Process Optimization and Characterization of Dielectric, Piezoelectric, and Ferroelectric Properties of ScAlN (30%) Thin Films

1. Introduction

Piezoelectric and ferroelectric materials play a pivotal role in microelectromechanical systems (MEMS), enabling critical functionalities in sensing [1,2,3], acoustic communication [4,5,6,7], and radio-frequency (RF) filtering [8,9,10]. Among these materials, aluminum nitride (AlN) has attracted substantial attention due to its high acoustic velocity [11], excellent thermal conductivity [11], and compatibility with complementary metal–oxide–semiconductor (CMOS) fabrication processes [12]. However, the relatively low piezoelectric coefficients of AlN limit its use in applications that require strong electromechanical coupling [13].

To address this limitation, extensive research has been conducted focused on enhancing the piezoelectric response of AlN through alloying and elemental doping. Among the various approaches, scandium (Sc) incorporation has proven particularly effective, significantly improving both the piezoelectric and ferroelectric properties of AlN and enabling its deployment in advanced acoustic and memory devices [14,15,16,17]. Early investigations using dual reactive co-sputtering demonstrated a noticeable enhancement in the piezoelectric response of Sc-doped AlN thin films [18,19,20]. Both ab initio models [21,22] and experimental studies [23] have shown that Sc concentrations up to 43% can yield up to a fourfold increase in the longitudinal piezoelectric coefficient ($d_{33}$), achieving values of approximately $27.6\mathrm{pC}/\mathrm{N}$. More recently, Lu et al. [24] reported a $d_{33}$ value of $31.6\mathrm{pC}/\mathrm{N}$ for 41% Sc-doped AlN, while Kenjita et al. [25] demonstrated that incorporating a lutetium (Lu) buffer layer can stabilize the wurtzite phase beyond the 43% Sc solubility limit, resulting in a $d_{33}$ of $35.5\mathrm{pC}/\mathrm{N}$ for 50.8% Sc-doped AlN.

Although extensive research has been conducted on optimizing ScAlN thin films to enhance their performance [26,27,28,29], a systematic study of how reactive sputtering process parameters affect film quality and performance is still lacking. In this work, the influence of process parameters—pressure, target–substrate distance, substrate temperature, and substrate RF power—on ScAlN thin films grown by reactive sputtering was experimentally investigated. To optimize the growth conditions, the mechanical, structural, and morphological properties of the films were characterized. Based on this analysis, an optimized deposition recipe was developed for this system and subsequently used to grow ScAlN thin films for detailed electrical and electromechanical characterization.

Nevertheless, most of the reported $d_{33}$ values in the literature have been obtained using a Berlincourt piezometer, which may not adequately account for substrate bending effects and electrode geometry [30,31]. In contrast, the double-beam laser interferometry (DBLI) technique employed in this work provides a more precise characterization of the piezoelectric response while compensating for substrate bending [32,33].

In addition, the transverse piezoelectric coefficient ($d_{31}$) was extracted using a digital holographic microscope (DHM) on a cantilever-based structure, where the tip displacement under electrical excitation was analyzed to compute $d_{31}$ via analytical formulations. Upon extracting the values of $d_{31}$, ${\epsilon }_o$, and $tan\left(\delta \right)$, the figure of merit (FOM) for the cantilever-based system for energy harvesters and sensors was then calculated.

Beyond their enhanced piezoelectric response, ScAlN films have also been shown to exhibit robust ferroelectric behavior [34,35,36,37], unlocking their potential in non-volatile memory and tunable RF components [38,39,40,41,42,43]. In this work, the dielectric (${\epsilon }_r$, $tan\delta$) and ferroelectric ($P_r$, $E_c$) properties of ScAlN thin films deposited under optimized process conditions were also investigated.

2. Deposition & Characterization Procedure

ScAlN thin films were deposited on 100 mm silicon (Si) (100) wafers using an Evatec® Clusterline-200 II magnetron sputtering system operated under high-vacuum conditions ($\sim 5\times {10}^{-8},\mathrm{mbar}$). A 12-inch casted Sc_0.3Al_0.7 alloy target was employed as the sputtering source. Deposition parameters—including chuck temperature, nitrogen (N₂) flow rate, and chuck height—were systematically varied to study their effects on film growth. The target was powered by a pulsed DC source at 5 kW, 100 kHz, with an 84% duty cycle.

