnO NRs/Au NPs under LED illumination with different wavelengths. The contribution of the LSPR absorption band is evidenced by the photoexcitation under 590 and 640 nm LED illumination.
To investigate the sensitivity of the ZnO NRs sensors towards oxidizing gases (i.e., NO 2 ), different concentrations of nitrogen dioxide (i.e., 0.5, 1, 2, 3, 5, and 10 ppm) were then applied to the chamber under LED illumination with different wavelengths. Before introducing the target gases, a 2 h stabilization step with SA in the chamber was conducted to achieve a reference baseline. Then, 10 ppm of NO 2 was introduced to the chamber for 30 min and then removed by introducing again SA until the next NO 2 level concentration. The same procedure was conducted for different wavelengths of LED illumination during the experiment. Figure 7, by referring to Figure 6b, shows the dynamic responses of ZnO NRs/Au NPs to different concentrations of NO 2 under blue, green, yellow, and red LED illumination. As expected, exposing ZnO (n-type) to an oxidizing gas (NO 2 ) leads to an increase in the resistance of the MOX with a response proportional to the gas concentration. Upon light illumination and depending on its energy, the sensor behavior is completely different. For example, in dark conditions, the resistance of the sample begins to rise once the NO 2 interacts with the surface of the ZnO NRs. After ceasing the exposure to NO 2 , the resistance does not recover to its original value, meaning that there is not enough energy to release the adsorbed NO 2 onto the surface of ZnO NRs (remember, samples are at room temperature). This behavior is reproduced with red and yellow illumination, where the energy of the photons is not sufficient to promote the desorption of the NO 2 . Only when green and blue light are used, the characteristic adsorption/desorption process can be observed, including the saturation regime after enough time. Since the adsorption/desorption process emerged for the blue and green LED, we calculated the sensitivity, response time, and recovery time for them. Figure 8 shows more detailed response differences between bare the ZnO NRs and ZnO NRs/Au-NPs. The sensitivity value of the sensors using blue and green LED are 891% and 191% at 10 ppm of NO 2 , respectively, while their response time for the blue and green LED is 18 min and 21 min, respectively.
Chemosensors 2022, 10, x FOR PEER REVIEW 7 of 1 leads to an increase in the resistance of the MOX with a response proportional to the ga concentration. Upon light illumination and depending on its energy, the sensor behavio is completely different. For example, in dark conditions, the resistance of the sample be gins to rise once the NO2 interacts with the surface of the ZnO NRs. After ceasing th exposure to NO2, the resistance does not recover to its original value, meaning that ther is not enough energy to release the adsorbed NO2 onto the surface of ZnO NRs (remembe samples are at room temperature). This behavior is reproduced with red and yellow illu mination, where the energy of the photons is not sufficient to promote the desorption o the NO2. Only when green and blue light are used, the characteristic adsorption/desorp tion process can be observed, including the saturation regime after enough time. Since th adsorption/desorption process emerged for the blue and green LED, we calculated th sensitivity, response time, and recovery time for them. Figure 8 shows more detailed re sponse differences between bare the ZnO NRs and ZnO NRs/Au-NPs. The sensitivit value of the sensors using blue and green LED are 891% and 191% at 10 ppm of NO respectively, while their response time for the blue and green LED is 18 min and 21 min respectively.  To appreciate the role of the Au NPs on the performance of the gas sensor, a detailed performance comparison between the bare ZnO NRs and the ZnO NRs/Au NPs were conducted at different wavelengths ( Figure 8). In both cases, we have the same trend described previously. However, the presence of Au NPs has, at least, three different contributions to the performance of the gas sensor: (1) to foster the adsorption of NO2 onto the  To appreciate the role of the Au NPs on the performance of the gas sensor, a detailed performance comparison between the bare ZnO NRs and the ZnO NRs/Au NPs were conducted at different wavelengths ( Figure 8). In both cases, we have the same trend described previously. However, the presence of Au NPs has, at least, three different contributions to the performance of the gas sensor: (1) to foster the adsorption of NO 2 onto the surface, both in dark and under