High-Performance MEMS Oxygen Sensors Based on Au/TiO 2 Films

: High-performance microelectromechanical system (MEMS) oxygen sensors were realized by successful preparation of Au nanoﬁlms over TiO 2 thin ﬁlms through successive sputtering on commercial MEMS microhotplates. Oxygen sensing performance of 3 and 6 nm thick Au over TiO 2 thin ﬁlms were compared with that of pure TiO 2 thin ﬁlms. It was shown that 6 nm thick Au over TiO 2 thin ﬁlms have the best sensitivity toward oxygen. The prepared TiO 2 thin ﬁlms were characterized using SEM, EDS, XPS, and a gas testing instrument. The results show that Au decoration has little inﬂuence on the surface morphologies of TiO 2 thin ﬁlms. However, Au decoration has a strong inﬂuence on the surface properties of the composite ﬁlms. The favorable performance of 6 nm Au-doped TiO 2 thin ﬁlms is attributed to factors such as catalytical performance, height of Schottky contact, and number of oxygen vacancies. This work makes contributions to low power consumption and high-performance oxygen sensors for Internet of Things applications.


Introduction
Oxygen gas is necessary for breathing. A shortage of oxygen can cause headache or even death. Thus, monitoring of concentration of oxygen gas using gas sensors is critical to human health, especially after emergency due to fire or explosion. Metal oxide semiconductor gas sensors have been widely applied to detect relevant gases such as CH 4 , H 2 , CO, or O 2 considering their high sensitivity, excellent stability, and low cost [1][2][3]. Various oxides like ZrO 2 , CeO 2 , and TiO 2 have been studied for oxygen sensing [4][5][6][7][8][9]. TiO 2 is one of the most important metal oxides for oxygen sensing due to its easy availability, low cost, chemical inertness, and excellent photocatalytic properties [10]. In 2000, TiO 2−x films were synthesized using the sol-gel and hydrogen reduction methods [8]. However, the working temperature of the prepared TiO 2−x film for oxygen sensing is as high as 800 • C. High temperatures result in high power consumption and possible danger of gas combustion in settings such as coal mines with high methane content. TiO 2 thin films were synthesized using the atomic layer deposition (ALD) method and the sensor worked at temperatures below 300 • C in 2019 [11]. UV-assisted oxygen sensing of TiO 2 has recently been studied, and the working temperature can be lowered down to room temperature [12,13]. Additionally, doped TiO 2 has been further studied for oxygen sensing. For example, TiO 2 nanoparticles were added to VO x nanoflakes using a thermal decomposition-assisted hydrothermal method. The composites show excellent temperature-independent oxygen sensing properties, which is a result of mixed valent states in VO x [14]. N-doped TiO 2 has much lower resistance and higher sensitivity compared to undoped TiO 2 , which may come from the thermionic emission conduction mechanism or tunneling current effect under completely depleted grain conditions of TiO 2 [15].
It is worth noting that most reported TiO 2 oxygen sensors are built on Al 2 O 3 substrates of relatively large size. Furthermore, Al 2 O 3 substrates normally lead to high power consumption when working at temperatures higher than ambient temperature. Silicon substrate-based micro TiO 2 oxygen sensors are favored compared to Al 2 O 3 substrate due to their smaller sizes, lower power consumption, and easier integration with integrated circuits. For example, a TiO 2 /Pd/TiO 2 layer was manufactured on a microhotplate by combining spin coating of TiO 2 and sputtering of a Pd layer. The sensor has a working range of concentrations between 0 and 20% oxygen. However, the power consumption is estimated to be 960 mW, which is quite high [16]. TiO 2 nanorod arrays were integrated on microelectrodes on silicon through an acid vapor oxidation process on a patterned strap of Ti/Pd films. The TiO 2 nanorod arrays can respond to oxygen at room temperature, though the base resistance is larger than 10 MΩ, resulting in circuit integration difficulties [17]. Furthermore, it has been reported that Au nanoparticles can activate oxygen at the Au/TiO 2 interface due to the catalytic effect of gold nanoparticles [18,19]. Au-TiO 2 composites have been reported to be sensitive to ammonia, acetaldehyde, acetone, and formaldehyde [20][21][22]. Nowadays, commercial microhotplates can consume around 30 mW when working at 400 • C. If high-performance oxygen-sensitive material could be combined with microhotplates, the sensors would be very good candidates for oxygen measurement in the environment or even for Internet of Things applications. Up to now, there are no reports on the integration of Au/TiO 2 films on microhotplates for oxygen sensors. The microhotplate-based oxygen sensors may be a good solution for sensor nodes of Internet of Things applications.
Sputtering is a very efficient method for depositing metal or metal oxide thin films over substrate. It can form a dense thin film with little cracking on various substrates. Additionally, it can be carried out at low temperatures in batches [16]. In this work, Au/TiO 2 films were successfully synthesized using the sputtering method on MEMS microhotplates. The sensing performance regarding oxygen was evaluated. The sensing mechanism was also studied in detail.

