We report on the molecular beam epitaxial growth and characterization of polarity-controlled single and multi-layer Scandium Aluminum Nitride (ScAlN) transduction structures grown directly on ScAlN templates deposited by physical vapor deposition (PVD) on Si(001) substrates. It is observed that direct epitaxial growth on PVD N-polar ScAlN leads to the flipping of polarity, resulting in metal (M)-polar ScAlN. By effectively removing the surface impurities, e.g., oxides, utilizing an in situ gallium (Ga)-assisted flushing technique, we show that high quality N-polar ScAlN epilayers can be achieved on PVD N-polar ScAlN templates. The polarity of ScAlN is confirmed by utilizing polarity-sensitive wet chemical etching and atomic-resolution scanning transmission electron microscopy. Through interface engineering, i.e., the controlled formation or removal of surface oxides, we have further demonstrated the ability to epitaxially grow an alternating tri-layer piezoelectric structure, consisting of N-polar, M-polar, and N-polar ScAlN layers. Such multi-layer, polarity-controlled ScAlN structures promise a manufacturable platform for the design and development of a broad range of acoustic and photonic devices.
The significant improvement in piezoelectric response achieved through the alloying of aluminum nitride (AlN) with scandium (Sc) has established ScAlN as the preferred material for a diverse range of piezoelectric applications, ranging from acoustic devices to next-generation micro-electromechanical systems (MEMs), sensors, energy harvesting, and beyond.1–7 Wurtzite ScAlN possesses metal (M) or nitrogen (N) polarity.8,9 Stacking multi-layer ScAlN with controlled polarity is critical for a broad range of device applications. For example, multi-layer periodically poled piezoelectric film (P3F) ScAlN enables the exploitation of high order modes to achieve high Q with comparatively large film thickness for acoustic resonators, thereby overcoming the frequency-thickness scaling requirements for the complex design of modern wireless communication systems.10–13 Previous approaches for achieving P3F largely rely on post-growth poling, mechanical transfer, and stacking of multi-layer structures. For example, Nam et al. reported an AlN/ScAlN/AlN multi-layer multi-mode bulk acoustic wave resonator (FBAR), wherein a higher order operation at 31 GHz is achieved by switching the polarity of the ScAlN layer post-growth.14,15 Kramer et al. reported the achievement of 50 GHz operation in a lithium niobate (LiNbO3) tri-layer structure through the mechanical transfer of individual layers with opposite polarities.16 Alternatively, the ability to epitaxially grow multi-layer structures with controlled polarity will provide a more manufacturable platform for the design and development of P3F resonators as well as many other emerging acoustic and nonlinear photonic devices.
To date, a wide range of techniques have been used to grow ScAlN, ranging from physical vapor deposition (e.g., sputtering) to epitaxial growth [e.g., molecular beam epitaxy (MBE) and metalorganic chemical vapor deposition (MOCVD)] on various substrates, including Si, SiC, sapphire, Mo, and Al.2,17,18 Owing to the differences in the migration length of impingent atoms with varying energies across different techniques, the resulting grains differ in sizes and orientation, which further translates into variations in material properties.19–21 Sputtering exhibits minimal sensitivity to epitaxial alignment with the underlying substrate, thus offering higher substrate flexibility; however, the grown films often have a fixed N-polarity.18 Recent studies have shown that the surface polarity of III-nitrides can be precisely controlled by utilizing MBE or MOCVD.22–24 It is therefore envisaged that epitaxial growth utilizing MBE and/or MOCVD can enable the achievement of multi-layer polarity controlled ScAlN transduction structures,25 which offers a more manufacturable and scalable platform, compared to conventional mechanical transfer, for applications including piezoelectric and photonic devices.
In this study, we report the MBE growth and characterization of polarity-controlled single and multi-layer ScAlN transduction structures directly on N-polar ScAlN templates grown on a Si(001) substrate by PVD. By employing an in situ Ga flushing technique to effectively remove the surface oxide, we first show that high quality N-polar MBE ScAlN can be grown directly on N-polar PVD ScAlN. By utilizing a combination of controlled formation and removal of surface oxidation, we further demonstrate the achievement of tri-layer ScAlN transduction structures consisting of alternating polarities. This unique growth process can be extended to polarity-controlled piezoelectric or ferroelectric transduction structures with virtually unlimited number of layers, thereby opening a new route for the scalable manufacturing of a new generation of piezoelectric and/or ferroelectric composite platforms for a broad range of device applications, including high frequency resonators and filters and nonlinear photonics.