light conditions; (2) to increase the response towards NO 2 under green and blue illuminations; and (3) to improve the desorption rate of NO 2 under green and blue illumination. Therefore, the contribution of the Au nanoparticles extends beyond the introduction of a new absorption band (around 600 nm). Cross-sensitivity experiments towards CO and CO2, which are common gases in the ambient air, were conducted with the aim of determining the sensing selectivity of ZnO NR/Au NPs. A typical experiment consisted of introducing, first, a certain amount of NO2 (10 ppm) followed by a certain concentration of either CO (100 ppm) or CO2 (500 ppm), and finally a mixture of both gases with the same previous concentrations. Figure 9a,b depicts the transient experiment of such experiment for CO (a) and CO2 (b). When the interfering gas is introduced into the chamber alone, the response of the sensor is practically null in the case of CO. However, when introducing CO2, a decrease in the resistance of the sensor is clearly appreciable even though the baseline is partially shadowing this effect. The observed reduction of the resistance is consistent with the fact that CO2 is a reducing gas. Finally, when both gases, NO2 and the interfering one, were blended into the chamber, no cross-interferences were evident in the case of CO, but a clear dampening of the response can be observed in the case of CO2. This cross-interfering phenomenon is the result of the competition between an oxidizing gas (NO2), which leads to an increase in resistance, and a reducing gas (CO2) with the opposite effect, a decrease in the resistance.
Moreover, another parameter that can influence the sensor performance, in photoactivated and low-power consumption applications, is the intensity of the LED illumination. Figure 9c shows the responses of the ZnO NRs/Au NPs sensor with different intensities of blue LED. By applying three different intensities (i.e., 1, 5, and 10 mW/cm 2 ) to 10 ppm of NO2, the response of the sensor was obviously different. Even though we cannot conclude that a lower intensity has a higher sensitivity, since there is no baseline of the resistance for each intensity, we can assure that the intensity of the LED had a clear effect on the shape of the response signal. Note that the higher the intensity, the faster the satu- Cross-sensitivity experiments towards CO and CO 2 , which are common gases in the ambient air, were conducted with the aim of determining the sensing selectivity of ZnO NR/Au NPs. A typical experiment consisted of introducing, first, a certain amount of NO 2 (10 ppm) followed by a certain concentration of either CO (100 ppm) or CO 2 (500 ppm), and finally a mixture of both gases with the same previous concentrations. Figure 9a,b depicts the transient experiment of such experiment for CO (a) and CO 2 (b). When the interfering gas is introduced into the chamber alone, the response of the sensor is practically null in the case of CO. However, when introducing CO 2 , a decrease in the resistance of the sensor is clearly appreciable even though the baseline is partially shadowing this effect. The observed reduction of the resistance is consistent with the fact that CO 2 is a reducing gas. Finally, when both gases, NO 2 and the interfering one, were blended into the chamber, no cross-interferences were evident in the case of CO, but a clear dampening of the response can be observed in the case of CO 2 . This cross-interfering phenomenon is the result of the competition between an oxidizing gas (NO 2 ), which leads to an increase in resistance, and a reducing gas (CO 2 ) with the opposite effect, a decrease in the resistance.
Moreover, another parameter that can influence the sensor performance, in photoactivated and low-power consumption applications, is the intensity of the LED illumination. Figure 9c shows the responses of the ZnO NRs/Au NPs sensor with different intensities of blue LED. By applying three different intensities (i.e., 1, 5, and 10 mW/cm 2 ) to 10 ppm of NO 2 , the response of the sensor was obviously different. Even though we cannot conclude that a lower intensity has a higher sensitivity, since there is no baseline of the resistance for each intensity, we can assure that the intensity of the LED had a clear effect on the shape of the response signal. Note that the higher the intensity, the faster the saturation regime is reached where the absorption and desorption processes achieve an equilibrium. Therefore, this kind of experiment is very important to estimate the minimum energy or power needed with the aim of minimizing power consumption.

