Vertical Etching of Scandium Aluminum Nitride Thin Films Using TMAH Solution

A wide bandgap, an enhanced piezoelectric coefficient, and low dielectric permittivity are some of the outstanding properties that have made ScxAl1−xN a promising material in numerous MEMS and optoelectronics applications. One of the substantial challenges of fabricating ScxAl1−xN devices is its difficulty in etching, specifically with higher scandium concentration. In this work, we have developed an experimental approach with high temperature annealing followed by a wet etching process using tetramethyl ammonium hydroxide (TMAH), which maintains etching uniformity across various Sc compositions. The experimental results of etching approximately 730 nm of ScxAl1−xN (x = 0.125, 0.20, 0.40) thin films show that the etch rate decreases with increasing scandium content. Nevertheless, sidewall verticality of 85°~90° (±0.2°) was maintained for all Sc compositions. Based on these experimental outcomes, it is anticipated that this etching procedure will be advantageous in the fabrication of acoustic, photonic, and piezoelectric devices.


Introduction
Group III-V materials are getting notable attraction for their diverse applications such as microelectromechanical systems (MEMS), piezoelectric transducers, resonators, and radio frequency (RF) acoustic filter devices [1][2][3][4]. Due to some of its promising qualities and simplicity of process integration, aluminum nitride (AlN) is widely employed in piezoelectric MEMS devices [5,6]. AlN can be doped with other metals to increase its piezoelectric properties [7,8], which advanced the success of Sc x Al 1−x N based optoelectronics devices [9][10][11]. At the earlier stage of twenty-first century, Takeuchi et al. used first-principles analysis to determine that wurtzite structure of Sc-IIIA-N alloys can be fabricated [12]. Later, Akiyama et al. demonstrated that by measuring co-sputtered Sc x Al 1−x N films [13][14][15] a piezoelectric coefficient of 27.6 pC/N could be achieved and is more than five times higher compared to AlN [13]. Furthermore, according to the experiments of Wingqvist et al. [16], the electromechanical coupling coefficient value of Sc x Al 1−x N film can be improved by up to 15%, with recent studies showing coupling coefficients exceeding 20%. Thus, using ScAlN thin films with increasing scandium concentration facilitates the fabrication of high-frequency and wideband acoustic devices [17][18][19]. However, Sc x Al 1−x N thin films become challenging to etch as the scandium concentration (x) increases, especially when using reactive ion etching (RIE) or inductively coupled plasma (ICP) etching [20]. Substantial research has been conducted on Sc x Al 1−x N to understand its growth and how different etching techniques can be used to fabricate piezoelectric devices [21]. Like any other group III-V material, Sc x Al 1−x N can be etched by dry or wet etching techniques. One of the known dry etching approaches is ion beam etching, which can be physical or chemical and can result in smooth etched surfaces at a suitable etch rate [22]. Table 1 summarizes experiments that dry etch Sc x Al 1−x N and reports sidewall verticality and etch rate. Luo et al. [23] performed ICP etching with thick S1818 photoresist (PR) as an etch mask. They demonstrated how RF power might control the sample's plasma etching energy, and how the energy of the Cl 2 /BCl 3 /N 2 plasma enhances the sidewall angle of Sc 0.06 Al 0.94 N. James et al. demonstrated that the reactive ion beam etching (RIBE) process of Sc x Al 1−x N etching is superior to ion