An N-polar ScAlN layer was first grown on a Si(001) substrate by sputter deposition at a temperature of 350 °C. Detailed growth parameters for the sputtered ScAlN film include a deposition rate of 2.8 Å/s, a substrate temperature of 350 °C, N2 flow rate of 27 sccm, and a DC power of 1000 and 545 W for Al and Sc, respectively.
The epitaxial growth of ScAlN was then performed utilizing a Veeco GENxplor MBE system equipped with a radio frequency (RF) plasma-assisted nitrogen source. Initially, the PVD grown ScAlN samples were treated with Buffered Hydrofluoric Acid (BHF) for 5 min to remove the surface oxides and immediately loaded into the MBE load-lock chamber. Prior to loading into the growth chamber, two outgassing processes were performed in the load-lock chamber and preparation chamber at temperatures of 200 and 600 °C, respectively. A significant obstacle in effectively implementing the MBE regrowth technique lies in overcoming the presence of the interfacial oxide layer on ScAlN.26,27 Due to the substantial oxygen affinity of Sc, an inevitable oxidation process occurs at the surface of ScAlN when exposed to air, forming an oxynitride layer.27,28 Traditionally, in situ surface cleaning before MBE growth involves thermal desorption of the oxide layer.29,30 However, such a technique alone is inadequate to remove surface oxide from ScAlN. Here, we conducted a detailed investigation of the role of Ga-assisted flushing in effectively removing the surface oxide from PVD grown ScAlN films.31,32 Compared to the commonly used Al-assisted flushing process to remove oxygen impurities from the surfaces of AlN and related alloys,33 the Ga-assisted flushing process can take place at a much lower substrate temperature, thereby minimizing any negative impact of high temperature on the crystal structures and associated material properties.31
Our Ga-assisted surface cleaning consists of multiple cycles of Ga adsorption and desorption on the surface of ScAlN in the growth chamber, wherein the substrate temperature is fixed at 650 °C. During each Ga adsorption/desorption cycle, the substrate was exposed to a Ga flux with a beam equivalent pressure (BEP) of ∼2 × 10−7 Torr for a duration of 60 s. This Ga-assisted flushing procedure was performed without active nitrogen species to prevent any inadvertent formation of GaN. The process of Ga adsorption and desorption was clearly observed by monitoring the time evolution of the reflection high-energy electron diffraction (RHEED) intensity. Figure 1(a) shows a schematic representation of the ScAlN PVD template with the presence of surface oxide, while Fig. 1(b) shows the diffused RHEED patterns from the oxidized ScAlN surface. Figure 1(c) shows a schematic illustration of the Ga-assisted surface oxide removal from ScAlN. For the first three cycles of flushing, sufficient time (3 min) was allowed for the Ga to desorb, following which a 1 min/1 min cycle was performed. This process was repeated for a total of 20 cycles.
The RHEED spot intensity first decreases upon Ga deposition and then gradually rises until reaching saturation as Ga desorbs. The recovery time (τr, defined as the time between Ga shutter close to the time when the RHEED intensity is the brightest) reduces with an increasing number of flushing cycles, whereas the RHEED intensity at the end of the desorption phase monotonically increases with an increasing number of flushing cycles. There is minimal variation in the evolution of the RHEED intensity during the last five cycles, which we interpret as an indication of the surface being adequately clean. Figure 1(d) shows bright-segmented RHEED patterns at the end of the 20th flushing cycle. The spots are more prominent with enhanced intensity post-Ga-assisted flushing, providing clear evidence of the removal of surface oxides.34 The evolution of the recovery time in this two-step flushing process has been illustrated as a function of flushing cycles in Fig. 1(e).
