Fabrication of an acetone gas sensor based on Si-doped WO3 nanorods prepared by reactive magnetron co-sputtering with OAD technique

Gas sensing technology is currently applied in a variety of applications. In medical applications, gas sensors can be used for the detection of the biomarker in various diseases, metabolic disorders, diabetes mellitus, asthma, renal, liver diseases, and lung cancer. In this study, we present acetone sensing characteristics of Si-doped WO3 nanorods prepared by a DC reactive magnetron co-sputtering with an oblique-angle deposition (OAD) technique. The composition of Si-doped in WO3 has been studied by varying the electrical input power applied to the Si sputtered target. The nanorods film was constructed at the glancing angle of 85°. After deposition, the films were annealed at 400 °C for 4 h in the air. The microstructures and phases of the materials were characterized by x-ray photoelectron spectroscopy (XPS), x-ray diffraction (XRD), and field-emission scanning electron microscopy (FESEM). The results showed that 1.43 wt% Si-doped WO3 thin film exhibited the maximum response of 5.92 towards 100 ppm of acetone at performing temperature (350 °C), purifying dry air carrier. The process exposed in this work demonstrated the potential of high sensitivity acetone gas sensor at low concentration and may be used as an effective tool for diabetes non-invasive monitoring.


Introduction
At present, the burden of diabetes has been rapidly increasing. The global prevalence of diabetes has been rising from 4.7% in 1980 to 8.5% in 2014. Over 420 million world populations are facing diabetes. Forecasting in 2030, demonstrates that more than 570 million people have been living with diabetes [1]. Diabetes is a chronic disease, caused by a deficiency of insulin secretion and insulin resistance. Diabetes patients who poorly control blood glucose lead to increased risks of other diseases including heart, retinopathy, nephropathy, and neuropathy and nerve, increasing limb amputation. The current standard diabetes diagnostic method is to analyze the amount of insulin by pricking blood at the fingertips. This method is an invasive technique, very painful, expensive, and may be infected. A non-invasive method by human breath analysis became more interesting for detecting various volatile organic compounds that were a biomarker for detecting different diseases such as metabolic disorders, diabetes mellitus, asthma, renal and liver diseases, and lung cancer. Human exhaled breath contains Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
We monitored blood sugar levels in Diabetes mellitus (DM). The blood-sugar level is related to the ketone bodies which consist of acetoacetate, beta-hydroxybutyrate, and acetone [4]. Acetone level in the breath for healthy people is in the range of 0.5-0.9 ppm and above 1.25-2.5 ppm for diabetes patients [5]. The medical reports showed that 40 ppm of acetone was found in normal people who eat ketogenic diet (high cholesterol), 360 ppm in children with seizures, and 1250 ppm in poorly controlled diabetic patients with ketoacidosis. Thus, the concentration of breath acetone below 0.9 ppm can be screening diabetes. There were several methods to measure breath acetone levels. The various techniques used to analyze acetone in the breath to diagnose diabetes such as Gas chromatography with mass spectrometry (GC-MS), Selected ion flow tube mass spectrometry (SIFT-MS), Proton transfer reaction with mass spectrometry (PTR-MS), Light-addressable potentiometric sensors (LPAS), and Semiconductor metal oxide sensor (SMOS) measured in part per million (ppm), part per billion (ppb), and part per trillion (ppt) [6].
In recent years, metal oxide semiconductor-based gas sensor is of great interest and widely used in portable acetone detection. Their advantages were a smaller size, higher sensitivity, lower cost, better reversibility, and easier operation compared to the other techniques, such as gas chromatography and mass spectrometry [7]. Therefore, many researchers have focused on the application of semiconductor metal oxide-based gas sensors in patients with diabetes. There are several metal oxides considered as a sensing material for monitoring acetone gas, including WO 3 , CuO, SnO 2 , and TiO 2 [8][9][10][11]. Their performances have been improved in terms of sensitivity and selectivity by additives or applied appropriate working temperature for target gas. Many studies have focused the sensing improvements via surface and microstructure modifications by doping and loading.