To optimize the deposition process, the target–substrate distance (chuck height) was adjusted while other conditions were kept constant. Deposition rates at each chuck height were determined under an N₂ flow of 20−35 sccm. Film thickness—and thus deposition rate—was measured using a Woollam ESM-300 ellipsometer, which evaluates thin-film thickness by analyzing polarization response as a function of thickness and refractive index [44]. Argon (Ar) was excluded to avoid abnormally oriented grain (AOG) formation, as reported in literature [20,24]. Surface morphology and AOG presence were examined using a Zeiss Supra 25 FE-SEM and Park XE-7 AFM. Depositions were performed in pure nitrogen (N₂) at 20–35 sccm while keeping chuck height and temperature constant to study the effect on stress and crystallinity. We characterized the deposited films with a series of measurements. Film stress was measured with a KLA-Tencor FLX-2320 by comparing substrate curvature before and after deposition and applying Stoney’s equation [45]. X-ray diffraction (XRD) was carried out on ScAlN thin films deposited on Si (100) substrates using a PANalytical X’Pert Pro diffractometer with a Philips PW3040/60 X-ray generator and detector. Diffraction patterns were collected using Cu-K$\alpha$ radiation ($\lambda =1.5418A^{\circ }$). 2$\theta$ scans identified the crystallographic planes, while rocking curve (fwhm of the $\omega$-scan curve) measurements assessed the film crystallinity. The correlation between c-axis alignment and piezoelectric performance in ScAlN films, reported by Tholander et al. [46], guided the optimization of deposition conditions.

A temperature sweep (250°C, 300°C, 350°C, and 400°C) was then performed to evaluate the influence of growth temperature on film quality. Once the deposition conditions yielding optimal crystallinity were identified, substrate RF power was adjusted to minimize residual stress. After optimizing the ScAlN deposition process, samples were fabricated for subsequent dielectric, piezoelectric, and ferroelectric characterization.

After optimizing the recipe for ScAlN thin films on Si 100 mm wafers, two different samples were prepared for the dielectric, ferroelectric, and piezoelectric measurements. Sample type-1 was used for dielectric, ferroelectric, and longitudinal piezoelectric coefficient ($d_{33}$) measurements. Figure 1(a) and Figure 1(b) represent the cross-sectional and top-view schematics of Sample type-1, respectively. It comprised a 50 nm blanket Pt bottom electrode, a 750 nm ScAlN layer, and a 50 nm Pt top electrode patterned to define metal–piezoelectric–metal (MPM) capacitors with circular and square geometries of various areas. Electrical access to the blanket bottom electrode was provided by deliberately removing a small region of the piezoelectric layer (by mechanical scratching) and applying indium (In) paste to this exposed area to ensure good grounding.

The longitudinal piezoelectric response of the ScAlN films was characterized using an aixACCT double-beam laser interferometer (DBLI). In this configuration, the $d_{33}$ coefficient was extracted, quantifying the voltage-dependent longitudinal displacement of the piezo layer while compensating for substrate bending. The ferroelectric properties—remnant polarization ($P_r$), and coercive field ($E_c$) of ScAlN film—were studied using the same aixACCT system. Dielectric relative permittivity (${ϵ}_r$) and dielectric loss (tan$\delta$) were also measured using this sample. To evaluate the consistency of the measurement results over different substrate sizes, we fabricated two more samples on 150 mm and 200 mm substrates using the same stacks, methods, and electrode patterns as sample type-1.

Meanwhile, as shown in Figure 1(c),Figure 1(d), sample type-2 was fabricated as an asymmetric laminate configured as a cantilever to extract the transverse piezoelectric coefficient ($d_{31}$). In this work, a cantilever-based approach was employed owing to its simplicity and its suitability for direct extraction of the transverse piezoelectric coefficient ($d_{31}$). The cross-sectional schematic and SEM top-view of sample type-2 are shown in Figure 1(c) and Figure 1(d), respectively. Prior studies have shown that a cantilever architecture provides a straightforward and reliable means of directly estimating the transverse piezoelectric coefficient of the piezoelectric layer in a multilayer stack [7,47].