Aside from the surface oxide formation, the re-growth on the polycrystalline ScAlN PVD film is challenging, owing to the absence of an epitaxial registry. The ScAlN growth was performed in an N-rich regime to avoid the formation of the Sc–Al intermetallic and Sc3AlN perovskite phase while maintaining the Sc and Al fluxes for a Sc composition of ∼0.2.35 The growth temperature was systematically tuned in the range of 400–600 °C for the epitaxial growth of the ScAlN layer on the N-polar PVD template, as shown in Fig. 2(a). The RHEED pattern was monitored and recorded in situ, as shown in Figs. 2(b)–2(d). The post-growth RHEED for the ScAlN layer at 400 °C shows the formation of ring-like patterns [Fig. 2(b)], whereas the RHEED pattern similar to that of the PVD ScAlN template is preserved for growth temperatures of 500 and 600 °C [shown in the insets of Figs. 2(c) and 2(d)], suggesting that a relatively high growth temperature is needed to promote the formation of crystalline ScAlN in this study. Similar ScAlN films were also grown on PVD templates with a reduced thickness of ∼50 nm, and no significant differences were observed compared to the 100 nm templates discussed in this study. However, while growing MBE films on 5 and 15 nm ScAlN PVD templates, the RHEED before the growth is almost invisible and does not improve upon Ga-flushing. Moreover, after the MBE growth, the RHEED did not show the regular wurtzite pattern, but rather showed clear ring-like features indicating poor material quality.
The surface morphology of the films was analyzed using a Bruker ICON atomic force microscopy (AFM) system and a Hitachi SU8000 scanning electron microscope (SEM), as shown in Figs. 2(b)–2(e). It is observed that the film grown at 400 °C exhibits a comparatively rough morphology, with a root mean square (rms) roughness of ∼1.8 nm. In contrast, the rms roughness for the films grown at 500 and 600 °C is ∼0.9 and 1 nm for a 1 × 1 μm2 scanning area, respectively. The surface roughness of the PVD template did not show any significant change when exposed to such temperatures for prolonged durations. Figure 2(e) shows the AFM image of the best surface morphology obtained at a growth temperature of 500 °C. X-ray diffraction (XRD) 2theta–omega (2θ/ω) and omega (ω) scans, i.e., rocking curve (XRC), were performed using a Rigaku SmartLab diffractometer with a Cu Kα1 radiation x-ray source (1.5406 Å) to examine the crystallinity of ScAlN films. Figure 2(f) presents the (0002) plane XRD 2θ/ω for the optimized regrown ScAlN sample at 500 °C and the PVD template. Figure 2(g) shows the comparison of the (0002) plane XRC full-width half maximum (FWHM) measured for the samples grown at different temperatures. The prominent characteristic peak at 36.07° in the 2θ/ω scan confirms the wurtzite structure. The underlying PVD template has an FWHM of 3.3°, whereas, for the MBE regrown films, the FWHM first decreases with increasing growth temperature, reaching a minimum value of 2.5° at 500 °C, beyond which it degrades to 3°. A similar temperature dependent linewidth behavior has been reported earlier for both AlN and ScAlN thin films, with significant lattice imperfections beyond a certain growth/annealing temperature.36,37 Given that the FWHM obtained for the MBE regrown ScAlN films encompasses the contribution from the underlying PVD ScAlN template (with an FWHM of 3.3°), its absolute value may not hold significant relevance. The relative value, however, indicates significantly improved crystallinity for MBE grown ScAlN, as shown in previous studies.21,35,38–41
Previous studies have underscored the significant role of surface-adsorbed oxygen in the transition from N-polarity to Al-polarity during AlN growth, across various growth techniques such as MBE, metalorganic vapor phase epitaxy (MOVPE), and sputtering.42,43 In the following experiments, we delve deeper into this phenomenon within our samples, aiming for a thorough examination of how Ga-assisted flushing can influence and potentially alleviate this polarity inversion. Wet chemical etching using a dilute solution of tetramethylammonium hydroxide (TMAH) or potassium hydroxide (KOH) is a commonly used technique to determine the polarity of a III-nitride film; the N-polar surface is more reactive and can be readily etched, whereas its M-polar counterpart is typically not affected by the etching process.44 We first examine two samples shown in Fig. 3: sample A (100 nm as-grown PVD ScAlN template) and sample B (100 nm MBE ScAlN grown on 100 nm PVD ScAlN with Ga flushing).
All the samples were patterned with standard photolithography to form openings, which were then exposed to a 20% TMAH solution for 10 min. The region exposed within the PVD ScAlN template underwent etching, providing clear confirmation of its N-polarity, shown in Figs. 3(a) and 3(b). In the case of sample B, grown by an initial Ga-assisted flushing, the exposed region gets etched away upon TMAH treatment, thus indicating the retention of N-polarity within the MBE-grown film, consistent with the underlying template, as shown in Figs. 3(c) and 3(d). This finding further strengthens our argument regarding the efficacy of the Ga-assisted flushing process in effectively removing surface-adsorbed oxygen contaminants, consequently ensuring the preservation of desired polarity during the subsequent regrowth of ScAlN via MBE.