Light-Activated and Sensing Mechanism
According to the UV-vis and PL spectra, the defect emission of ZnO NRs was attributed to the electron transition from a deep-level trap to the valence band of ZnO [3,5,18]. Due to LSPR, the absorption spectrum of Au NPs overlaps with the defect-related emission bands of ZnO NRs in the visible range, and it is impossible to directly couple with the NBE of ZnO NRs. In other words, resonant energy coupling must occur between the defect emission of ZnO NRs and LSPR. In addition, Au NPs can absorb the visible light photon energy emitted by ZnO NRs, which excites the electrons of the Au NPs to a higher energy state and then transfers them to DL emission through the CB of ZnO NRs. The energy transfer is in control of the change of defect energy level emission. With the support of the resonance energy coupling process, electrons with higher energy can be well transferred to the conduction band of ZnO NRs, resulting in enhanced NBE and DL emissions of ZnO NRs, as shown in the schematic diagram in Figure 10. Particularly, the energy emitted by photonic ZnO NRs DL can be resonantly absorbed by Au NPs to generate LSPR, which excites the electrons of Au NPs to a higher energy state and then transfers them to the conduction band of ZnO NRs. According to the schematic diagram, the enhancement of the emission intensity of NBE and DL is related to the following factors: The comparison of the sensor response in this work and the previously published reports is shown in Table 1. The contribution of Au NPs decoration on ZnO-based gas sensors indicates a high performance under particular conditions, especially under visiblelight-driven conditions at room temperature. The increased absorption after Au NPs decoration leads to the generation of the energetic charge on the surface of the sensors that enhances the sensitivity of the sensors.

Light-Activated and Sensing Mechanism
According to the UV-vis and PL spectra, the defect emission of ZnO NRs was attributed to the electron transition from a deep-level trap to the valence band of ZnO [3,5,18]. Due to LSPR, the absorption spectrum of Au NPs overlaps with the defect-related emission bands of ZnO NRs in the visible range, and it is impossible to directly couple with the NBE of ZnO NRs. In other words, resonant energy coupling must occur between the defect emission of ZnO NRs and LSPR. In addition, Au NPs can absorb the visible light photon energy emitted by ZnO NRs, which excites the electrons of the Au NPs to a higher energy state and then transfers them to DL emission through the CB of ZnO NRs. The energy transfer is in control of the change of defect energy level emission. With the support of the resonance energy coupling process, electrons with higher energy can be well transferred to the conduction band of ZnO NRs, resulting in enhanced NBE and DL emissions of ZnO NRs, as shown in the schematic diagram in Figure 10. Particularly, the energy emitted by photonic ZnO NRs DL can be resonantly absorbed by Au NPs to generate LSPR, which excites the electrons of Au NPs to a higher energy state and then transfers them to the conduction band of ZnO NRs. According to the schematic diagram, the enhancement of the emission intensity of NBE and DL is related to the following factors: (1) the coupling efficiency indicated by the degree of overlap between the emission spectrum of ZnO NRs defect and the absorption spectrum of Au NPs; (2) the coupling intensity indicates the absorption of Au NPs; and (3) the local field caused by the hot-spots effect the enhancement is affected by the morphology and spatial density of Au NPs on the surface of ZnO NRs [34].  Generally, the working principle of metal oxide-based gas sensors depends on the chemically adsorbed (adsorption and desorption) oxygen molecules on the surface, which are ionized into substances, such as O , O and O , by capturing electrons near the surfaces of the metal oxides. Moreover, it is known that O , O and O ionsorption species dominate at <150 °C, between 150 and 400 °C, and at >400 °C, respectively [43]. When ZnO NRs are exposed to the air at room temperature in dark conditions, some oxygen molecules are absorbed by the surface and then trap free electrons from the conduction band of ZnO NRs O ( ) + e ↔ O ( ) [26]. Correspondingly, the