beam etching (IBE) in terms of etching rate, selectivity, and sidewall angle (73 • ) (See Table 1). The Sc x Al 1−x N etching rate and selectivity degrade when the identical beam parameters (See Table 1) are used without the reactive gas [24]. Wang et al. [18] presented the design, fabrication, and characterization of Sc 0.20 Al 0.80 N thin films used for piezoelectric micromachined ultrasound transducers (PMUTs) by using RIE as an etching process (See Table 1). They claimed that the etched layer has good verticality and concluded that increasing the scandium concentration would enhance PMUT performance [18]. Hardy et al. attempted ICP etching using argon (Ar) rather than the more commonly used nitrogen and found that the etch selectivity is significantly higher for Sc x Al 1−x N relative to AlN and the surface roughness can be kept unchanged after etching process [25]. Furthermore, Shao et al. successfully fabricated a lamb wave resonator with a quality factor of around 1000 with Sc 0.22 Al 0.78 N by achieving a 77 • sidewall [26]. They also concluded that Cl 2 gas increases the anisotropy of the etching process [26] since the dry etching process is mainly Cl 2 /BCl 3 based. However, due to ScCl 3 's poor volatility, performing dry etching of Sc x Al 1−x N can be an ineffective process [20]. On the contrary, it takes more etching power and ion bombardment to get the etching rate back to a considerable level [18,27]. Moreover, it was found that excessive etching into the bottom layer might be a concern for device fabrication, and a lower etching rate as well as high-power consumption are the fundamental causes of poor selectivity to various mask materials, which has accelerated the wet etching trials in the research community [20]. Table 2 summarizes some of the wet etch results of Sc x Al 1−x N reported in literature. The etching rate of Sc 0.15 Al 0.85 N was found to be approximately 50 nm/min (at 60-70 • C) with MIF-319 developer that generally contains 2~5% TMAH [18]. Another work demonstrated that employing 25% KOH at 80 • C, a 500 nm Sc 0.36 Al 0.64 N layer could be etched in only 15 s [28], which indicates KOH as a promising etchant. Several authors highlighted the issues with wet etching lie in the lateral etching behind the mask and the generation of sidewall angle [20,29,30]. Wet etching becomes a highly appealing choice if the lateral etching that occurs throughout the process is not an issue or if it can be minimized to acceptable levels. Another notable observation from Airola et al., is that alkaline etchants will outperform acidic etchants in terms of sidewall roughness and etching rate [20]. They have used different etch masks (Mo/SiO 2 /SiN x ) with 25% TMAH at 80 • C and concluded that the mask did not make any difference in the etching results [20]. In the recent work of Tang et al. [21], vertical etching of Sc 0.125 Al 0.875 N have been demonstrated by using 10% KOH at the temperature of 65 • C. They successfully etched down 800 nm of Sc 0.125 Al 0.875 N, but with a significant amount of lateral etching (approximately 395 nm). In addition, the process is not suitable for etching Sc x Al 1−x N with higher concentrations as it could not result vertical sidewalls [21]. In this work, we present a wet etching process by using TMAH solution for etching Sc x Al 1−x N thin films (x = 0.125, 0.20, 0.40), introducing an intermediary thermal annealing process at 650 • C in a nitrogen atmosphere (The experimental approach and etching results demonstrate a ''Universal Method" to etch Sc x Al 1−x N with sidewall verticality between 85 • and 90 • .

