The Structural Evolution of Semipolar (11−22) Plane AlN Tem-Plate on m-Plane Sapphire Prepared by Sputtering and High Temperature Annealing

In this work, the epitaxial semipolar (11–22) AlN was prepared on nonpolar m-sapphire substrate by combining sputtering and high-temperature annealing. According to our systematic measurements and analysis from XRD, Raman spectra, and AFM, the evolution of crystalline structure and morphology was investigated upon increasing AlN thickness and annealing duration. The annealing operation intensively resets the lattice and improves the crystalline quality. By varying the film thickness, the contribution from the AlN-sapphire interface on crystalline quality and lattice parameters during the annealing process was investigated, and its contribution was found to be not so obvious when the thickness increased from 300 nm to 1000 nm. When the annealing was performed under durations from 1 to 5 h, the crystalline quality was found unchanged; meanwhile, the evolution of morphology was pronounced, and it means the crystalline reorganization happens prior to morphology reset. Finally, the annealing treatment enabled a zig-zag morphology on the AlN template along the sapphire [0001] direction in the plane, which potentially affects the subsequent device epitaxy process. Therefore, our results act as important experience for the semipolar nitride semiconductor laser device preparation, particularly for the epitaxy of microcavity structure through providing the crystalline evolution.


Introduction
The spread of COVID-19 has intensively boosted the development of high-efficiency sterilization methods; particularly, the pandemic explosion draws considerable attentions into AlGaN-based ultraviolet-C (UVC, <280 nm) luminescent devices due to their efficiency towards coronaviruses [1,2]. Although the AlGaN based UVC-luminescent devices show encouraging superiority in the aspects of fast sterilization, nontoxicity, flexible installability, and portability [3], the wall-plug efficiency (WPE) is still less than 10%, which is still far away from the value of GaN-based blue-LED [3], and it is ascribed to both material and physical reasons. From the view of material, the demand of UVC-transparent meanwhile lattice-match acts as the prerequisite for UVC-LED epitaxy, and as a result, only bulk AlN Materials 2022, 15, 2945 2 of 10 substrate and AlN/Sapphire template are qualified. However, the directly grown AlN on flat sapphire usually exhibits thread dislocation density (TDD) above 10 10 /cm 2 , and those dislocations would penetrate into the quantum well region and annihilate the emission. Fortunately, such an obstacle was subsequently solved by the epitaxial lateral overgrowth (ELOG) and high-temperature annealing (HTA) techniques, which both effectively suppress the TDD down to the 10 7~8 /cm 2 level [4,5]. Whereas, from the physical aspect, one of the dominant obstacles is the strong polarization of the nitride semiconductor along the [0001] axis, and it causes a strong quantum-confined stark effect (QCSE), which prohibits the overlap of hole and electron wavefunction [6][7][8][9]. Actually, the QCSE-caused emission degradation has been solved by utilizing nonpolar or semipolar templates; however, such strategies have been limitedly carried out in AlGaN-based UVC-LEDs which suffer much more from the QCSE than in conventional GaN-based blue-LEDs due to increased Al content-enhanced polarization. In addition, in order to ensure the quality of the epitaxial UVC-luminescent device structure, e.g., the microcavity and Bragg reflector region in the laser device, an outstanding crystalline quality is necessary.
Therefore, exploring the avenue which spontaneously solves above-mentioned issues is of significance for further improving performance, in particular the emission power of AlGaN-based UVC-LEDs. The preparation of high crystalline quality nonpolar or semipolar AlN templates acts as a promising candidate to settle the bottleneck, which has been pursued by several groups [10][11][12][13][14]. Particularly, in order to achieve outstanding crystallinity, the strategy of HTA is preferred rather than ELOG due to the challenge of manipulating epitaxy orientation during the MOCVD growth [13,[15][16][17]. However, the systematic investigation on the evolution of strain inside the HTA AlN layer on m-plane sapphire is still limited.
In the present work, by fully utilizing the magneto-sputtering combined with the hightemperature annealing technique, we obtained the single-crystalline semipolar (11−22) AlN template on nonpolar m-plane sapphire substrate. Moreover, the structural information was systematically studied to explore the effect from HTA, including the crystallinity as well as morphology. In particular, by varying the thickness and annealing time, the changing of crystallinity and structure were both tracked to conclude the influence from the interface. As a consequence, our results act as the solid and meaningful experience for lateral studies on semipolar UVC-luminescent devices, particularly with specific structures, e.g., microcavity in the laser device.