Film Deposition and Characterization
HHC1000 MEMS microhotplates were brought from HMNST (Hefei, China). Before sputtering, bare microhotplates were bonded to a ceramic package using Au wires for conduction. The Au and TiO 2 targets were 99.99% pure and used for sputtering without further treatment. The process of sputtering is illustrated in Figure 1. TiO 2 and Au films were deposited on the microhotplates in sequence to function as oxygen-sensitive films. A radio frequency table sputtering system (Hefei Jusheng Co., Hefei, China) was employed for deposition of TiO 2 or Au film. For deposition of TiO 2 film, the flow rate of Ar was 35.4 sccm, the power of RF source was 40 W, and the temperature of the stage during deposition was 400 • C. The thickness of the film was controlled by deposition time. The deposition time of TiO 2 was 2 h and finally 300 nm of TiO 2 could be obtained. For deposition of Au film, the flow rate of Ar was 35.4 sccm, the power of RF source was 50 W, and the deposition stage was at room temperature. The deposition time of Au was 15 or 30 s in order to obtain an ultrathin non-continuous Au layer over the TiO 2 layer. The thickness of the Au films was measured at 3 or 6 nm for 15 or 30 s processing. The final samples obtained were named TiO2-300 nm, TiO2-300 nm-Au-3 nm, TiO2-300 nm-Au-6 nm. The thickness of the thin films was measured by a Dektak XT Profilometer (Bruker, Billerica, MA, USA). The morphologies and EDS information of the films were characterized by field emission scanning electron microscopy (TESCAN MAIA3, Brno,  The final samples obtained were named TiO 2 -300 nm, TiO 2 -300 nm-Au-3 nm, TiO 2 -300 nm-Au-6 nm. The thickness of the thin films was measured by a Dektak XT Profilometer (Bruker, Billerica, MA, USA). The morphologies and EDS information of the films were characterized by field emission scanning electron microscopy (TESCAN MAIA3, Brno, Czech Republic). The surface chemical properties of the films were checked by an X-ray Photoelectron Spectrometer (Thermo Fisher ESCALAB 250Xi, Waltham, MA, USA).

Device Measurement
The sensing performance of the fabricated sensors was investigated in a lab-made testing system (HGS9010, designed by Hefei Micro/Nano Sensing Technology Co. Ltd., Hefei, China), which consists of a potentiometric arrangement, a testing gas chamber with a volume of 1 L, and a mixing fan. During measurement, the MEMS-based sensor was held in the gas chamber and heated using a current source (GWINSTEK, GPD-3303S, New Taipei City 236, Taiwan). The measurement was carried out in dynamic mode. The total flow rate was 500 sccm. The pure oxygen and nitrogen gas were mixed at different ratios to realize different oxygen concentrations. The resistances of the MEMS-based sensors under various temperatures and atmospheres were recorded using the testing system. To avoid the effect of the residual oxygen gas, the tests on fabricated MEMS-based sensors were conducted in the order of low to high concentrations. The sensitivity (S) of the gas sensor was calculated by S = R g /R 0 , where R g is the resistance in atmosphere of mixed oxygen and nitrogen, and R 0 is the resistance in pure nitrogen, respectively. Response time and recovery time were defined according to the reported work.