Sc x Al 1−x N Film Deposition
An SPTS Sigma 200 deposition system was used to reactively sputter the Sc x Al 1−x N films onto 150-mm Si <100> wafers. The conditions for Sc x Al 1−x N deposition consist of using 5 kW of target power at 350 • C with a mixture of Ar/N 2 . Abnormally oriented grains (AOG) nucleate during film deposition and the density varies based on deposition conditions as well as the templating surface. The AOGs increase surface roughness (See Figure 1) and prevents uniform patterning over Sc x Al 1−x N for the hard mask needed to define the features to be etched. To quantify the Al/Sc concentration in the films, energydispersive spectroscopy (EDS) was conducted, and the Al/Sc ratios fall within the expected values (See Supplementary Materials Sections S1-S4, Figures S1-S24, Tables S1-S18). The Sc 0.80 Al 0.20 N film results were compared to a study by Esteves et al. [31] that used the same sputter target and deposition system to deposit Al 0.80 Sc 0.20 that had the composition verified by X-ray photoelectron spectroscopy (XPS).

SiO 2 Film Deposition
A SiO 2 film was used as a hard mask (See Figure 2) for etching Sc x Al 1−x N since SiO 2 is comparatively much more accessible, well-known in semiconductor processing, and provides good etch selectivity when removing the mask layer post Sc x Al 1−x N etch. Before the SiO 2 deposition, the samples were cleaned with acetone, isopropyl alcohol (IPA), and diluted (DI) water. Though piranha solution (mixture of H 2 SO 4 and H 2 O 2 ) can be used as an alternative cleaning approach [32][33][34], it was found in our experiments that piranha also etches the Sc x Al 1−x N film. After cleaning, samples were dried up with a nitrogen gun. The SiO 2 layer was deposited using a CHA Mark-40 dielectric evaporator. The base chamber pressure reached 1.8 × 10 −7 Torr before deposition and the filament and beam current was maintained at 0.36 A and 6.3~7.5 mA, respectively. The O 2 backflow (40 sccm) was conducted at the pressure level of 2.5 × 10 −5 Torr. Two different oxide thicknesses of 330 nm and 1 µm were used, and though not shown, the etching quality does not depend upon the mask layer thickness.

Film Deposition
A SiO film was used as a hard mask (See Figure 2) for etching Sc Al N since SiO is comparatively much more accessible, well-known in semiconductor processing, and provides good etch selectivity when removing the mask layer post Sc Al N etch. Before the SiO deposition, the samples were cleaned with acetone, isopropyl alcohol (IPA), and diluted (DI) water. Though piranha solution (mixture of H SO and H O ) can be used as an alternative cleaning approach [32][33][34], it was found in our experiments that piranha also etches the Sc Al N film. After cleaning, samples were dried up with a nitrogen gun. The SiO layer was deposited using a CHA Mark-40 dielectric evaporator. The base chamber pressure reached 1.8 × 10 Torr before deposition and the filament and beam current was maintained at 0.36 A and 6.3~7.5 mA, respectively. The O backflow (40 sccm) was conducted at the pressure level of 2.5 × 10 Torr. Two different oxide thicknesses of 330 nm and 1 μm were used, and though not shown, the etching quality does not depend upon the mask layer thickness.

Film Deposition
A SiO film was used as a hard mask (See Figure 2) for etching Sc Al N since SiO is comparatively much more accessible, well-known in semiconductor processing, and provides good etch selectivity when removing the mask layer post Sc Al N etch. Before the SiO deposition, the samples were cleaned with acetone, isopropyl alcohol (IPA), and diluted (DI) water. Though piranha solution (mixture of H SO and H O ) can be used as an alternative cleaning approach [32][33][34], it was found in our experiments that piranha also etches the Sc Al N film. After cleaning, samples were dried up with a nitrogen gun. The SiO layer was deposited using a CHA Mark-40 dielectric evaporator. The base chamber pressure reached 1.8 × 10 Torr before deposition and the filament and beam current was maintained at 0.36 A and 6.3~7.5 mA, respectively. The O backflow (40 sccm) was conducted at the pressure level of 2.5 × 10 Torr. Two different oxide thicknesses of 330 nm and 1 μm were used, and though not shown, the etching quality does not depend upon the mask layer thickness.

Nickel Mask Preparation and Lift-Off
To pattern the SiO 2 hard mask, a Ni metal film was used as the mask that was patterned using a lift-off process. After SiO 2 deposition, lithography was conducted using an MLA 150 Advanced Maskless Aligner Heidelberg Instrument. During the process, AZ5214E photoresist (PR) is spin coated onto the sample at 5000 rpm, resulting in a thickness of 1.2~1.4 microns to prepare for a nickel lift-off process for SiO 2 patterning. The sample(s) were pre-baked at 110 • C for 4 min, then exposed at 405 nm UV light at 135 mJ/cm 2 , developed in AZMIF-300 for 45~50 s, and rinsed in diluted water for 25-30 s. Subsequently, using e-beam evaporation, a 100-nm thick nickel (Ni) film was deposited, which will eventually serve as a mask during the etching of SiO 2 layer (See Figure 2).
During the Ni-metallization process, the base chamber pressure and evaporation pressure were maintained at 3.3 × 10 −7 Torr, and 1.2 × 10 −6 Torr, respectively. The deposition process was performed at 0.5 Å/s. After Ni deposition, the samples were immersed in an acetone solution for 15 min to remove the lift-off photoresist and expose the required regions for SiO 2 etching. A post-clean was later used on all samples that consisted of an acetone, IPA, and DI water rinse.