To further confirm the polarity inversion and retention phenomenon at the atomic scale, cross-sectional transmission electron microscopy (TEM) samples were prepared using a FEI Helios NanoLab 600i DualBeam scanning electron microscope/focused ion beam (SEM/FIB). High-angle annular dark-field (HAADF) imaging in scanning TEM (STEM) was performed using a Themis Z (Thermo Fisher) Cs probe corrected microscope with an acceleration voltage of 200 kV and a probe semi-convergence angle of 22 mrad. For differentiated differential phase contrast (dDPC) imaging, a collection angle of 11–63 mrad and a beam current of 100 pA were used. A radial Wiener filter was applied to enhance the contrast of the dDPC images, thereby better resolving the atomic structure. dDPC is an extension of the integrated DPC (iDPC) STEM and conventional DPC STEM techniques.45,46 DPC utilizes a four quadrant segmented detector to collect signal from the STEM probe. dDPC images are calculated as the differential of the collected DPC signal such that the dDPC signal is equal to the Laplacian of the iDPC signal.47,48 In dDPC images, the contrast is reserved compared to HAADF, with the dark circles representing atomic columns and the light background representing the vacuum.
Figure 3(e) shows a HAADF-STEM image of the MBE/PVD ScAlN heterostructure grown with the proposed Ga-assisted flushing. A high magnification dDPC image of the MBE film, recorded from the yellow box in Fig. 3(e), is shown in Fig. 3(f), clearly revealing N-polarity in the MBE film, consistent with the polarity of the underlying PVD template. This further corroborates the efficacy of the proposed Ga-assisted flushing method in preserving polarity at the atomic scale.
We have subsequently investigated the MBE growth of M-polar ScAlN on the N-polar ScAlN template, achieved by utilizing the oxygen-assisted polarity reversal at the interface of the PVD/MBE film, as shown in Fig. 4(a). To explain the role of surface adsorbed oxygen in flipping the polarity, it should be noted that at lower concentrations of O2, a defect cluster consisting of [VAl+ON] is formed, which appears to have a minimal effect on the stability of the wurtzite structure, preventing polarity inversion. However, as the O2 concentration increases, the concentration of [VAl+ON] defect clusters also rises, reaching a critical point where the aluminum coordination shifts from a tetrahedral to an octahedral geometry.22,49 This transition promotes the formation of inverse domain boundaries (IDBs), which ultimately facilitate the inversion of polarity. Zhang et al. reported the oxygen-assisted polarity flipping behavior in AlN films, with a well-defined inversion domain boundary (IDB) ∼2 nm thick.42 This IDB can be described by two interpenetrating N-polar and Al-polar wurtzite lattices sharing a common anion sub-lattice.43
When examining the sample by a similar wet etching technique for polarity confirmation, no discernible patterns emerged following the etching process, as evidenced in Fig. 4(b). This outcome serves to solidify the proposition that the presence of surface-adsorbed oxygen plays a pivotal role in inducing a polarity inversion within MBE-grown ScAlN, transitioning from its native N-polarity to M-polarity across the interface aligning closely with previous findings.42,43,50
Going a step further, we investigated the feasibility of growing a tri-layer ScAlN structure with controlled polarity for each layer by utilizing this oxygen-assisted polarity flipping technique. Initially, an M-polar MBE layer was grown atop the N-polar PVD template without Ga-assisted flushing, following which the sample was taken out and exposed to air for 2 days to form a surface oxide layer. The sample was reloaded in the MBE chamber for regrowth. As the MBE regrowth is done without any Ga flushing, the polarity is expected to flip owing to the surface adsorbed oxygen, i.e., leading to the formation of N-polar ScAlN across the interface.