Tungsten oxide (WO 3 ) is one of the promising candidate metal oxides to be used as a sensing layer because of its physical and chemical properties, especially in acetone gas sensing applications. Haiyun Xu et al [12] have reported acetone gas sensors based on mesoporous WO 3 nanofiber with the crystalline framework and aqueous phase by an electrospinning process. The result showed a fast response time (24 s) and recovery time (27 s) to a relatively low acetone concentration of 1 ppm at 300°C. Similarly, Chang X et al [13] created WO 3 nanosheets of inorganic fullerene-like WS 2 particle prepared by sulfurization of WO 3 nanoparticle that showed the fast response time (6 s) and recovery time (9 s) to a slightly low level of acetone concentration (0.17 ppm at 300°C .11 M). Imran et al [14] used acetone sensor-based non-stoichiometric tungsten oxide (WO 3 -x) nanofibers prepared by electrospinning. It showed that the sensor had repeatability of more than 99% towards low acetone concentration at 350°C. Moreover, J Shi et al [15] have developed WO 3 nanocrystals by the sol-gel method for acetone detection. The sensor exhibited a super low acetone detection limit at 0.05 ppm at 300°C.
There was a report demonstrating that WO 3 surface modification by Si additive resulted in improvement of acetone response as shown in table 1. M Righettoni et al [16] have reported acetone gas sensor based on Si-doped έWO 3 phase prepared by flame spray pyrolysis. The sensor response is 4.63 at 600 ppb in dry air and 0.7 in 90% RH at the temperature of 400°C. In addition, A Rydosz et al [17] have reported acetone gas sensors based on Sidoped WO 3 prepared by glancing angle DC magnetron sputtering method. The sensor showed the maximum response of 40.53 towards 8 ppm acetone at 425°C.
In this work, we prepared Si-doped WO 3 nanorods by magnetron sputtering as a sensitive layer for acetone detection. The magnetron sputtering presented advantages with a homogeneous morphology, easily controlled the thickness of the films, and could be manipulated the composition of each deposited material by controlling the applied power to the sputtering sources [22,23]. The acetone gas sensing performance has been characterized to demonstrate the potential of Si-doped WO 3 nanorods as a high sensitivity acetone gas sensor at low concentration and can be used as a non-invasive method for diabetes monitoring.

Design and fabrication of an acetone gas sensor based on Si-doped WO3 nanorods
The design of the sensor was shown in figure 1(a). It consisted of a sensitive element and pairs of gold (Au) interdigitated electrodes, with interspacing between fingers of 100 μm and had an active area with approximately 3250 μm×4000 μm, fabricated on a silicon wafer with a SiO 2 buffer layer. The fabrication process, after standard cleaning of Si wafer, 1 μmthick SiO 2 insulating layer was grown by thermal oxidation protecting the electrical connection between the substrate (Si) and the electrode. The interdigitated electrodes were patterned by photolithography and followed by lift-off of deposited 150 nm-thick Au by electron beam evaporation. Then, a sensitive layer made of WO 3 was deposited on top of Au interdigitated electrodes using reactive magnetron cosputtering with oblique angle deposition (OAD) technique as described in figure 2. It was used for single-step film deposition via oblique angle physical vapor deposition with precision substrate rotation. This process provides nanorod fabrication with good dispersion [24]. During the film deposition process, the nanorods could be formed on the surface at oblique angles. Due to the shadowing effect, the particles were captured at higher surface points [25,26], leading to form rougher surfaces with columnar structures as shown in figure 3. In the reactive magnetron co-sputtering with OAD   procedure, nanorods were fabricated at oblique angle deposition producing arrays of nanorods with rough surfaces. The fabricated sensor used in this work was shown in figure 1(b).