ICP Etching for SiO 2 Layer
The SiO 2 layer was etched using an ICP dry etch process after depositing Ni as the hard mask. Initially, the chamber cleaning process was performed for 10 min, where 50 sccm of O 2 was used with 30 mTorr chamber pressure. The SiO 2 etch used CHF 3 (45 sccm) and O 2 (5 sccm) gases as the primary etchants during the process, where ICP, RF power, and chamber pressure were, respectively set at 400 W, 100 W, and 5 m Torr. Etch times were based on the SiO 2 thickness where a 1 min 50 s etch was used for 330-nm thick layer, and 5 min for 1-µm thick layer. Any residual Ni leftover after ICP process was removed away during the wet etching and HF cleaning process.

High Temperature Annealing
In general, the ions generated by the plasma bombardment and the etched surface transmit kinetic energy during any ICP operation of the SiO 2 etch. The etched surface can contain a high degree of impurities and defects that originated during the etch process. Wet etching of Sc x Al 1−x N immediately after etching the SiO 2 layer led to a significant amount of surface roughness (See Supplementary Materials Figure S26). Therefore, an intermediate anneal process (Before Sc x Al 1−x N etching) was implemented to remove the potential embedded impurities within the Sc x Al 1−x N film as well as repair the damaged ions. Airola et al. also demonstrated that high temperature thermal annealing process could be a prospective solution to this challenge [20]. Following the etching of the SiO 2 layer, we placed our samples in the high-temperature annealing furnace chamber in a nitrogen environment using 40-45 sccm of gas flow and annealed for 1 h at 650 • C. There is a possibility to further induce stress into Sc x Al 1−x N during annealing processs, which occurs with the thermal expansion mismatch of the substrate and the Sc x Al 1−x N [20]. Therefore, we employed a very modest temperature gradient (20 • C/min) for ramping up the process and cooling down the samples. One of the major concerns about high temperature annealing is that whether it is degrading the film quality or not. Hence, after thermal annealing we checked our sample(s) in SEM and found that there were no significant visual changes. In fact, the AOGs were also clearly visible (See Supplementary Materials Figure S27).

TMAH Wet Etching
The wet etch step used a 25% concentrated TMAH (TMAH: Water in 1:3 ratio) solution at 78 • C~82 • C to etch the Sc x Al 1−x N (x = 12.5%, 20%, 40%) films. The etching rate of Sc 0.125 Al 0.875 N, Sc 0.20 Al 0.80 N, and Sc 0.40 Al 0.60 N was found to be approximately 365 nm/min, 243 nm/min, and 81 nm/min, respectively. We observed that the etching rate is significantly lower compared to the etching rate presented for AlN (1500 nm/min) in [20]. Like other group III-V materials, Sc x Al 1−x N is expected to form oxides compound while reacting with TMAH [N(CH 3 ) 4 + OH − ]. When Sc x Al 1−x N and the TMAH solution interact, AlN and ScN form separately and further react to form oxides and amphoteric substance. The following chemical reactions are subjected to happen during the wet etch process: Nanomaterials 2023, 13, 274 The formation of the scandium hydroxide residues occurs at a temperature of <85 • C, and it slows down the etching process of Sc x Al 1−x N. During the chemical reaction process, other components such as CH 4 , NO and NO 2 can be formed as well, usually at higher temperatures (>300 • C). Thus, we kept the temperature of the TMAH solution at 80 • C during the etching process to have control of the etching rates of Sc x Al 1−x N. Finally, the hydroxides that eventually remain on the sample,) can be removed using HF solution with DI water.