The expected polarities of this tri-layer ScAlN structure are further confirmed by atomic-resolution STEM studies. Figure 4(c) displays a low-magnification HAADF-STEM image, revealing three ScAlN layers: a nitrogen polar (N-polar) PVD film, a metal polar (M-polar) MBE film, followed by an N-polar MBE film on top. High-magnification dDPC images in Figs. 4(d)–4(f), recorded from the colored squares marked in Fig. 4(c), were utilized to resolve both the metal and nitrogen atomic columns and therefore determine the local polarity of the sample in the three layers. Shown in Figs. 4(d) and 4(e) are the dDPC images of the bottom PVD ScAlN film (red box) and the adjacent MBE film (blue box), revealing the polarity change from N-polar to M-polar. There is a clear swap in nitrogen atomic column placement indicating the polarity switch. Figure 4(f) shows a high magnification dDPC image of the interface between the two MBE films (pink box). The bottom MBE film remains entirely M-polar, while the top MBE film shows a partial change in polarity with both M- and N-polar domains being visible. The partial polarity conversion is attributed to the uncontrolled and non-uniform nature of oxidation by exposure to air. Thus, it can be inferred that the surface adsorbed oxygen plays a critical role in flipping the polarity of the subsequently grown layer. Beyond a certain concentration of the surface adsorbed oxygen, the oxynitride layer at the interface gives rise to inversion domain boundaries (IDBs), as has been commonly observed in AlN.42,43 Given the oxygen affinity of Sc and Al, it is almost impossible to stop the formation of the native oxide layer on top of the exposed ScAlN; however, the Ga-assisted flushing technique proposed in this work offers a route to effectively remove the interfacial oxide, thus preventing the formation of IDBs during the epitaxy. Moreover, the polarity could be flipped back from M-polar to N-polar with the formation of surface oxides (e.g., exposure to air in this study). Controllably switching polarization in multilayer composite ScAlN structures offers a new avenue to tailor the performance and functionality in acoustic devices.14,15
In summary, we have demonstrated multi-layer ScAlN transduction structures with controlled polarity for each constituting layer. This was achieved through interface engineering; the process of surface oxidation (or the incorporation of other impurity adatoms, e.g., Mg and Si, at the growth interface) enables the achievement of controlled polarity switching in multilayer composite structures. Such multi-layer piezoelectric, ferroelectric, and/or dielectric composite structures are important for a wide range of device applications. For example, they can streamline the fabrication process and boost device performance for acoustic resonators and filters to support multiple harmonic modes without compromising performance across various frequencies. The controlled polarity of the ScAlN multi-layer structures, as demonstrated in this work, also promises an effective route to develop scalable ScAlN based P3F technology with enhanced electromechanical coupling and power handling capabilities, as well as multi-level memory cells with significantly enhanced storage capacity and functionality. This unique approach can be further extended to substrates beyond Si, such as SiC, sapphire, and various metal templates/substrates. In the present study, the combination of MBE and PVD methods is utilized to achieve multi-layer ScAlN transduction structures; however, such polarity-controlled multi-layer structures can also be realized by utilizing MOCVD or other growth techniques. In addition, the ability to achieve a pristine growth interface with the use of the in situ Ga-assisted flushing method holds potential for broader applications in epitaxial regrowth processes beyond ScAlN, including but not limited to Ga(Al)N ohmic contact regrowth and patterned channel regrowth, within electronic devices such as high electron mobility transistors (HEMTs). Such application ensures the achievement of an oxide-free initial growth surface, thereby enhancing the crystalline quality of the regrown layer significantly.
This work was supported by the Department of Defense Advanced Research Projects Agency (Award No. HR00112390018), Army Reserach Office (Award No. W911NF-24-2-0210), and College of Engineering, University of Michigan.
AUTHOR DECLARATIONS
Conflict of Interest
The authors have no conflicts to disclose.
Author Contributions
Shubham Mondal: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal). Eitan Hershkovitz: Data curation (equal); Investigation (equal); Writing – original draft (equal); Writing – review & editing (equal). Garrett E. Baucom: Data curation (equal); Investigation (equal); Writing – original draft (equal). Md Mehedi Hasan Tanim: Data curation (equal). Shaurya Dabas: Investigation (equal); Methodology (equal). Baibhab Chatterjee: Funding acquisition (equal); Validation (equal). Honggyu Kim: Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Writing – review & editing (equal). Roozbeh Tabrizian: Funding acquisition (equal); Resources (equal). Zetian Mi: Conceptualization (lead); Funding acquisition (equal); Methodology (equal); Writing – review & editing (equal).
DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding author upon reasonable request.