The fabricated cantilever consists of a rectangular Metal–Piezoelectric–Metal (MPM) stack. Unlike conventional designs that use a $Si{O}_2$ structural layer and symmetric top/bottom electrodes of equal thickness, an asymmetric electrode configuration was employed here to facilitate excitation of the $d_{31}$ bending mode. Specifically, a 200 nm Pt layer was deposited as the bottom electrode, followed by the piezoelectric film, and finally a 100 nm Pt layer as the top electrode.

To account for the influence of these additional layers in the mechanical and electromechanical response, the multimorph model is employed [47]. In this framework, the first step is to determine the location of the neutral axis for bending in the multilayered structure. This neutral axis is obtained by evaluating the torque equilibrium across all constituent layers, expressed as:

$$z_M={\displaystyle \frac{{\sum }_i{z}_i{E}_i{A}_i}{{\sum }_i{E}_i{A}_i}}$$

(1)

where $z_M$ denotes the z-position of the torque neutral axis (with z defined as the direction perpendicular to the beam surface). The summation is taken over all layers in the device stack. For each layer i, $z_i$ represents the position of its centroid, $E_i$ is its Young’s modulus, and $A_i=w_i{t}_i$ is its cross-sectional area, where $w_i$ and $t_i$ are the width and thickness of layer i, respectively. The displacement of the tip of the cantilever can be expressed as:

$$\delta ={\displaystyle \frac{E_p{Z}_p{w}_p{d}_{31}{L}^{2}V}{2{\sum }_i{E}_i{A}_i(\frac{t_i^2}{12}+Z_i^2)}}=m{d}_{31}V$$

(2)

where $Z_i$ is the position of the centroid of the layer i with respect to the neutral axis, and V is the applied voltage. From this relation, the tip displacement can be treated as a quantity that is directly proportional to the applied voltage V. When a DC bias is applied across the cantilever beam, the resulting tip deflection can be recorded to generate a $\delta$–V curve, which is subsequently fitted using a linear regression. The transverse piezoelectric coefficient $d_{31}$ can then be extracted from the fitted slope according to

$$d_{31}={\displaystyle \frac{k_{\mathrm{fitted}}}{m}},$$

(3)

where $k_{\mathrm{fitted}}$ is the experimentally obtained slope of the $\delta$–V relation, and m is the geometric factor determined by the multimorph model.

Furthermore, upon extracting the values of $d_{31}$, ${\epsilon }_o$, and $tan\left(\delta \right)$, the figure of merit (FOM) for the cantilever-based system was calculated. To compare the efficiency of piezoelectric layers in cantilever-based systems for energy harvesters and sensors, a suitable FOM method was defined by Priya [48,49], which can be expressed as:

$$FoM={\displaystyle \frac{d_{31}^2}{{\epsilon }_r*tan\left(\delta \right)}}$$

(4)

This serves as a useful parameter for applications in energy harvesting and low noise sensors, which we use to quantify our films.

3. Results

3.1. Process Optimization

For stress optimization, ScAlN films were deposited on 100 mm Si (100) substrates and subsequently analyzed. Ellipsometric measurements confirmed a film thickness of approximately 500 nm with a deviation of $\pm 5\%$. Figure 2 presents the stress behavior of the films under varying process conditions, including nitrogen flow rate, chuck height, substrate temperature, and substrate RF power.

From Figure 2(a), it can be seen that at nitrogen flow rates of 20 sccm and 25 sccm, the films exhibited slightly compressive stresses of about −20 MPa, while at 30 sccm and 35 sccm, the stress shifted to slightly tensile values of +23 MPa and +11 MPa, respectively. This indicates that increasing the gas flow rate has a minimal influence on the overall stress state of the films. In contrast, varying the target–substrate distance (chuck height) from 10 mm to 30 mm significantly changed the stress from −252 MPa (compressive) to +191 MPa (tensile) (Figure 2(b)). At constant nitrogen flow rate and temperature, the stress tends to become more tensile with increasing target–substrate distance.