The Si-doped WO 3 nanorods have been deposited by reactive magnetron co-sputtering (AJA international, Inc.; ATC 2000-F) with the OAD configuration. In the sputtering deposition process, the vacuum chamber was initially evacuated to a base pressure of 6.0×10-6 Tor. Sputtered targets made of 2-inch-diameter and 0.250inch-thick Tungsten (99.99% purity), and 2 inch-diameter and 0.250-inch-thick Silicon (99.99% purity) was used as sputtered targets. Both sputtered targets aligned to the center of the substrate surface in OAD geometry with a deposition angle of 85°. The W target was set at a distance of 69 mm from the centreline of the substrate. A sputtering pulse DC power of 150 W was initially applied to the tungsten target while varying input DC power from 0-30 W to silicon target. The chamber was filled with Argon (Ar) and oxygen (O 2 ) to ignite the plasma and act as a reactive gas. The Si-doped WO3 films were deposited with a constant ratio of 68% Ar/32% O2. The flows of Ar and O 2 were precisely controlled by mass flow controllers. The deposition pressure was maintained at 5×10 -3 Tor and the duration of deposition pressure was set at 72 min to achieve 500 nm film thickness. The sample was deposited at room temperature. Subsequently, the as-prepared samples were annealed at 400°C in the air for 4 h.
The physical properties of the films were characterized by using an x-ray diffractometer (XRD) to point out the amorphous and crystalline states of the films.
A field-emission scanning electron microscope (FESEM) was used to examine their superficial morphologies and nanostructures. The film's composition was analyzed by using Energy Dispersive x-ray Spectroscopy (EDS). The oxidation state of element composition was characterized by x-ray Photoelectron Spectroscopy (XPS). The binding energy was calibrated with C 1 s reference (284.8 eV).

Gas-sensing measurement
The acetone gas sensing performance of Si-doped WO 3 nanorods was studied by using a gas sensing measurement system presented in figure 4. Briefly, the target acetone concentration was achieved by mixing clean dry air (Air zero) with an acetone gas balanced in dry air at a T-junction connector using multichannel  mass flow controllers. The targeted concentration has flowed to the test chamber (2000 sccm) with a flow rate of 2 l/min). The electrical resistance of the sensor was monitored using a digital electrometer. The sampling time was set to 1 s. The resistance was measured at an operating temperature ranging from 250°C to 400°C as a function of acetone concentration in a range from 50 ppm to 100 ppm. The sensor response(S) to acetone was determined as the resistance ratio S=Rair/Rgas, where Rair was the resistance of the sensor in the air zero and Rgas was the resistance of the sensor upon exposure to the acetone gas. The response time is given by a time that attains 90% of the stabilized signal after gas exposure while the recovery time is defined by a time that attains 90% of the stabilized signal after air recovery.

Structure and micro-morphology characterization
The XRD patterns of Si-doped WO 3 at different Si weight% compositions by varying the input power to Si target (5-30 W) are shown in figure 5. It was obvious that strong and sharp diffraction peaks were observed indicating the sample was highly crystalline. The XRD patterns of the prepared samples could be matched to a monoclinic phase of WO 3 (JCPDS no. 43-1035), displaying the dominant planes of (020), (200), (120) (220), (312), (140), (420) and (317) as shown in figure 5. Furthermore, the XRD pattern of Si-doped WO 3 nanorods confirms that Si was presented in metallic Si (Si°) state, which was indexed to the face centered cubic structure of Si (JCPDS no.75-0589), with the plane of (111). The appearance of the Si diffraction peaks indicates that the silicon particles may be supported on the surface of tungsten trioxide.
