Effects of NH3 Plasma and Mg Doping on InGaZnO pH Sensing Membrane

In this study, the effects of magnesium (Mg) doping and Ammonia (NH3) plasma on the pH sensing capabilities of InGaZnO membranes were investigated. Undoped InGaZnO and Mg-doped pH sensing membranes with NH3 plasma were examined with multiple material analyses including X-ray diffraction, X-ray photoelectron spectroscopy, secondary ion mass spectroscopy and transmission electron microscope, and pH sensing behaviors of the membrane in electrolyte-insulator-semiconductors. Results indicate that Mg doping and NH3 plasma treatment could superpositionally enhance crystallization in fine nanostructures, and strengthen chemical bindings. Results indicate these material improvements increased pH sensing capability significantly. Plasma-treated Mg-doped InGaZnO pH sensing membranes show promise for future pH sensing biosensors.


Introduction
The ion sensitive field-effect transistor (ISFET) with a sensing membrane for use in biosensing applications was demonstrated by P. Bergveld in 1970 [1]. When integrated with complementary metal oxide semiconductor technology, ISFETs are capable of detecting ion activity in the human body [2]. Concentrations of biologically related ions such H + , Na + , and K + are important health indexes associated with disease monitoring. Recently, various high-k dielectrics such as Ta 2 O 5 [3], Pr 2 O 3 [4], and Er 2 O 3 [5] have been proposed as sensing membrane materials to detect these crucial ions in the human body [6,7]. To further enhance ion-sensing capabilities, it is worthwhile to explore materials, alternative fabrication processes, and treatments such as addition of nanoparticles or modulation of the membrane thickness [8,9]. In addition, incorporating post treatments such as Magnesium (Mg) doping and ammonia (NH 3 ) plasma treatment may optimize the membrane performance [7,10,11]. In this study, InGaZnO films, which can function as a transparent conductive oxide, have been demonstrated as ion-sensing membranes in electrolyte-insulator-semiconductor (EIS) structures. Furthermore, magnesium atoms were doped into InGaZnO films and NH 3 plasma treatment was incorporated into the membrane fabrication process [12,13] to boost ion-sensing behavior. Based on previous studies [10,14,15], magnesium atoms can fill in vacancies and reduce dangling bonds, and NH 3 plasma treatment can include N atoms into the membrane to mitigate defects. Therefore, combination of Mg doping and NH 3 plasma treatment may effectively improve material quality and ion-sensing capability [13]. To examine an InGaZnO membrane with these treatments, multiple material characterization techniques were performed, and sensing behaviors were evaluated [16,17]. To observe the morphologies both on the surface and in the cross section, scanning electron microscopy (SEM), atomic force microscopy (AFM), and transmission electron microscopy (TEM) images were taken. X-ray diffraction (XRD) was used to monitor crystalline structures, X-ray photoelectron spectroscopy (XPS) was used to study chemical bindings, and secondary ion mass spectrometry (SIMS) was used to evaluate element composition along the depth. Results indicate that Mg doping and NH 3 plasma treatment could include both Mg and N atoms into the membrane to reinforce the chemical binding and strengthen crystallization [18]. Notably, AFM and SEM images revealed grainization on the membrane surface, and TEM images indicated passivation of cracks and separation lines in the cross section.
Consistent with material improvements, pH sensing capability was enhanced, and hysteresis voltage and drift rate were suppressed with the incorporation of Mg doping and NH3 plasma treatment in the membrane fabrication process [19,20]. This study confirms that incorporation of Mg doping and NH3 plasma treatment can work together to optimize membrane material properties and sensing performance [21,22]. Furthermore, based our previous study [23], appropriate annealing could effective improve the material quality and sensing behaviors. In this study, different from the annealing treatment, compared with the as-deposited samples, NH3 plasma treatment could boost InGaZnO samples with or without Mg doping in terms of sensing capability more effectively. Therefore, NH3 plasma treatment in this study is more favorable than annealing treatment in terms of improving the pH sensing behaviors. InGaZnO membranes with Mg doping and NH3 plasma treatment can achieve excellent sensing performance and are promising for fabrication of future portable biosensing devices [24].

