Comparison of Magnesium and Titanium Doping on Material Properties and pH Sensing Performance on Sb2O3 Membranes in Electrolyte-Insulator-Semiconductor Structure

In this research, electrolyte-insulator-semiconductor (EIS) capacitors with Sb2O3 sensing membranes were fabricated. The results indicate that Mg doping and Ti-doped Sb2O3 membranes with appropriate annealing had improved material quality and sensing performance. Multiple material characterizations and sensing measurements of Mg-doped and Ti doping on Sb2O3 sensing membranes were conducted, including of X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS) and transmission electron microscopy (TEM). These detailed studies indicate that silicate and defects in the membrane could be suppressed by doping and annealing. Moreover, compactness enhancement, crystallization and grainization, which reinforced the surface sites on the membrane and boosted the sensing factor, could be achieved by doping and annealing. Among all of the samples, Mg doped membrane with annealing at 400 °C had the most preferable material properties and sensing behaviors. Mg-doped Sb2O3-based with appropriate annealing are promising for future industrial ionsensing devices and for possible integration with Sb2O3-based semiconductor devices.


Introduction
Over the past fifty years, growing attention has been paid to the development of the chemical sensing of ion concentrations in various solutions. Measurements of ion concentrations, such as pH sensing, are crucial to monitor human health, food safety and environmental pollution. Owing to rapid detection, fast response and reliable longterm operations, semiconductor-based ion sensitive devices have been proposed, such as ion-sensitive field-effect transistors (ISFETs), electrolyte-insulator -semiconductor (EIS) capacitors and light-addressable potentiometric sensors (LAPS-) [1,2]. Among these devices, EIS capacitors, with their advantages of compact size, low cost and simple fabrication, have been demonstrated as multianalyte ion and solute sensing devices [3,4]. In an EIS capacitor, the key component is the sensing membrane, capable of detecting ions in solutions. For conventional EIS devices, SiO 2 is commonly used as the dielectric for the membrane, though materials such as Ta 2 O 5 [5], Gd 2 O 3 [6] and Zr 2 O 3 have also been studied as possible replacements for traditional SiO 2 . In order to further boost the EIS capacitor's performance and provide the possibility for future device integration, it is worthwhile to explore new materials and treatments for the membrane material in order to fabricate EIS capacitors with pH-sensing behaviors. Based on previous reports [7,8], Sb 2 O 3 has been used for transparent conducting films, varistors and photocatalysts. In addition, the usage of Sb 2 O 3 can improve electrical and optical device properties. Based on recent studies [9][10][11][12][13][14][15][16][17][18][19] Sb 2 O 3 has been used for gas and liquid solution sensors. However, Sb 2 O 3 EIS sensors have not been clearly reported yet. Compared with conventional semiconductor materials, Sb 2 O 3 with a wide band gap of around 3 eV [20] can withstand high breakdown voltage. As for new treatments, incorporating various atoms such as F [21], N [22], Ti [23] and Mg [24] by plasma treatment [25] or co-sputtering [26] into EIS membranes to strengthen their material quality has been intensively investigated.  [27][28][29][30][31][32][33][34][35] as the membrane material have not been clearly reported yet. Since Mg has low electron affinity and low electronegativity, and the radius of Mg 2+ is similar to that of Sb 3+ , Mg can perfectly replace the lattice site of Sb [36][37][38][39][40][41][42][43]. In addition, many studies have also shown that Mg doping can increase the energy gap and reduce redundant oxygen vacancies. Moreover, annealing treatment in an O 2 ambient can further improve sensing performance and device reliability. This is because filling the oxygen vacancy and reducing the defects in an oxygen environment repairs the dangling bonds and releases the strain bonds, while more oxygen atoms are added to the surface and the oxygen vacancies in the lattice are filled. Therefore, Mg-doped Sb 2 O 3 EIS sensors with annealing at 400 • C can achieve a high sensitivity of 60.17 mV/pH, which is above the Nernst limit [44]. To gain insight into the effects of annealing, multiple material analyses including X-ray diffraction (XRD), X-ray photoelectron sptroscopy (XPS) and field-effect scanning electron microscopy (FESEM) were used to examine the surface morphologies and material properties of Sb 2 O 3 membranes. Material characterizations reveal that annealing at an appropriate temperature can enhance Sb 2 O 3 crystallization and Ti doping of the Sb 2 O 3, and Mg doped of the Sb 2 O 3 can suppress the formation of silicate grainization. Therefore, the EIS pH-sensing capability could be boosted and reliability issues such as hysteresis and drift voltage can be mitigated. Mg-doped Sb 2 O 3 -based EIS capacitors are promising for versatile integration in chemical ion-sensing applications with other Sb 2 O 3 -based devices [45][46][47][48].