In order to verify surface morphologies and nanostructures, the composition and particularly the doping content of an obtained sample, FE-SEM and EDS were carried out as shown in figures 6-7. Figure 6. shows the typical morphology and cross-section of pure and Si-doped WO 3 films. The SEM image of pure WO 3 ( figure 6(a)) shows that the surface contains lots of nanorods homogenously distributed on the substrate. For Sidoped WO 3 ( figure 6(b)), some nanorods were connected. The film thickness was found to be approximately 500 nm and nanorods were deposited on the substrate with an angle of 42 degrees Figures 7(a)-(e) illustrated the EDS maps and EDS spectra taken from the top-view surface morphology of Si-doped WO 3 (5 W Si/WO 3 sample) after deposition and annealing at 400°C in air. The corresponding EDS maps of W and O elements indicated homogeneous distributions of the components over the scanned region, while the distribution of Si element is visible with low intensity. The EDS spectra indicate the characteristic x-ray peaks corresponding to O, W, and Si atoms present in the sample and a small peak from C, which is derived from carbon tape used to hold the sample during the measurement. From the EDS analysis, the amount of Si was found to be 1.43 weight% (1.97 atomic%) for sample deposited input power 5 watts of Silicon target which is in good agreement with the low intensive Si distribution presented in the EDS map. The Si content tends to increase with increasing the input power of the Si source ( figure 7(f)).
The surface chemical compositions and oxidation state of elements existing in 5 W Si/WO 3 film after testing were illustrated in figure 8. The survey scan spectrum ( figure 8(a)) confirms that the expected material elements on the surface included W 4 f, O 1 s, and Si 2p as well as Carbon (C 1 s) due to surface contamination. Considering to W 4 f element ( figure 8(b)), W 4f 7/2 and W 4f 5/2 core levels can be individually split into one doublet pair at binding energies of 35.60 and 37.72 eV, respectively. The binding energy difference of W 4 f doublet peaks is 2.12 eV, which could be assigned to the W 6+ oxidation state [27]. For the oxygen ( figure 8(c)), the curve of O 1 s peak can be decomposed into three peaks located at 530.50, 531.50, and 532.78 eV. The main O 1 s peak could be attributed to lattice oxygen (O 2− ) while the middle and the last ones may be associated with chemisorbed surface oxygen (O − ) or weakly bonded oxygen species and hydroxide species since water molecules adsorbed on the surface, respectively [28]. Regarding the Si element ( figure 8(d)), the Si 2p peak was found at the binding energy of 103.10 eV, indicating that the oxidation state of Si in the sample is Si°or metallic Si [29]. Therefore, there was the presence of metallic Si distributing on the WO 3 surface.

Gas-sensing properties
The gas sensing properties were characterized using a flow-through gas sensing system (figure 4). The operating temperature was an important factor influencing the surface states of the metal oxide and the chemical reaction. The response of the sensors based on Si-doped WO 3 nanorods towards 100 ppm acetone vapor tested at a different operating temperature ranging from 250°C to 400°C is presented in figure 9. It was obviously shown that the acetone response of each sensor initially increased to the highest response at an optimal temperature before decreasing when temperature increases. The optimum operating temperature for acetone gas-sensing examinations of various Si-doped WO 3 was around 350°C. In addition, the acetone response was sustained significantly as the small amount of Si doping but then became decreasing as the Si doping content increased at all working temperatures. Especially, Si-doped WO 3 sensor displays the highest response of approximately 5.92 was deposited from input power 5 watts to silicon target and co-sputtering. Figure 10 demonstrated changes in resistance of WO 3 sensors with different Si doping concentrations exposed to various acetone concentrations (50-100 ppm) at the optimum working temperature. The baseline resistance of the Si-doped WO 3 sensor seems to decrease with increasing Si doping contents. Reduction of WO 3 resistance by doping Si attributes the electrons transfer at the heterojunction. The electrons are transferred from Si particles to the WO 3 nanorods, then electrons accumulate at the surface of WO 3 , leading to a reduction of the baseline resistance [30]. After being subjected to acetone vapor, the resistance of all sensors decreased, especially a resistance change of Si-doped WO 3 sensor significantly enhanced with Si doping at 5 W power of Si source before declining at the higher Si doping. It was an acetone detection at low concentration, the noise was observed during acetone exposure due to the fluctuation of acetone concentration from the bubbling process. It could be seen that the increasing response of acetone for all sensors when increasing acetone concentration. The result Figure 11. (a) Selectivity characterization of 5 W Si/WO 3 sensor exposed to various gases and (b) long-term stability of 5 W Si/WO 3 sensor towards 100 ppm acetone at 350°C.