Experimental
To fabricate InGaZnO sensing membranes on electrolyte-insulator-semiconductor structures, the structure was deposited on 4-inch n-type (100) silicon wafers with resistivity of 5-10 Ω-cm [25]. To remove any native oxide, the wafers were cleaned using HF (HF:H 2 O = 1:100). Next, 50-nm SiO 2 was grown by thermal wet oxidation. In our experiment, InGaZnO and Mg targets were purchased from Gredmann Company, Taiwan. Then, in the condition, 50-nm InGaZnO was deposited on the wafer by radio frequency (RF) reactive sputtering with a mixture of Ar and O 2 (Ar:O 2 = 23:2) ambient during sputtering [23]. In the second condition, a 50-nm Mg-doped InGaZnO sensing membrane was deposited by co-sputtering on the wafer. During the reactive sputtering, InGaZnO and Mg targets were used in an ambient of Ar:O 2 at 23:2 with RF power at 80 W and ambient pressure of 1.3 Pa. Both undoped and Mg-doped samples were subjected to a post-NH 3 plasma treatment in a plasma-enhanced chemical vapor deposition (PECVD) system with an RF power of 30 W for 1 min and 3 min, respectively. The NH 3 plasma treatment was performed by a PECVD and the model and make of the PECVD was Sancom and PD-240. An Al film 300 nm in thickness was then deposited on the backside of the wafer. Next, adhesive silicone gel was used to define a sensing window. Finally, the samples were fabricated in silver gel on the copper lines of a printed circuit board. An epoxy package was used to separate the EIS structure and the copper lines. Since the EIS structure is not stable because defects between interface layer and silicon. To overcome the problem, incorporation of NH 3 plasma treatement were used to optimize the sensing performance. The detailed fabrication process is illustrated in Figure 1. To evaluate the sensing behavior of a membrane, the ionic consumption reactions between the solution/sensing membrane can be explained by the site-binding model [26][27][28]. The voltage of the surface potential (ψ) depends on the pH concentration of the electrolyte and the sensing factor β. The value of (ψ) can be figured out using Equation (1).
(1) (k is Boltzmann's constant, T is the temperature, q is the elementary charge, pHpzc is the pH value with no charge). β is a factor related to the sensitivity of the gate membrane. Furthermore, the β is linked to the density of surface hydroxyl groups, as described in Equation (2). Ns is the number of surface site/area and CDL is the double layer capacitance based on the Gouy-Chapman-Stern model [29].  To evaluate the sensing behavior of a membrane, the ionic consumption reactions between the solution/sensing membrane can be explained by the site-binding model [26][27][28]. The voltage of the surface potential (ψ) depends on the pH concentration of the electrolyte and the sensing factor β. The value of (ψ) can be figured out using Equation (1).
(k is Boltzmann's constant, T is the temperature, q is the elementary charge, pH pzc is the pH value with no charge). β is a factor related to the sensitivity of the gate membrane. Furthermore, the β is linked to the density of surface hydroxyl groups, as described in Equation (2). N s is the number of surface site/area and C DL is the double layer capacitance based on the Gouy-Chapman-Stern model [29].
To analyze the membrane films, multiple material analyzing techniques are performed. The make and model of SEM and TEM are JEOL JSM-7500F and JEOL JEM 2100 PLUS, respectively. The operating voltages for SEM and TEM were 15 kV and 200 kV, respectively. In addition, the make and model of PL is HITACHI F-4500. The excitation laser wavelength was 325 nm with a laser spot diameter of 1 µm. The PL spectral range was 330~1000 nm (CCD sensor) and 1000~1500 nm (InGaAs sensor). As for the XRD apparatus, the make and model is Bruker D8 Discover. For XRD analysis of the samples, the grazing incidence of X-ray beam CuKa (k = 1.542 Å) radiation is used with an incidence angle step of 0.5 • in the diffraction angle range (2θ) from 20 • to 60 • . The make and model of SIMS was CAMECA IMS-7f with O 2+ ion source and image resolution 10,000, mass range 300. The SIMS instrument was used by an internally yielded beam ions focused on a sample surface to produce secondary ions. The generated ions were then passed through a mass spectrometer across a high electrostatic potential. The depth profiles of elemental and molecular species could be assessed by SIMS analysis. The AFM model and make are Bruker Dimension Icon. The image resolution of X-Y noise was less than 0.15 nm and Z noise less than 30 pm (Close Loop). The scanning range (X,Y) was 1 µm * 1 µm. The AFM in Bruker Dimension Icon modes with intermittent contact was using a silicon tip with a 10 pN/nm spring constant. A sample area of 3 × 3 µm was scanned with actuation rates up to 8 kHz in air and fluid.
The XPS model and make are XPS ULVAC-PHI and PHI 5000. The XPS spectra were carried out by a VG ESCA Scientific Theta Probe. As for the XPS instrument condition, the X-ray spot size was about 15 µm, the take-off angle was around 53 • and the pass energy was set as 50 eV. The X-ray source for the XPS measurement was Al Kα (1486.6 eV). Furthermore, the sputtering argon ion beam with a beam energy of 3 kV was operated at a current density of 1 µA/mm 2 .