As shown in Figure 2(c), the stress remains relatively stable between ${300}^{\circ }\mathrm{C}$ and ${400}^{\circ }\mathrm{C}$; however, a compressive stress of −181 MPa is observed at ${250}^{\circ }\mathrm{C}$. Although the lowest overall stress was obtained at a chuck height of 20 mm, determining the optimal process condition requires considering additional deposition parameters and material characterization results. From Figure 2(d), it is evident that the film stress demonstrates a strong dependence on RF power. Increasing the substrate RF power from 0 to 100 W drives the stress from near-zero values to approximately –1.8 GPa compressive. However, the RF power applied to the substrate does not have any impact on the film’s crystallinity [50]. Therefore, the stress could be optimized by tuning RF power on the substrate without affecting the film quality.

Figure 3(a)–(c) illustrate the variation in the peak position of the ScAlN wurtzite (002) phase under different process conditions. As shown in Figure 3(a)–(c), the (002) diffraction peak shifts toward higher angles with increasing $N_2$ flow rate, target–substrate distance, and temperature, indicating a reduction in lattice spacing of the ScAlN crystals.

The crystallinity of the films was further assessed by analyzing the full width at half maximum (FWHM) values obtained from XRD $\omega$-scans (rocking curves) under different process conditions, as depicted in Figure 3(d)–(f). Figure 3(d) highlights the effect of $N_2$ flow rate on the film quality. The best crystallinity (lowest RC) was observed at 20 sccm, yielding an FWHM value of $1.8^{\circ }$. Higher nitrogen flow rates produced broader peaks (2.8°, 3.4°, and 3.8° for 25, 30, and 35 sccm, respectively), reflecting a deterioration in crystalline quality. Considering the strong correlation between crystallinity and device performance [51], a nitrogen flow rate of 20 sccm was selected for subsequent process optimization.

Figure 3(e) presents the effect of target–substrate distance on crystallinity. Optimal film quality was achieved at distances of 20 mm and 25 mm, yielding a rocking curve (RC) value of $2.0^{\circ }$. Both shorter and larger distances resulted in peak broadening, indicating inferior crystallinity. Hence, a range of 20–25 mm was identified as the preferred chuck height for future depositions to balance stress and crystallinity.

The influence of deposition temperature, shown in Figure 3(f), was found to be relatively minor compared to the other parameters. The optimal temperature was determined to be ${300}^{\circ }\mathrm{C}$, with an RC value of $1.8^{\circ }$.

To identify the crystallographic phases present in the deposited ScAlN film and to verify the reproducibility of the optimized parameters, films were also deposited on Si/Pt substrates. A gonio (2$\theta$) scan was performed for sample type-1, as shown in Figure 4(a), to determine the crystallographic phases. As shown in Figure 4(b), the crystalline quality remained consistent, with RC values ranging from $1.{76}^{\circ }$ to $1.9^{\circ }$ across the 100 mm, 150 mm, and 200 mm substrates, confirming the scalability and uniformity of the deposition process.

The presence of abnormally oriented grains (AOGs) was observed by SEM to assess the surface quality of films deposited on Si (100) substrates. Previous studies report that the density of AOGs increases with Sc content in the film [52,53]. A high concentration of AOGs leads to increased surface roughness, degraded film quality, and a reduction in both piezoelectric response and quality factor in piezoelectric acoustic resonators [54,55].

To understand how the deposition parameters influence AOG formation, the SEM images were analyzed using an image recognition algorithm specifically developed for the quantitative assessment of AOGs [56]. This method enables pixel-level detection of individual grains on the substrate and allows precise calculation of their average area and surface coverage. Figure 5 shows the resulting AOG area distributions. A leftward shift of the peak in the distribution indicates a higher fraction of smaller AOGs. Narrow, sharp peaks correspond to a more uniform grain size, whereas broader curves reflect a wider spread of AOG areas. These quantitative distributions support the qualitative impressions obtained from direct visual inspection of the SEM images. Figure 5(a) and (b) illustrate the transformation from a raw SEM image, where AOGs are visually apparent, to the corresponding pixel-level grain detection map.