electron carrier density in ZnO NRs decreases, forming an electron depletion layer to compensate for the negative charge of surface oxygen species (see Figure 11a) [44]. When the ZnO NRs are exposed to vis-light, electron-hole pairs will be generated, and then the holes can migrate to the surface of ZnO NRs and be neutralized by pre-chemisorb oxygen ion h + O ( ) ↔ O ( ) [29]. At the same time, photo-generated electrons can combine with oxygen molecules in the air to form additional photo-generated oxygen ions, as in the following equation (see Figure 11b): Consequently, the electrical resistance of ZnO NRs decreases in this reaction. The adsorption and desorption of oxygen molecules on the surface of ZnO NRs can gradually reach a stable equilibrium state, and the carrier density of ZnO NRs under vis-light illumination is still much higher than that in dark [45]. A similar reaction also occurs in ZnO NRs/Au NPs since there is a contribution of electron transfer from Au NPs due to LSPR (Figure 11d). Therefore, a large photocurrent was observed under the vis-light irradiation with the energetic electron (hot electron) and more density of charge (Figure 11e). The increment of energetic charge on the surface leads to more sensitivity of the sensors. In the opposite case, the density of free electrons reduces quickly after illumination of the vis-light is turned off, which leads to a decrease in the photocurrent and an increase in the electrical resistance (see Figure 6). All the chemisorption processes affect the width change of the EDL. Generally, the working principle of metal oxide-based gas sensors depends on the chemically adsorbed (adsorption and desorption) oxygen molecules on the surface, which are ionized into substances, such as O − 2 , O − and O 2− , by capturing electrons near the surfaces of the metal oxides. Moreover, it is known that O − 2 , O − and O 2− ionsorption species dominate at <150 • C, between 150 and 400 • C, and at >400 • C, respectively [43]. When ZnO NRs are exposed to the air at room temperature in dark conditions, some oxygen molecules are absorbed by the surface and then trap free electrons from the conduction band of ZnO NRs (O 2 (g) + e − ↔ O − 2(ads) ) [26]. Correspondingly, the electron carrier density in ZnO NRs decreases, forming an electron depletion layer to compensate for the negative charge of surface oxygen species (see Figure 11a) [44]. When the ZnO NRs are exposed to vis-light, electron-hole pairs will be generated, and then the holes can migrate to the surface of ZnO NRs and be neutralized by pre-chemisorb oxygen ion (h + + O − 2 (ads) ↔ O 2(g) ) [29]. At the same time, photo-generated electrons can combine with oxygen molecules in the air to form additional photo-generated oxygen ions, as in the following equation (see Figure 11b): Consequently, the electrical resistance of ZnO NRs decreases in this reaction. The adsorption and desorption of oxygen molecules on the surface of ZnO NRs can gradually reach a stable equilibrium state, and the carrier density of ZnO NRs under vis-light illumination is still much higher than that in dark [45]. A similar reaction also occurs in ZnO NRs/Au NPs since there is a contribution of electron transfer from Au NPs due to LSPR (Figure 11d). Therefore, a large photocurrent was observed under the vis-light irradiation with the energetic electron (hot electron) and more density of charge (Figure 11e). The increment of energetic charge on the surface leads to more sensitivity of the sensors. In the opposite case, the density of free electrons reduces quickly after illumination of the vis-light is turned off, which leads to a decrease in the photocurrent and an increase in the electrical resistance (see Figure 6). All the chemisorption processes affect the width change of the EDL.  