Results and Discussion
During the wet etch process, we heated the 25% TMAH solution to 80 • C and then immersed the Sc x Al 1−x N samples in the solution. Figure 3a-d shows the wet etch results of the Sc 0.20 Al 0.80 N film after 3 min of etching. The 730-nm thick layer was removed in 3 min, resulting in an etch rate of 243 nm/min with 87 •~9 0 • vertical sidewalls. This result demonstrates that wet etching using TMAH can be used as an alternative to achieve sidewall angles better than 80 • , which can be considered state-of-the-art compared to studies in Table 2.  Tables 1 and 2. For the lower concentration, the verticality resulted to be 88.2 • ± 0.2 • (Figure 4d), which is practically the same for what we found for the Sc 0.20 Al 0.80 N. However, as shown in Figure 4b, the profile of the Sc 0.40 Al 0.60 N shows a verticality of~85 • from the provided SEM image.
As the scandium concentration increases, usually the undercut (lateral) etching increases [20,21]. Chen et al., demonstrated the reason behind the dependency of lateral etch with the AOGs [35]. The experiment demonstrated that AOGs are grains that contain ScAlN unit cells that have their c-axis tilted from the normal direction of the film surface and do not precisely nucleate from the bottom of the film [34]. The SiO 2 mask is resistant to the etching process and the Sc x Al 1−x N film is etched laterally. The result is a suspended SiO 2 layer. Thus, we performed high-temperature annealing (thermal diffusion) in the nitrogen atmosphere to recover the surface damage by minimizing the effect of ion bombardment. Annealing is expected to increase the piezoelectric properties of the samples as well [20,[36][37][38]. In addition to this, AOGs generated during the growth process increases etch resistivity, especially in wet etch processes [20,39]. In our works, we were successful in reducing the lateral etching with almost vertical sidewalls for higher Sc concentrations (See Figure 4a). This was possible thanks to the combination of the optimized annealing process (temperature and nitrogen concentration) and wet etching recipes (TMAH concentration, temperature, etch time).
annealing process (temperature and nitrogen concentration) and wet etching recipes (TMAH concentration, temperature, etch time). One critical observation of this wet etching process is that when the Sc x Al 1−x N sample is withdrawn from the TMAH solution, the low solubility of scandium in alkaline solutions caused the formation of residues. Considering the TMAH chemistry and the work conducted in [20], it is very likely to be that the residues were in the form of ScO x H y on the sample surface [20], which appeared in the form of morphological substance or bumps (Figure 3a,b,d and Figure 4a). As the scandium content increases, the more significant number of residues will be deposited (See Figure 3a and Figure 4a) that lead to a slower etch. Airola et al. mentioned that there is a dependency of the residues on the rinsing method and etching process [20]; and we have also found that continuously stirring the Sc x Al 1−x N sample during the etching process in TMAH reduces the formation of residues. Therefore, after the TMAH etching process, an effective DI water rinsing is recommended to remove the residues or TMAH will continue to keep the reaction ongoing. Although it will not remove the residues completely in this process, an HF-based cleaning (rinsing) process is highly recommended. Before using HF, we soaked the sample in DI water for 15 min. One of the major reasons behind this process is that if TMAH content remains in the samples during HF cleaning, there is a possibility of generating tetramethylammonium (TMA) salts through the reaction shown in (5). One critical observation of this wet etching process is that when the Sc x Al 1−x N sample is withdrawn from the TMAH solution, the low solubility of scandium in alkaline solutions caused the formation of residues. Considering the TMAH chemistry and the work conducted in [20], it is very likely to be that the residues were in the form of ScO x H y on the sample surface [20], which appeared in the form of morphological substance or bumps (Figure 3a,b,d and Figure 4a). As the scandium content increases, the more significant number of residues will be deposited (See Figures 3a and 4a) that lead to a slower etch. Airola et al. mentioned that there is a dependency of the residues on the rinsing method and etching process [20]; and we have also found that continuously stirring the Sc x Al 1−x N sample during the etching process in TMAH reduces the formation of residues. Therefore, after the TMAH etching process, an effective DI water rinsing is recommended to remove the residues or TMAH will continue to keep the reaction ongoing. Although it will not remove the residues completely in this process, an HF-based cleaning (rinsing) process is highly recommended. Before using HF, we soaked the sample in DI water for 15 min. One of the major reasons behind this process is that if TMAH content remains in the samples during HF cleaning, there is a possibility of generating tetramethylammonium (TMA) salts through the reaction shown in (5).