Figure 5(c) presents the AOG area probability density as a function of N₂ flow rate. At 30 sccm and 35 sccm, the distribution shifts to smaller areas and becomes narrower compared to 20 sccm and 25 sccm, indicating that higher N₂ flow promotes the formation of smaller grains. However, the overall probability of AOG occurrence is higher at these elevated N₂ flow rates. At 20 sccm, the average AOG size is larger, but the probability of AOG formation is comparatively low. At 25 sccm, the AOG density remains relatively low compared to 30–35 sccm, but the range of AOG areas is broader. Taken together, these observations suggest that 20 sccm N₂ offers a more favorable trade-off than higher flow rates in terms of minimizing AOG formation.

The effect of target–substrate distance is shown in Figure 5(d). Varying the distance from 10 mm to 25 mm has only a modest impact on AOG probability; however, at 30 mm the AOG formation probability increases, with a larger number of smaller AOGs appearing. Temperature exhibits a similarly limited influence on AOG density (Figure 5(e)). At higher temperatures, the overall AOG probability decreases slightly, but the distribution broadens, indicating a wider range of AOG sizes.

In contrast, the RF substrate power sweep, shown in Figure 5(f), has a much more pronounced effect. Increasing the RF power substantially reduces the AOG formation probability, ultimately yielding films that are nearly AOG-free at 100 W. This improvement, however, comes at the cost of increased compressive stress, which reaches approximately −1.8 GPa at 100 W shown in Figure 2(d). Considering this trade-off between surface quality and residual stress, no RF power was applied to the substrate in the final optimized process recipe used in this study.

Overall, the best process conditions for achieving superior crystallinity in ScAlN thin films on 100 mm substrates, considering stress−AOG growth probability tradeoff, were identified with a nitrogen flow rate of 20 sccm, a target–substrate distance of 20 mm, and a deposition temperature of ${300}^{\circ }\mathrm{C}$.

Since one of the goals of this ScAlN film optimization was to design a device that requires a bimorph ScAlN structure, determining the influence of vacuum breaking between successive depositions for each layer growth is a critical factor for film quality. The surface morphology of two ScAlN samples deposited on Si/Pt substrates was analyzed using atomic force microscopy (AFM) over a scan area of $2\mu \mathrm{m}\times 2\mu \mathrm{m}$. In this experiment, two ScAlN layers were sequentially deposited on Si/Pt substrates, separated by an intermediate 50 nm platinum layer, to evaluate the impact of vacuum interruption on film growth. For these depositions, the N₂ flow was set to 20 sccm, the chuck height to 20 mm, and the substrate temperature to 350°C, while no RF power was applied to the substrate.

Figure 6(a) presents the top surface morphology of the second ScAlN layer deposited after breaking the vacuum between the successive deposition steps, whereas Figure 6(b) illustrates the corresponding surface obtained under in situ conditions (continuous deposition without vacuum interruption). The rougher surface observed in Figure 6(a) demonstrates the adverse effect of vacuum breaking on bimorph ScAlN film growth, resulting in larger grain size and reduced grain density.

To mitigate surface roughness and grain enlargement, the deposition temperature was reduced from ${350}^{\circ }\mathrm{C}$ to ${300}^{\circ }\mathrm{C}$. As reported by Jin et al. [57], elevated substrate temperatures as well as low temperatures can suppress uniform columnar growth in AlN films. Figure 6(c) shows the ScAlN film surface, featuring smaller, well-aligned columnar grains with higher grain density and a reduced RMS roughness of 1.6 nm. These improvements confirm that optimized temperature control is essential for achieving high-quality bimorph ScAlN layers. Table 1 refers to the investigation range and optimized process parameters for the ScAlN thin film deposition procedure.

3.2. Dielectric Properties

Figure 7(b),(c) illustrate the relative permittivity (${\epsilon }_r$) and dielectric loss tangent ($tan\delta$) of the ScAlN films deposited on substrates of different diameters. The relative permittivity values obtained from DBLI measurements are ${\epsilon }_r=$ 19.4, 18.2, and 18.0 for the 100 mm, 150 mm, and 200 mm samples, respectively. The dielectric loss tangent showed the highest value ($tan\delta =0.012$) for the 100 mm film, while the 150 mm and 200 mm films exhibited lower losses as $tan\delta =0.004$ and $tan\delta =0.007$, respectively.