implied that an increased number of acetone molecules interacted with adsorbed oxygen species on the surface of WO 3 . Moreover, the sensor response improves substantially at the Si doping content at the 5 W power source, which offers the highest acetone response of approximately 5.92 at 100 ppm. Regarding the response dependency on acetone concentration, the optimal sensor shows a low detection limit at 50 ppm with a response of around 5.20. Lastly, the acetone selectivity of the optical sensor (5 W Si/WO 3 ) was investigated against 100 ppm H 2 S, CO 2 , H 2 O, and NH 3 at the working temperature of 350°C, as illustrated in figure 11(a). It is seen that the sensor exhibited a higher acetone response than other gases, revealing that the sensor presents excellent acetone selectivity. Figure 11(b) demonstrates the sensor response of the 5 W Si/WO 3 sensor towards 100 ppm acetone at 350°C for 15 days. It shows that the acetone response of the 5 W Si/WO 3 sensor is relatively stable with a fluctuation of ∼20%. Therefore, Si-doped WO 3 sensor can be an attractive choice for acetone detection that has a great potential for breath analysis.
The acetone gas sensitivity was investigated from the prepared Si-doped WO 3 nanorods at different electrical input power applied to the Si sputtered target. Mechanism of the gas sensor based on the analysis of the characteristics of above gas sensing. The sensing mechanism of the Si-doped WO 3 nanorod-based sensor could be created by using the surface-controlled model. WO 3 nanorods were normally n-type semiconductors, in which electrons became the major charge carriers and played an essential role in electrical properties. Then, oxygen molecules could be easily formed chemisorbed oxygen species (O − or O 2− ) [21,31,32] by trapping one or two electrons at the surface of WO 3 nanorods, which caused an increase of electron depletion region of the surface of WO 3 nanorods and increasing its resistance.
Under the acetone gas exposure ( figure 12(a)), the acetone molecules reacted with chemisorbed oxygen partners on the surface of WO 3 nanorods, releasing free-electron back to the WO 3 nanorods. Consequently, the thickness of the depletion region decreases, leading to a reduction of resistance of WO 3 nanorods, which is in good agreement with the experimental result in figure 10. The reaction can be expressed by following equations (1)-(2) [13,15,33,34]. With Si doping on the WO 3 nanorods, Si particles can improve acetone sensing performance in two different ways. Firstly, Si particles can be acted as catalytic for the oxygen spill-over process in Si-WO 3 nanorods [34]. When the air exposure to the system, Si-doped WO 3 nanorods can dissociate O 2 into O − , and then O − overflow adsorbs onto the WO 3 surface ( figure 12(b)). Then Si particles reduced the activation energy required for the reaction and further enhance its response to acetone. Secondly, doping with Si can induce a large surface area. The gas response is commonly proportional to the gas-surface area interaction [34]. More and more surfaceactive sites by optimal Si dopants led to improve the acetone response. Therefore, the presence of Si-doped WO 3 nanorods promotes additional surface-active sites and regeneration of electrons by enhancement of gaschemisorbed oxygen interactions, leading to the overall improvement of the response.
The fast response-recovery time is greatly ascribed to more surface-active sites to promote interaction with acetone. The reactive magnetron co-sputtering with OAD technique provided the possibility for fabrication of well-ordered and sophisticated nanostructure, e.g. nanorods, nanocolumns by manipulating the deposition angle and substrate angle and fabrication Si-doped WO 3 nanorods prepared by magnetron sputtering as a sensitive layer for acetone detection.

Conclusion
The gas-sensing properties of Si-doped WO 3 nanorod films deposited by reactive magnetron co-sputtering with OAD and annealed at 400°C for 4 h in the air were characterized in the presence of acetone at various concentrations. The results showed that 1.43% of Si-doped WO 3 nanorods exhibited the maximum response (S=5.92) in the actual 100 ppm acetone at operating temperature 350°C, purified dry air carrier. The process exposed in this work demonstrated the potential of high sensitivity acetone gas sensor at low concentration and can be used as a supplementary tool for diabetes monitoring.