Experimental
To fabricate Sb 2 O 3 -based EIS capacitors, the films were deposited on 4-inch n-type (100) wafers with a resistivity of 5-10 Ω-cm. The Sb 2 O 3 was sputtered, and Mg or Ti was co-sputtered on the wafers in an Ar:O 2 = 20:5 ambient, respectively. The RF power was l00 W and the pressure inside the chamber was l0 mTorr. Based on the measurements, the thickness of the undoped Sb 2 O 3 was 61.53 nm, the thickness of Mg-doped Sb 2 O 3 was 63.40 nm and the thickness of the Ti-doped Sb 2 O 3 was 69.25 nm. The rapid thermal annealing (RTA) process was carried out for 30 s in O 2 ambient at temperatures of 400, 500 and 600 • C, respectively. Then, a 300 nm aluminum film was deposited on the backside of the wafer. The backside aluminum film was grown by e-beam evaporation. After that, an epoxy bond was used to determine the sensing area. Finally, silver gel was used to attach samples on the copper wires of the printed circuit board (PCB). The detailed EIS structure is shown in Figure 1. Analyses using CV measurements were performed. Using a reference capacitance of 0.4 C max in the CV curves, the correlation between the substrate bias and the electrolyte concentration was calculated. In addition, the change of substrate bias voltage variation caused by the change of electrolyte concentration can be illustrated by the site-binding model [49,50], with the flat band voltage shift proportional to the electrolyte concentration, as in the following equation: E Re f is the reference electrode potential, and χ sol is the surface dipole potential of the solution. Φ Si is the work function of silicon, and Ψ 0 is the liquid junction potential difference. All of the terms in the equation are constant, except for Ψ 0 , which makes the membrane sensitive to the electrolyte due to polarization and forms the potential barrier. Furthermore, Ψ 0 is closely related to the surface sites on the membrane.
Moreover, the hydrogen ionic reaction with the membrane interface is illustrated in the site-binding model shown in Equation (2) [51][52][53]. The surface potential (ψ) can be related to the membrane parameter β. k is Boltzmann's constant, q is the elementary charge, T is the temperature, pH pzc is the pH value with zero charge on the surface and β is a factor that points to the sensitivity of the gate membrane.
Furthermore, β is closely related to the density of surface hydroxyl groups, as shown in (3). N s is the number of surface sites per unit surface area and C DL is the double layer capacitance, according to the Gouy-Chapman-Stern model [54].
In this research, Mg-doped Sb 2 O 3 -based EIS capacitors were demonstrated as the best pH-sensing devices, as shown in Figure 2. The tighter inner layer becomes the Helmholtz layer, which causes the positive charge to be absorbed by the Helmholtz layer and is not affected by the potential difference. Therefore, the outer layer is still a diffusion layer, and the resulting capacitance C DL is reduced. It is known from the Moss-Berstein effect that, when the capacitance C DL decreases, the β value increases, so, the higher β is, the better the sensing capability will be [55,56].
Based on (3), the higher the surface site density of N s is, the higher β is and the better the sensing capability will be. The grainization, crystallization and compactness of the material structure on the sensing membrane may reinforce the quality and quantity of the surface sites.

Results and Discussion
To investigate the effects of Ti doping and Mg doping with annealing on a Sb 2 O 3 membrane, XRD was used to monitor the crystalline phases of the differently treated films. The XRD patterns of the undoped, Ti-doped and Mg-doped samples are shown in Figure 3a-c, respectively. As shown in Figure 3a-c, Sb 2 O 3 (222), (400) and (440) phases can be observed in the XRD patterns of all of the samples. Among the undoped samples, the film annealed at 500 • C had the strongest crystalline phases. In addition, among Ti-doped and Mg-doped samples, the samples annealed at 400 • C had the strongest crystallized phases. Furthermore, Ti-doped and Mg-doped samples had stronger XRD peaks than the undoped samples, indicating that the combination of doping could further enhance the crystallization. However, as the anneal temperature increased to 400 and 500 • C, all of the Sb 2 O 3 peaks increased, indicative of the crystalline phases strengthening and the crystallization enhancing. As the annealing temperature increased to 600 • C, the Sb 2 O 3 decreased, and the crystalline structures might be deteriorated in undoped, Ti-doped and Mg-doped samples owing to the RTA pH-sensing devices at 600 • C.