Results and Discussion
In this study, the as-deposited InGaZnO samples were subjected to the NH 3 plasma treatment, material characterizations and sensing measurements were performed on the as-deposited samples and the samples with NH 3 plasma treatment. To examine the surface morphologies both on the surface and through the cross section, SEM images on the surface and TEM images on the cross section of the InGaZnO sample were taken. The InGaZnO sample with NH 3 plasma treatment and the Mg-doped InGaZnO sample with plasma treatment are shown in Figures 2 and 3. Both AFM and SEM images revealed that some crystals were generated on the surface of the samples treated with NH 3 plasma compared with the untreated InGaZnO samples. Moreover, line-shaped stripes in the untreated InGaZnO sample were reduced by NH 3 plasma treatment, and even eliminated by Mg doping plus NH 3 plasma treatment [3], as shown in Figures 2 and 3. Based on previous studies, NH 3 plasma treatment can roughen the surface and hence enhance the crystallization and grain size [11]. Therefore, the sensing factor β could be increased with the increased roughness value as shown on the AFM images ( Figure 3) because the number of sites exposed to the solution was enlarged. Moreover, the N atoms caused by NH 3 plasma incorporated into the film could mitigate the dangling bonds and reduce the traps [13].  Furthermore, the photoluminescence (PL) peak was enhanced and blue-shifted, which may result from crystallization. In addition, contact angle measurements revealed  Furthermore, the photoluminescence (PL) peak was enhanced and blue-shifted which may result from crystallization. In addition, contact angle measurements revealed Furthermore, the photoluminescence (PL) peak was enhanced and blue-shifted, which may result from crystallization. In addition, contact angle measurements revealed that the enhancement of hydrophilic properties might result from modification of the surface due to Mg doping, as shown in Figure 4.  In addition to studying morphologies, XRD, XPS, and SIMS were used to investigate the crystalline phases, chemical bindings, and incorporated atoms versus depth, respectively [4]. Figure 5a,b show XRD patterns for the undoped InGaZnO and Mg-doped InGaZnO membranes with various NH3 plasma treatments, showing that Mg-doping and plasma treatment could enhance the crystalline phase. The strongest crystallization occurred in the 3-min plasma treatment of the Mg-doped samples. As the plasma treatment time increased to 6 min, the crystalline phase became weaker. In addition, Mg co-sputtering can form magnesium oxide (MgO) crystals into IGZO as shown in Figure 5a,b and therefore, the sensing behavior can be improved [30,31]. The electronegativity of Mg (χ = 1.31) is low and stablize the binding with oxygen [32]. Moreover, magnesium atoms can replace the vacancies and repair dangling bonds to passivate defects in oxides [33]. Therefore. the membrane films with better material quality can have better pH sensing capability.
XPS spectra for the undoped and Mg-doped samples shown in Figure 5c,d show that an N-chemical binding peak emerged with the NH3 plasma treatment. Moreover, the strongest N-binding occurred at 3 min of plasma treatment for both the undoped and Mgdoped samples. This analysis is in line with the XRD patterns. Furthermore, SIMS data revealed that the incorporation N and Mg atoms along the depth of the undoped and Mgdoped samples, as shown in Figure 5e,f. Results indicate that N atoms could be uniformly distributed along the depth of the membrane to passivate defects, as shown in Figure 5e,f. Moreover, high concentrations of Mg could be found in the Mg-doped membrane. SIMS analysis of the showed that In, Ga, and Zn distribution spiked near the membrane/Si interface were mitigated and the defects were suppressed in the doped samples, likely due to Mg doping. This suppression may mitigate defects near the interface, and hence further boost sensing behaviors for the Mg-doped membrane [5]. In addition to studying morphologies, XRD, XPS, and SIMS were used to investigate the crystalline phases, chemical bindings, and incorporated atoms versus depth, respectively [4]. Figure 5a,b show XRD patterns for the undoped InGaZnO and Mg-doped InGaZnO membranes with various NH 3 plasma treatments, showing that Mg-doping and plasma treatment could enhance the crystalline phase. The strongest crystallization occurred in the 3-min plasma treatment of the Mg-doped samples. As the plasma treatment time increased to 6 min, the crystalline phase became weaker. In addition, Mg co-sputtering can form magnesium oxide (MgO) crystals into IGZO as shown in Figure 5a,b and therefore, the sensing behavior can be improved [30,31]. The electronegativity of Mg (χ = 1.31) is low and stablize the binding with oxygen [32]. Moreover, magnesium atoms can replace the vacancies and repair dangling bonds to passivate defects in oxides [33]. Therefore. the membrane films with better material quality can have better pH sensing capability.