Figure 11 shows a schematic of the proposed sensing mechanisms when the sensor is in dark conditions (before and after Au NPs decoration) and exposed to the target gas (NO2) under visible-light irradiation together with the energy band diagram. When the chemisorbed oxygen molecules reach an equilibrium state, the new electron depletion layer is formed under the surface of Au NPs (Figure 11e). Once the sensors are exposed to NO2 gas, as the electrons extracted from the conduction band or the remaining photogenerated electrons are transferred to the surface and increase the resistance of the sensor, a stronger chemical adsorption species O ( ) (ℎ ) is formed, and the depletion layer becomes thicker [7] as shown in Figure 11c,f. These reactions can be simply described as follows:  Figure 11 shows a schematic of the proposed sensing mechanisms when the sensor is in dark conditions (before and after Au NPs decoration) and exposed to the target gas (NO 2 ) under visible-light irradiation together with the energy band diagram. When the chemisorbed oxygen molecules reach an equilibrium state, the new electron depletion layer is formed under the surface of Au NPs (Figure 11e). Once the sensors are exposed to NO 2 gas, as the electrons extracted from the conduction band or the remaining photogenerated electrons are transferred to the surface and increase the resistance of the sensor, a stronger chemical adsorption species O − 2(ads) (hν) is formed, and the depletion layer becomes thicker [7] as shown in Figure 11c,f. These reactions can be simply described as follows: NO 2(g) + O − 2(ads) (hν) ↔ NO − 2(ads) (hν) + O 2(g) NO 2(g) + e − (hν) ↔ NO − 2(ads) (hν) (4) NO 2(g) + O − 2(ads) ↔ NO − 2(ads) + O 2(g) (5) NO 2(g) + e − ↔ NO − 2(ads) The species NO − 2(ads) (hν) is feebly bound to the surface and can be simply removed since the electron affinity of NO 2 (2.27 eV) is higher than that of O 2 (0.44 eV) [7]. When NO 2 gas is removed from the chamber, the photo-generated holes recombine with the electron captured by the chemisorbed NO − 2(ads) (hν) and O − 2(ads) (hν) recover the surface of nanoparticles via reaction (2), which returns the depletion layer to the initial state that corresponds to the decreased electrical resistance of the sensors. Therefore, under the actual experiment, the detection of NO 2 is controlled by the competition between oxygen molecules (O 2(g) ), pre-chemisorbed oxygen ion (O − 2(ads) ), photo-absorbed oxygen ion (O − 2(ads) (hν)) and nitrogen dioxide (NO − 2(ads) (hν)) for the absorption process (reaction 2-6) and desorption process (reaction 7) as follows: Thus, visible light acts as a regulator of the competition between these reaction pathways [1]. These mechanisms quantitatively predict the sensitivity of the sensors S(%) to different concentrations of nitrogen dioxide (NO 2 ) on the surface of ZnO NRs observed experimentally under different light intensities [27]. The oxygen molecules play an important role in the chemisorption (adsorption and desorption) process of NO 2 molecules on the ZnO NRs surface. The coverage degrees of O − 2(ads) (hν) directly influences performances (i.e., sensitivity, response, and recovery time) of photo-activated metal oxide gas sensors.

Conclusions
A visible-light-driven room temperature NO 2 gas sensor based on localized surface plasmon resonance (LSPR) was successfully fabricated by the electrophoretic deposition of gold nanoparticles (Au NPs) on zinc oxide nanorods (ZnO NRs) grown by a hydrothermal process. According to UV-vis and photoluminescence (PL) spectra, Au NPs decoration has a significant contribution to the improvement of the sensitivity of the sensor due to increasing energetic charge (hot electron) on the surface and LSPR phenomenon. The sensitivity of the sensor reached up to around 46 times of 891% for blue LED at 10 ppm of NO 2 after Au NPs decoration. Finally, a detailed description of the phenomenon governing the reaction mechanisms and the performance of the sensors were provided, proving that the role of the Au NPs extends beyond the mere extension of the light absorption to the visible region due to the LSPR phenomenon, by fostering the adsorption of NO 2 , and improving the desorption rate and increasing the response of NO 2 under green and blue illumination conditions.
The results reported that LSPR could potentially lead to the development of goodperformance chemi-resistive gas sensors operating at room temperature, which are energyefficient, suitable, and safer to use in a hazardous environment. Moreover, one aspect that should be noted is that another factor that greatly influences the performance of the sensor in photo-activated, low power consumption, and RT working temperature regime, is light intensity. This will be an interesting challenge for sensor technology in the future.