The observed variation in dielectric loss directly influences the measured piezoelectric coefficient. As shown in Figure 7(a), samples with lower dielectric loss exhibited higher $d_{33,f}$ values, highlighting the strong dependence between dielectric dissipation and effective piezoelectric response in ScAlN thin films.

3.3. Measurement of Piezoelectric Strain Coefficients

3.3.1. Longitudinal Piezoelectric Coefficient ($d_{33,f}$)

The longitudinal piezoelectric coefficient ($d_{33,f}$) of ScAlN films deposited on substrates of three different diameters was measured using double-beam laser interferometry (DBLI). The extracted $d_{33,f}$ values ranged from 17.6 pm/V to 18.7 pm/V, as shown in Figure 7(a) for 100 mm, 150 mm, and 200 mm substrates. During measurement, a 1 kHz AC voltage was applied to the circular top electrode to induce piezoelectric displacement, while the bottom electrode was grounded. The input voltage amplitude was maintained below the polarization switching threshold, ensuring that the displacement response remained linear with the applied electric field, consistent with conventional piezoelectric behavior shown in Figure 8(a).

However, Yuan et al. [24] demonstrated that the density of abnormally oriented grains (AOGs) degrades the piezoelectric coupling in ScAlN, with films exhibiting more than 100 AOGs per $0.01\mu m^2$ showing a reduction in $d_{33}$ by approximately $10\%$ compared to well-oriented crystals. In the present study, SEM and AFM analysis shown in Figure 5 and Figure 6 revealed an AOG density below this threshold, suggesting that the measured $d_{33,f}$ values are close to those expected for an ideal $S{c}_{0.3}A{l}_{0.7}N$ wurtzite crystal.

3.3.2. Transverse Piezoelectric Coefficient ($d_{31,f}$)

After device release for sample type 2, a DC bias voltage was applied between the top and bottom electrodes, and the resulting out-of-plane displacement at the cantilever tip was measured using digital holographic microscopy (DHM). The measured displacement–voltage response, shown in Figure 7(d), was then fitted to extract the effective transverse piezoelectric coefficient of the film, $d_{31,f}$, at each measurement point using Equation (2).

By repeating this procedure for cantilevers of varying lengths, a clear convergence trend was observed in Figure 7(e): as the aspect ratio ($L/W$) increases, the extracted $d_{31,f}$ shifts from a plane-strain-dominated value (approximately 30% higher than the uniaxial constant $d_{31}$) towards the intrinsic uniaxial material constant of the film, $d_{31}$. This behavior is consistent with the expected behavior reported by Hake et al. [49]. For structures with small $L/W$ ratios, the mechanical response deviates from pure beam behavior and becomes more plate-like, and the effective $d_{31,f}$ converges toward the plane-strain value, approaching $-9\mathrm{pC}/\mathrm{N}$. As $L/W$ increases, the geometry more closely approximates the uniaxial stress state assumed in the beam theory, and the fitted $d_{31,f}$ converges toward the intrinsic material property, reaching approximately $-6.22\mathrm{pC}/\mathrm{N}$.

The theoretical $d_{31}$ value for 30% ScAlN reported by Caro et al. [22] is also plotted in Figure 7(e) (shown as a horizontal red line). This correlation validates both the extraction method and the material parameters used.

3.4. Ferroelectric Properties

Using the same DBLI experimental setup, dynamic hysteresis measurements (DHM) were performed to evaluate the polarization switching characteristics of the ScAlN films. The resulting P–E hysteresis loop (Figure 8(c)) clearly demonstrates ferroelectric behavior, showing an average remnant polarization of $\Delta P_r/2=130\mu \mathrm{C}/{\mathrm{cm}}^2$ and an average coercive field of $\Delta E_c/2=3.5\mathrm{MV}/\mathrm{cm}$. This reversible switching of polarization under an alternating electric field confirms the intrinsic ferroelectric nature of the material. Additionally, the nonlinear Voltage−Displacement curve (Butterfly Curve) presented in Figure 8(b) is evidence of the ferroelectric nature of ScAlN. This graph illustrates the nonlinear piezoelectric response, demonstrating the material’s displacement when subjected to an electric field surpassing the coercive field (Ec). The characteristic butterfly-shaped voltage–displacement curve indicates the ferroelectric behavior of ScAlN. The observed nonlinearity results come from the reorientation of dipoles under electric fields exceeding Ec, where the material demonstrates maximum displacement while considerable ferroelectric domain switching takes place.