N(CH ) OH + HF → (CH ) F + H O
We used diluted HF solution (3~5%) to remove the hard mask layer (SiO 2 ). As shown in Figures 3b,c and 4b As expected, the TMAH also etches the Si substrate once it etches the Sc x Al 1−x N film and the etching of Si is very aggressive compared to Sc x Al 1−x N (See Figure 4d). This means that it might be possible to use this method to release the Sc x Al 1−x N from the Si substrate if the Sc x Al 1−x N thin film is protected (both upper and underneath) by a layer that is resistant to TMAH. We used diluted HF solution (3~5%) to remove the hard mask layer (SiO ). As shown in Figure 3b Figure   4d). This means that it might be possible to use this method to release the Sc Al N from the Si substrate if the Sc Al N thin film is protected (both upper and underneath) by a layer that is resistant to TMAH.

Conclusions
In this work, we have successfully developed an efficient process that is capable of etching Sc Al N (x = 0.125, 0.20, 0.40), ensuring vertical sidewalls 85°~90° (±0.2°) as well as reducing the degree of undercut. Sc Al N etching results were found to be independent of the SiO2 hard mask thickness, based on the two thicknesses used of 330 nm and 1 μm. We have also analyzed the prospective reasons behind the factors that affect verticality during etching and demonstrated how the annealing process significantly improves the surface damage introduced into the Sc x Al 1−x N by reducing the ion-bombardment effect, which ultimately prevents lateral etching. The more the scandium content increases, the more difficult it is to etch due to the formation of ScO x H y residues with AOG potentially further slowing the wet etch rate. Additionally, the impact of high-temperature annealing can be an additional variable to tune for further optimization of the etching profile. Nevertheless, the reported procedure can yield vertical sidewalls in

Conclusions
In this work, we have successfully developed an efficient process that is capable of etching Sc x Al 1−x N (x = 0.125, 0.20, 0.40), ensuring vertical sidewalls 85 •~9 0 • (±0.2 • ) as well as reducing the degree of undercut. Sc x Al 1−x N etching results were found to be independent of the SiO 2 hard mask thickness, based on the two thicknesses used of 330 nm and 1 µm. We have also analyzed the prospective reasons behind the factors that affect verticality during etching and demonstrated how the annealing process significantly improves the surface damage introduced into the Sc x Al 1−x N by reducing the ion-bombardment effect, which ultimately prevents lateral etching. The more the scandium content increases, the more difficult it is to etch due to the formation of ScO x H y residues with AOG potentially further slowing the wet etch rate. Additionally, the impact of high-temperature annealing can be an additional variable to tune for further optimization of the etching profile. Nevertheless, the reported procedure can yield vertical sidewalls in Sc x Al 1−x N, which is helpful in the fabrication of piezoelectric devices using Sc x Al 1−x N that are sensitive to sidewall angles.
Funding: This research was partially funded by the Department of Energy, Sandia Laboratories Academic Alliance. Sandia National Laboratories is a multi-mission laboratory managed and operated by National Technology & Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy's National Nuclear Security Administration under contract DE-NA0003525. This paper describes objective technical results and analysis. Any subjective views or opinions that might be expressed in the paper do not necessarily represent the views of the U.S. Department of Energy or the United States Government.

Data Availability Statement:
The data reported in this manuscript are available on request from the corresponding author.
subjective views or opinions that might be expressed in the paper do not necessarily represent the views of the U.S. Department of Energy or the United States Government.

Conflicts of Interest:
The authors declare no conflict of interest.