Furthermore, to monitor the chemical bindings and element compositions, the O 1s XPS analysis was performed on the undoped, Ti-doped and Mg-doped samples, as shown in Figure 4a-c, respectively. Figure 4a reveals that an RTA at an appropriate temperature of 500 • C could effectively suppress the formation of silicate and optimize the sensing device's performance. However, as the annealing temperature further increased to 600 • C, the amount of silicate increased again. Similarly, annealing at 400 • C could effectively suppress the formation of silicate, as shown in the XPS spectra of Ti-doped and Mg-doped samples, as shown in Figure 4b Figure 4d-f, respectively. The Sb-O doublet peaks can be observed on these spectra. The Sb 3d XPS spectra also show that the Sb-O bonds could be strengthened with an appropriate annealing temperature of 500 • C for the undoped sample, as shown in Figure 4d. In addition, Mg doping could further enhance the Sb-O bonds in terms of Sb-O doublet peaks compared with the undoped samples. Consistent with the XRD patterns, silicate could be mitigated and chemical bindings could be enhanced by annealing and Mg doping [57]. Since XPS could complement the XRD analysis and was the technique used to identify the presence of the Mg and Ti bindings, the measurements of the Ti 2p XPS spectra for Ti-doped samples, as shown in Figure 4g, and the Mg 2p spectra for Mg-doped samples, as shown in Figure 4h, were performed, respectively. The results indicate that the XPS peaks for Ti and Mg did not vary a lot in various annealing conditions, but the peaks did represent the presence of Mg and Ti binding.     To examine the nanostructures of the Sb 2 O 3 samples, TEM and HRTEM were used to analyze the film in nanometer scale [58]. The TEM and HRTEM images of the as-deposited Sb 2 O 3 , the Ti-doped Sb 2 O 3 sample annealed at 400 • C, the as-deposited Mg-doped Sb 2 O 3 sample and the Mg-doped Sb 2 O 3 sample annealed at 400 • C are shown in Figure 5a-d, respectively. The crystalline interval of the as-deposited sample is not clear. As for the annealed Ti-doped sample, the interval spacing can be observed on the TEM image, as shown in Figure 5b. In addition, the enlarged HRTEM image of the interval spacing reveals that the width interval spacing is 0.327 nm, which is close to the 0.322 nm of Sb 2 O 3 (222) in a HRTEM image based on a previous report [59]. In addition, the interval spacing of the as-deposited Ti-doped sample can be seen in both the TEM and HRTEM images of Figure 5c. The morphology of the interval spacing of the annealed Ti-doped sample and the as-deposited Mg-doped sample are similar. Compared with the annealed Ti-doped sample and the as-deposited Mg-doped sample, the Mg-doped annealed sample exhibits clearer interval spacing in the TEM and HRTEM images, as shown in Figure 5d. Corresponding with the XRD patterns, as shown in Figure 2, the Mg-doped film annealed at 400 • C had the strongest Sb 2 O 3 (222) phases, implying that the combination of Mg doping and annealing could optimize the film's material quality. To measure the sensitivity and linearity of EIS capacitors, a Ketheley 2400 Source Meter was used to evaluate the C-V curves of the samples treated in various conditions. With 0.4 C max set as the reference capacitance, the sensitivity and linearity could be calculated by extracting the points of various pH values with this reference capacitance.  Consistent with the material characterizations, the pH sensitivity and linearity of the as-deposited membrane could be improved by annealing at 500 • C, as shown in Figure 6a,b. Moreover, the sensitivity and linearity could also be boosted by Ti doping and Mg doping without annealing, as shown in Figure 6c,e. In addition, appropriate annealing at 500 • C could further enhance the sensitivity of the Ti-doped samples from 48.94 to 54.08 mV/pH and the Mg-doped samples from 55.94 to 60.17 mV/pH, respectively, as shown in Figure 6d  As for the undoped samples, annealing at the appropriate temperatures of 400 • C and 500 • C effectively enhanced the sensitivity and linearity of the membrane, as shown in Figure 8a. Moreover, as shown in Figure 7b,c, in line with all of the material analyses, the combination of doping and appropriate annealing at 400 • C superpositionally increased both sensitivity and linearity for the Ti-doped and Mg-doped samples. Annealing with a high annealing temperature of 600 • C would degrade the material quality and the sensing behaviors of the doped Sb 2 O 3 membranes. Material quality improvements enhanced the device sensing properties. Among all of the membranes, the Mg-doped membrane annealed at 400 • C with the strongest crystallization, grainization, and chemical bindings had the best pH sensitivity of 60.17 mV/pH and a high linearity of 99.06% [60,61].  