Results
A SEM image of the whole sensor is shown in Figure 2A. The length of the squareshaped microhotplate chip is 1.0 mm. The integrated electrodes area in the middle of the square microhotplate chip is 0.1 × 0.1 mm 2 . As seen in Figure 2B-D, there is no obvious difference in morphologies between bare TiO 2 layer and TiO 2 layers with Au decoration. This is because the thickness of the Au layer is so thin that the Au layer has little effect on the surface morphologies of TiO 2 layer. The final samples obtained were named TiO2-300 nm, TiO2-300 nm-Au-3 nm, TiO2-300 nm-Au-6 nm. The thickness of the thin films was measured by a Dektak XT Profilometer (Bruker, Billerica, MA, USA). The morphologies and EDS information of the films were characterized by field emission scanning electron microscopy (TESCAN MAIA3, Brno, Czech Republic). The surface chemical properties of the films were checked by an X-ray Photoelectron Spectrometer (Thermo Fisher ESCALAB 250Xi, Waltham, MA, USA).

Device Measurement
The sensing performance of the fabricated sensors was investigated in a lab-made testing system (HGS9010, designed by Hefei Micro/Nano Sensing Technology Co. Ltd., Hefei, China), which consists of a potentiometric arrangement, a testing gas chamber with a volume of 1 L, and a mixing fan. During measurement, the MEMS-based sensor was held in the gas chamber and heated using a current source (GWINSTEK, GPD-3303S, New Taipei City 236, Taiwan). The measurement was carried out in dynamic mode. The total flow rate was 500 sccm. The pure oxygen and nitrogen gas were mixed at different ratios to realize different oxygen concentrations. The resistances of the MEMS-based sensors under various temperatures and atmospheres were recorded using the testing system. To avoid the effect of the residual oxygen gas, the tests on fabricated MEMS-based sensors were conducted in the order of low to high concentrations. The sensitivity (S) of the gas sensor was calculated by S = Rg/R0, where Rg is the resistance in atmosphere of mixed oxygen and nitrogen, and R0 is the resistance in pure nitrogen, respectively. Response time and recovery time were defined according to the reported work.