XPS spectra for the undoped and Mg-doped samples shown in Figure 5c,d show that an N-chemical binding peak emerged with the NH 3 plasma treatment. Moreover, the strongest N-binding occurred at 3 min of plasma treatment for both the undoped and Mg-doped samples. This analysis is in line with the XRD patterns. Furthermore, SIMS data revealed that the incorporation N and Mg atoms along the depth of the undoped and Mg-doped samples, as shown in Figure 5e,f. Results indicate that N atoms could be uniformly distributed along the depth of the membrane to passivate defects, as shown in Figure 5e,f. Moreover, high concentrations of Mg could be found in the Mg-doped membrane. SIMS analysis of the showed that In, Ga, and Zn distribution spiked near the membrane/Si interface were mitigated and the defects were suppressed in the doped samples, likely due to Mg doping. This suppression may mitigate defects near the interface, and hence further boost sensing behaviors for the Mg-doped membrane [5].   After morphological and material characterizations, sensing behaviors of t doped and Mg-doped membranes with NH3 plasma treatment were evaluated [ evaluate the pH sensing behavior of a sensing membrance C-V curves were take capacitance changed with gate bias voltage swept for 3 V interval. As a reference tance of 0.4 Cmax is set, the values of reference voltages versus pHvaues could tracted from the C-V curves. The pH sensitivity and linearity could be calculated fr subfigure of reference voltages versus pH values. The pH senesing evaluated fro curves. After morphological and material characterizations, sensing behaviors of the undoped and Mg-doped membranes with NH3 plasma treatment were evaluated [17]. To evaluate the pH sensing behavior of a sensing membrance C-V curves were taken. The capacitance changed with gate bias voltage swept for 3 V interval. As a reference capacitance of 0.4 Cmax is set, the values of reference voltages versus pHvaues could be extracted from the C-V curves. The pH sensitivity and linearity could be calculated from the subfigure of reference voltages versus pH values. The pH senesing evaluated from C-V curves. Figure 6a,b show the sensing capability measurements for the undoped and Mg-doped membranes with NH 3 plasma treatment for 3 min. Results indicate that the pH sensing sensitivity for the undoped InGaZnO membrane achieved 62.28 mV/pH, which was above Nernst limit. However, the sensitivity of the Mg-doped membrane was as high as 65.85 mV/pH, indicating that both Mg doping and NH 3 plasma treatment could effectively enhance crystallization, reduce defects in the bulk and interface, and reinforce membrane material quality. As shown in Table 1, the sensitivity of NH 3 plasma treated InGaZnO sample and Mg doped InGaZnO sample have pH sensitivity of 62.28 and 65.85 mV/pH, which are above Nernst limit (around 60 mV/pH), while the InGaZnO and Mg-doped InGaZnO sample with appropriate annealing at 500 • C have the pH sensitivity of 56.51 and 59.3 mV/pH, which are below Nernst limit. Compared with plasma treatment with other gases, NH 3 might be more effective. Since nitrogen (N) was similar to oxygen (O) in regard to ionic radius and acts as a better compensator, NH 3 plasma was used in the post-treatment of sputter film. Therefore, we investigated the impact of NH 3 plasma treatment on the sensing behavior, surface morphology, and crystal structure. Due to the NH 3 plasma treatment, plasma-induced morphological changes and increment of grain size were observed, favoring the increase of surface roughness and number of surface defect sites, and thus resulting in higher sensitivity and linearity. The surface charge density was mainly related to the ionic activity in the solution. To assess the stability of the tested membranes, the hysteresis voltages and the drift effect were evaluated. To investigtate the hysteresis effects, the membranes were immersed in solutions with various pH values of 7, 4, 10, and 7 in an alternate time sequence. The submerging time was five minutes in each solution. The hysteresis voltage could be calculated by the voltage deviation between the initial and the terminal voltages taken in the pH loop. The dangling bonds could bind with the ions in the solutions, hysteresis response could be observed.