To study stability and long-term reliability issues, the hysteresis voltages and the drift voltages were assessed. To calculate the hysteresis voltages, the samples were immersed in solutions with various pH values of 7, 4, 7, 10 in an alternating cycle with an immersion time of 5 min. The hysteresis voltage of the Ti-doped and Mg-doped samples with various annealing treatments are shown in Figure 8a,b, respectively. Consistent with the material characterizations and sensing measurements, the hysteresis voltages for the Ti-doped sample and the Mg-doped sample annealed at 400 • C had the lowest values of 27.28, 4.84 and 4.29 mV, respectively. Mg doping incorporated with annealing at 400 • C had the most reliable response. Since the dangling bonds and traps might capture H + or OHions in solutions, a membrane of better material quality might exhibit lower hysteresis voltages. Furthermore, to evaluate the capacitor's long-term reliability, all of the tested samples were immersed in pH7 buffer solutions for 12 h, and the drift voltages were calculated. Similarly, the hysteresis voltages for the Ti-doped sample and the Mg-doped sample annealed at 400 • C had the smallest values of 3.01 and 1.99 mV/hr, respectively, as shown in Figure 8c,d. Since post-annealing at 400 • C incorporating Mg doping could effectively reduce vacancies and defects, the drift voltage of the Mg-doped sample annealed at 400 • C was greatly suppressed. These reliability tests were consistent with all of the other electrical measurements and material analyses [62].
Finally, to compare the sensing sensitivity of different ions on the EIS structure, the sensitivity and linearity of the K + and Na+ ions of the undoped, the Ti-doped and Mgdoped samples were measured. As shown in Figure 9a-f, the sensitivity and linearity measurement of K + ion for the samples reveal that the Mg-doped sample annealed at 400 • C had the optimized sensitivity of 17.3 mV/pK and linearity of 96.15%.  Similarly, the sensitivity and linearity measurements of the Na + ion for the samples reveal that the Mg-doped sample annealed at 400 • C had an optimized sensitivity of 21.01 mV/pK and a linearity of 96.15%, as shown in Figure 10a-f. Since the radius and mass of H + ions are much smaller than the radius and mass of K + and Na + , the sensitivity of the K + and Na + ions is smaller than that of the H + ions.
Finally, the sensitivity, linearity, hysteresis characteristics and drift characteristics of the appropriately annealed undoped, Ti-doped and Mg-doped Sb 2 O 3 membranes were com-pared. The Mg-doped Sb 2 O 3 membranes annealed at 400 • C still had the most preferable sensing characteristics compared with all of the other samples, as shown in Figure 11.

Conclusions
In this study, undoped, Ti-doped and Mg-doped Sb 2 O 3 sensing membranes with various annealing conditions were fabricated. Multiple material characterizations and sensing measurements were conducted to study the annealing and doping effects on the membranes. The results indicate that Mg doping incorporating annealing at an appropriate temperature of 400 • C could optimize the material quality and enhance the sensing behaviors due to the suppression of silicate, the enhancement of crystallization, the boosting of sensing factors and the removal of defects. Mg-doped Sb 2 O 3 -based membranes with appropriate annealing are promising for future industrial ion-sensing devices and for possible integration with Sb 2 O 3 -based devices.

Data Availability Statement:
The data used to support the findings of this study are available from the corresponding author upon request.