Results
A SEM image of the whole sensor is shown in Figure 2A. The length of the squareshaped microhotplate chip is 1.0 mm. The integrated electrodes area in the middle of the square microhotplate chip is 0.1 × 0.1 mm 2 . As seen in Figure 2B-D, there is no obvious difference in morphologies between bare TiO2 layer and TiO2 layers with Au decoration. This is because the thickness of the Au layer is so thin that the Au layer has little effect on the surface morphologies of TiO2 layer. In Figure 3A, there is no appearance of a Au element peak for pure TiO 2 thin films. The Pt peak comes from Pt electrodes underneath the TiO 2 thin films. The Au peak appears in the TiO 2 -300 nm-Au-3 nm sample and contributes to 3.95% of the total weight of the sample, as shown in Figure 3B. As seen in Figure 3C, content of the Au element increases to 12.88% as the thickness of Au becomes 6 nm. The data from EDS element analysis proves that the Au element was introduced over the surface of the TiO 2 successfully.
In Figure 3A, there is no appearance of a Au element peak for pure TiO2 thin films. The Pt peak comes from Pt electrodes underneath the TiO2 thin films. The Au peak appears in the TiO2-300 nm-Au-3 nm sample and contributes to 3.95% of the total weight of the sample, as shown in Figure 3B. As seen in Figure 3C, content of the Au element increases to 12.88% as the thickness of Au becomes 6 nm. The data from EDS element analysis proves that the Au element was introduced over the surface of the TiO2 successfully. As seen in Figure 4A, pure TiO2 shows a rather strong Ti element signal peak (459.34 and 464.84 eV) in the XPS diagram. Even with decoration of only a 3 nm Au layer over the TiO2 layer, the signal of the TiO2 element disappears considering that XPS is a surfacesensitive characterization technique. As seen in Figure 4B, no Au element signal was detected in the pure TiO2 sample. On the contrary, the Au signal (84.15 and 87.79 eV) appears in both Au-TiO2 samples and becomes higher for the TiO2 composite film with the thicker Au layer. As seen in Figure 4C, the as-prepared three samples have very different signals for oxygen species. The signal at 530.54 eV belongs to lattice oxygen, while 531.69 eV belongs to vacancy oxygen and 532.34 eV belongs to OH oxygen. Pure TiO2 thin film has two oxygen species, lattice oxygen and OH oxygen, which originate from TiO2 and surface adsorption of water molecules in the air. TiO2-300 nm-Au 3 nm shows no distinct lattice oxygen peak, which is the result of coverage of the surface Au layer on the surface of the TiO2. The oxygen vacancy's peak appears for this sample while it does not appear for pure TiO2. Since Au has a strong density of electrons, this layer can attract oxygen vacancypositive species due to its electrostatic force. TiO2-300 nm-Au 6 nm presents a similar peak As seen in Figure 4A, pure TiO 2 shows a rather strong Ti element signal peak (459.34 and 464.84 eV) in the XPS diagram. Even with decoration of only a 3 nm Au layer over the TiO 2 layer, the signal of the TiO 2 element disappears considering that XPS is a surfacesensitive characterization technique. As seen in Figure 4B, no Au element signal was detected in the pure TiO 2 sample. On the contrary, the Au signal (84.15 and 87.79 eV) appears in both Au-TiO 2 samples and becomes higher for the TiO 2 composite film with the thicker Au layer. As seen in Figure 4C, the as-prepared three samples have very different signals for oxygen species. The signal at 530.54 eV belongs to lattice oxygen, while 531.69 eV belongs to vacancy oxygen and 532.34 eV belongs to OH oxygen. Pure TiO 2 thin film has two oxygen species, lattice oxygen and OH oxygen, which originate from TiO 2 and surface adsorption of water molecules in the air. TiO 2 -300 nm-Au 3 nm shows no distinct lattice oxygen peak, which is the result of coverage of the surface Au layer on the surface of the TiO 2 . The oxygen vacancy's peak appears for this sample while it does not appear for pure TiO 2 . Since Au has a strong density of electrons, this layer can attract oxygen vacancy-positive species due to its electrostatic force. TiO 2 -300 nm-Au 6 nm presents a similar peak position to TiO 2 -300 nm-Au 3 nm. However, the number of oxygen vacancies becomes less for TiO 2 -300 nm-Au 6 nm compared to that of TiO 2 -300 nm-Au 3 nm.
As shown in Figure 5A, the 300 nm TiO 2 -Au 6nm sample shows the best response to oxygen among the three samples. Compared to bare TiO 2 thin films, few nm Au decoration shows great enhancement in response to oxygen gas. The TiO 2 -Au 6 nm sample has the best performance for oxygen sensing of the three samples, which is due to the influence of Schottky contact and oxygen vacancy. Figure 5B shows that the oxygen concentration and response have a linear relationship.
Chemosensors 2023, 11, x FOR PEER REVIEW 5 of 8 position to TiO2-300 nm-Au 3 nm. However, the number of oxygen vacancies becomes less for TiO2-300 nm-Au 6 nm compared to that of TiO2-300 nm-Au 3 nm. As shown in Figure 5A, the 300 nm TiO2-Au 6nm sample shows the best response to oxygen among the three samples. Compared to bare TiO2 thin films, few nm Au decoration shows great enhancement in response to oxygen gas. The TiO2-Au 6 nm sample has the best performance for oxygen sensing of the three samples, which is due to the influence of Schottky contact and oxygen vacancy. Figure 5B shows that the oxygen concentration and response have a linear relationship.   As shown in Figure 5A, the 300 nm TiO2-Au 6nm sample shows the best response to oxygen among the three samples. Compared to bare TiO2 thin films, few nm Au decoration shows great enhancement in response to oxygen gas. The TiO2-Au 6 nm sample has the best performance for oxygen sensing of the three samples, which is due to the influence of Schottky contact and oxygen vacancy. Figure 5B shows that the oxygen concentration and response have a linear relationship.  As shown in Table 1, our Au-TiO 2 thin films exhibit superior performance with other related works. The response time and recovery time of our samples are also shorter than those of TiO 2 nanotubes by anodic oxidation. Considering that our samples can be obtained in a facial and batch-scale way, our TiO 2 oxygen sensors are ready for commercial production with comparable performance at lower costs.