Hysteresis voltage measurements were then taken for the undoped and Mg-doped membranes, shown in Figure 6c,d, respectively, show that NH 3 plasma treatment with a time of 3 min can effectively suppress hysteresis voltage, possibly from the removal of the dangling bonds and traps [34]. Compared with the as-deposited InGaZnO film and the as-deposited Mg doped film, the NH 3 plasma treatment could lower the hysteresis voltage from 18.42 to 1.68 eV for the undoped sample and from 16.62 to 1.01 eV for the Mg-doped sample, respectively.  Finally To examine the drift rate voltage for long-time reliability, the samples were submerged in a pH7 buffer solution for 12 h. Consistent with the material and sensing characterizations, drift voltage rate measurements for the undoped and Mg-doped samples revealed that NH 3 plasma treatment could effectively remove the defective bindings as shown in Figure 6e,f. Plasma treatment could potentially cause the membrane to bind with ions in the solution. Moreover, Mg doping could further lower both the hysteresis voltage and drift rate, since incorporated Mg atoms might fill in the vacancies and bind with dangling bonds to enhance material quality. The drift voltage shift could also be induced by the ions captured by the dangling bonds, too. To illustrate the voltge shifts in the hysteresis and drift voltage measurements, Figure 7 explain the mechansim for hysteresis and drift voltage shifts caused by dangling bonds. In the beginning, no ions were attached on the dangling bonds. As the measuring time passed by, more and more ions were captured by the dangling bonds and the gate voltage shifts might occur as shown in Figure 7.  Based on previous reports [23,[35][36][37], Mg doping can generate Mg 2+ near the surface of the membrane and attract some OH − ions in the electrolyte solution. Therefore, the OH − in the diffusion layer of the electrolyte solution might decrease, and therefore the CDL (diffusion capacitance) might decrease as well. The illustration of CDL decrease is shown in Figure 8. Then, the membrane parameter β, the surface potential, and overall sensitivity could also be boosted. However, though NH3 plasma treatment could incorporate N atoms into the membrane and potentially decrease defects, it could not cause Mg atoms in the membrane to be uniformly distributed in the same way that Rapid Thermal Anneaing (RTA) annealing did in our previous research. Therefore, linearity could be decreased due to the uneven distribution of Mg atoms with the plasma treatment. In the future, a combination of RTA annealing and NH3 plasma treatment could be conducted to further improve the sensitivity and linearity of the membrane. As the plasma treatment time increased to 6 min, the hysteresis voltage and drift rate increased drastically. Since plasma treatment could damage the film by etching away the surface, plasma treatment with a time longer than 6 min could deteriorate the surface material quality consistent with the XRD analysis as shown in Figure 5. Therefore, plasma treatment with an appropriate time could optimize the material properties and device performance. Based on previous reports [23,[35][36][37], Mg doping can generate Mg 2+ near the surface of the membrane and attract some OH − ions in the electrolyte solution. Therefore, the OH − in the diffusion layer of the electrolyte solution might decrease, and therefore the C DL (diffusion capacitance) might decrease as well. The illustration of C DL decrease is shown in Figure 8. Then, the membrane parameter β, the surface potential, and overall sensitivity could also be boosted. However, though NH 3 plasma treatment could incorporate N atoms into the membrane and potentially decrease defects, it could not cause Mg atoms in the membrane to be uniformly distributed in the same way that Rapid Thermal Anneaing (RTA) annealing did in our previous research. Therefore, linearity could be decreased due to the uneven distribution of Mg atoms with the plasma treatment. In the future, a combination of RTA annealing and NH 3 plasma treatment could be conducted to further improve the sensitivity and linearity of the membrane. As the plasma treatment time increased to 6 min, the hysteresis voltage and drift rate increased drastically. Since plasma treatment could damage the film by etching away the surface, plasma treatment with a time longer than 6 min could deteriorate the surface material quality consistent with the XRD analysis as shown in Figure 5. Therefore, plasma treatment with an appropriate time could optimize the material properties and device performance. bination of RTA annealing and NH3 plasma treatment could be conducted to further improve the sensitivity and linearity of the membrane. As the plasma treatment time increased to 6 min, the hysteresis voltage and drift rate increased drastically. Since plasma treatment could damage the film by etching away the surface, plasma treatment with a time longer than 6 min could deteriorate the surface material quality consistent with the XRD analysis as shown in Figure 5. Therefore, plasma treatment with an appropriate time could optimize the material properties and device performance.

Conclusions
In this study, Mg doping and NH 3 plasma treatment were incorporated into the fabrication process of InGaZnO membranes. Inclusion of Mg and N atoms into the membrane could enhance crystallization, strengthen chemical binding, and reduce defects, as revealed in morphological and material characterizations. Evaluation of the sensing behavior also indicated that Mg doping and NH 3 plasma treatment with a time of 3 min could boost the sensing behaviors above the Nernst limit, and also enhance low hysteresis voltage and drift rate. Our results indicate that Mg-doped InGaZnO membranes with NH 3 plasma treatment show promise for future industrial pH sensing EIS-based biosensors.

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