To further analyze the switching dynamics, Dynamic Hysteresis Measurement (DHM) and Positive-Up-Negative-Down (PUND) measurements were conducted, as shown in Figure 8(c),(d). As the applied field exceeds the coercive field, domain walls progressively align with the external field direction, producing a characteristic rise in polarization. When the field polarity was reversed, domains switched correspondingly from N-polar to M-polar orientation (or vice versa).

Distinct switching current peaks were observed at the coercive field points in both Figure 8(c),(d), indicating rapid polarization reversal once the energy barrier for domain wall movement was overcome [58]. These peaks arise from displacement currents proportional to the rate of polarization change, confirming the active ferroelectric switching process. A slight asymmetry was observed in Figure 8(c), between the positive and negative coercive fields ($E_c^{+}=3.3\mathrm{MV}/\mathrm{cm}$, $E_c^{-}=3.7\mathrm{MV}/\mathrm{cm}$), consistent with findings by Liu et al. [59], who reported similar asymmetry even in capacitors with identical Al/AlScN/Al electrodes. This suggests that factors such as stress distribution within the film [60] may contribute to asymmetric switching behavior.

In the present study, higher remnant polarization and coercive field values were observed under negative bias, suggesting the presence of a built-in potential induced by residual stress, which impedes switching in the opposite direction. Additionally, as indicated by the green lines in Figure 8(d), leakage current was evident in both bias directions. This leakage contribution likely inflated the apparent remnant polarization value, resulting in the measured $2P_r=260\mu \mathrm{C}/{\mathrm{cm}}^2$. Although the precise quantitative influence of leakage current was not measured in the present PUND analysis, its contribution to the measured polarization value is evident.

Table 2 showcases a comparative evaluation of the dielectric, piezoelectric, and ferroelectric properties of ScAlN along with the figure of merit (FOM).

4. Conclusion

This work has demonstrated a process-optimization methodology for ScAlN thin films grown by reactive pulsed DC magnetron sputtering on Si(100) substrates using a 12 inch casted ${\mathrm{Sc}}_{0.3}{\mathrm{Al}}_{0.7}$ target. The influence of key process parameters −N₂ flow rate, target−substrate distance, substrate temperature, and RF power applied to the substrate on film stress, crystallinity, and surface morphology has been investigated systematically. On the basis of measured stress, XRD rocking curve values along the c-axis (002), and the estimated roughness and probability of abnormally oriented grains (AOG) formation from SEM and AFM analysis, an optimized deposition recipe was established while carefully balancing the trade-offs between stress, crystallinity, and AOG density. Using this optimized recipe (Table 1), two sets of samples were fabricated: one used for dielectric, ferroelectric, and longitudinal piezoelectric ($d_{33,f}$) characterization, and another for transverse piezoelectric ($d_{31}$) measurements. To assess the scalability and consistency of the process, $d_{33,f}$, $tan\delta$, and ${\epsilon }_r$ were measured on films deposited on 100, 150, and 200 mm substrates. DBLI measurements yielded $d_{33,f}$ values in the range of $17.6$–$18.7\mathrm{pm}/\mathrm{V}$, while the relative permittivity ${\epsilon }_r$ ranged from 18 to 19.4 and the loss tangent $tan\delta$ from 0.4% to 1.1% across the three substrate sizes. Cantilever-based $d_{31}$ measurements provided a value of $-6.22\mathrm{pC}/\mathrm{N}$, in good agreement with previously reported simulation data. The optimized ScAlN films also exhibited a ferroelectric response, with an average remanent polarization $\Delta P_r/2=130\mu \mathrm{C}/{\mathrm{cm}}^2$ and coercive fields ranging from $E_c^{+}=3.3\mathrm{MV}/\mathrm{cm}$ to $E_c^{-}=3.7\mathrm{MV}/\mathrm{cm}$. These results collectively demonstrate that the developed process enables the reproducibility of high-quality ScAlN thin film deposition over different wafer-scale substrates, making them strong candidates for next-generation MEMS devices for mass production.