Discussion
The main sensing mechanism of Au-TiO 2 nanofilms to oxygen is attributed to three parts: the catalytical effect of Au, electrical effect originating from the Schottky contact effect from Au-TiO 2 , and the oxygen vacancy effect from TiO 2 . The catalytical effect of Au is often referred to as the spill-over effect. Briefly speaking, Au, as a noble metal catalyst, can lower the activation energy of dissociation of oxygen molecules and facilitate the dissociation of oxygen molecules into oxygen atoms, making oxygen atom chemisorption onto the surface of TiO 2 easier. This effect increases the effective concentration of reactive oxygen species so as to enhance the sensitivity of Au/TiO 2 to oxygen gas, as seen in Figure 6A. At the same time, an increase in Au content on the surface enhances the Schottky contact effect over the surface due to the creation of more Schottky barriers, thus increasing the response, as shown in Figure 6B. As shown in Table 1, our Au-TiO2 thin films exhibit superior performance with other related works. The response time and recovery time of our samples are also shorter than those of TiO2 nanotubes by anodic oxidation. Considering that our samples can be obtained in a facial and batch-scale way, our TiO2 oxygen sensors are ready for commercial production with comparable performance at lower costs.

Discussion
The main sensing mechanism of Au-TiO2 nanofilms to oxygen is attributed to three parts: the catalytical effect of Au, electrical effect originating from the Schottky contact effect from Au-TiO2, and the oxygen vacancy effect from TiO2. The catalytical effect of Au is often referred to as the spill-over effect. Briefly speaking, Au, as a noble metal catalyst, can lower the activation energy of dissociation of oxygen molecules and facilitate the dissociation of oxygen molecules into oxygen atoms, making oxygen atom chemisorption onto the surface of TiO2 easier. This effect increases the effective concentration of reactive oxygen species so as to enhance the sensitivity of Au/TiO2 to oxygen gas, as seen in Figure  6A. At the same time, an increase in Au content on the surface enhances the Schottky contact effect over the surface due to the creation of more Schottky barriers, thus increasing the response, as shown in Figure 6B.  The reaction between the O 2 and oxygen vacancies can be written as the following equation: (1) The response of the Au-TiO 2 nanofilm oxygen sensor is influenced by the concentration of oxygen vacancies in the TiO 2 film. The Au layer on the surface generally decreases the ratio of oxygen vacancies over the whole component on the Au-TiO 2 surface, thus having a Chemosensors 2023, 11, 476 7 of 8 negative effect on the response of the sample. The final response is a combination of the spill-over effect of Au, Schottky effect of Au/TiO 2 , and oxygen vacancies effect of TiO 2 . As shown from the results, TiO 2 -300 nm-Au-6 nm has higher response to oxygen than that of TiO 2 -300 nm-Au-3 nm.

Conclusions
It has been shown that nanometer-thick Au thin films can be added over TiO 2 thin films on microhotplate substrates to greatly promote the oxygen-sensing performance of the thin films. In our experiments, the sensing performance of 3 and 6 nm thick Au over TiO 2 thin films were compared with that of pure TiO 2 thin films. It was shown that TiO 2 -300 nm-Au-6 nm has the best sensitivity toward oxygen. The best performance of TiO 2 -300 nm-Au-6 nm is due to the contributions of the combination of spill-over of Au, the Schottky contact of Au-TiO 2 , and the oxygen vacancy effect of TiO 2 . TiO 2 -300 nm-Au-3 nm has more oxygen vacancies than TiO 2 -300 nm-Au-6 nm. However, TiO 2 -300 nm-Au-6 nm has stronger spill-over and Schottky effects. Overall, TiO 2 -300 nm-Au-6 nm has better performance than TiO 2 -300 nm-Au-3 nm. Our sensor, based on TiO 2 -300 nm-Au-6 nm, has a response value of 61.3 to 5% oxygen. The response time and recovery time to 5% oxygen are 37 and 40 s. The performance of our best sensor is superior to most reported TiO 2 sensors. Considering the facial and batch production nature of our methods, the oxygen sensor proposed is very promising for commercial applications.