In addition to XRD results, the Raman spectroscopy is also a powerful tool to detect the crystalline information in the lattice. As a reference, a Raman spectrum of m-plane sapphire is present as shown in Figure 4a. According to previous studies, we can see the phonon scatterings from the AlN ranged from 600 cm −1 to 700 cm −1 . After subtracting the E(g) Raman signal from m-sapphire at 645 cm −1 , for the 300 nm as-grown AlN film, the phonon-scattering peak of E 2 H was the most intensive, while only the signs of A 1 (TO) and E 1 (TO) modes were present. The high-temperature annealing fully activated the A 1 (TO) and E 1 (TO) signals due to the improvement of crystalline quality [21]. In addition, when compared with the as-grown sample, the E 2 H peak exhibited a blueshift from 654 cm −1 to 663 cm −1 in the annealed sample. Such a Raman blueshift again confirmed a transition from tensile strain in the as-grown sample into compressive strain in the annealed sample [15], which is consistent with the XRD description in Figure 3. With increasing AlN thickness, the A 1 (TO) peak became more and more intensive and separated in the as-grown sample. Despite the intensity of the E 1 (TO) mode being continuously enhanced as the thickness rose, the signal still mostly merged with the E 2 H signal. The annealing successfully separated the three phonon-scattering peaks when the thickness gradually increased. Actually, in addition to the E 2 H mode in the 300-nm sample, the other two AlN-scattering modes both showed blueshift behavior resulted from the annealing, again verifying the evolution of strain statue.
creased. Actually, in addition to the E2 H mode in the 300-nm sample, the other two AlN-scattering modes both showed blueshift behavior resulted from the annealing, again verifying the evolution of strain statue. The high-temperature operation largely reordered the lattice matrix during the annealing process. Therefore, the evolution of lattice constant could be captured during the annealing process. In order to track the crystallinity improvement, the 2theta-omega scans of samples annealed from 1 to 5 h are shown in Figure 5. It is clearly seen that the annealing time as short as only one hour was more than enough to sharpen the diffraction peak of (11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22) plane, and it suggests that the crystalline quality reset happens very soon when the annealing temperature is above the threshold of lattice reorder. However, as the annealing goes on, the position of the (11-22) diffraction peak successively experienced a shift during the 5 h, and it indicated that the lattice expands in the initial 2 h and subsequently continuously shrinks to 2.645 Å until 5 h, which is observed in Figure 5. Anyway, the lattice parameter is larger than the value of bulk AlN. The high-temperature operation largely reordered the lattice matrix during the annealing process. Therefore, the evolution of lattice constant could be captured during the annealing process. In order to track the crystallinity improvement, the 2theta-omega scans of samples annealed from 1 to 5 h are shown in Figure 5. It is clearly seen that the annealing time as short as only one hour was more than enough to sharpen the diffraction peak of (11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22) plane, and it suggests that the crystalline quality reset happens very soon when the annealing temperature is above the threshold of lattice reorder. However, as the annealing goes on, the position of the (11-22) diffraction peak successively experienced a shift during the 5 h, and it indicated that the lattice expands in the initial 2 h and subsequently continuously shrinks to 2.645 Å until 5 h, which is observed in Figure 5. Anyway, the lattice parameter is larger than the value of bulk AlN. The crystalline quality evolution upon increasing annealing duration was reexplored by XRD RCs as shown in Figure 6. As shown in Figure 6, the crystalline quality evaluated by RC curves did not gradually improve in the time scale of hours, and the 1 h annealed sample showed the (11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22) plane RC FWHM as low as 1080 arcsec. However, when the The crystalline quality evolution upon increasing annealing duration was reexplored by XRD RCs as shown in Figure 6. As shown in Figure 6, the crystalline quality evaluated by RC curves did not gradually improve in the time scale of hours, and the 1 h annealed sample showed the (11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22) plane RC FWHM as low as 1080 arcsec. However, when the annealing duration was 3 h, the value increased back up to 2080 arcsec. When a larger annealing duration was performed above 3 h, the FWHM returned down to 810 and 930 arcsec when it was annealed under 4 and 5 h, respectively. In addition to the RC results, the Raman spectra were also measured to explore the stress evolution as plotted in Figure 7. Actually, corresponding to the RC results, only 1 h annealing was more than enough to expose the incisive E 2 H and E 1 (TO) signals, indicating that only one-hour-high temperature treatment successfully recrystallized the sample. The crystalline quality evolution upon increasing annealing duration was reexplored by XRD RCs as shown in Figure 6. As shown in Figure 6, the crystalline quality evaluated by RC curves did not gradually improve in the time scale of hours, and the 1 h annealed sample showed the (11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22) plane RC FWHM as low as 1080 arcsec. However, when the annealing duration was 3 h, the value increased back up to 2080 arcsec. When a larger annealing duration was performed above 3 h, the FWHM returned down to 810 and 930 arcsec when it was annealed under 4 and 5 h, respectively. In addition to the RC results, the Raman spectra were also measured to explore the stress evolution as plotted in Figure  7. Actually, corresponding to the RC results, only 1 h annealing was more than enough to expose the incisive E2 H and E1(TO) signals, indicating that only one-hour-high temperature treatment successfully recrystallized the sample.  In order to explore the surface morphology of semipolar AlN before and after 5 h high-temperature annealing, the 2 × 2 μm 2 AFM measurements were carried out as shown in Figure 8. Before the annealing, the as-grown sample presented stripe-like morphology, which is vertical to the sapphire c axis direction, due to the orientation of AlN crystallization. According to previous studies, such an orientation-preferential growth mode has been observed by the TEM method [15,16], and it originates from the anisotropic lateral growth rate of the grains [22]. Moreover, such a striation morphology which is along the sapphire [0001] direction seems to be strongly related with the grown temperature [15], and our preparation condition just agreed with the qualification. The width and vertical distance of one wire-like structure are ~20 nm and ~1.4 nm, respectively. The calculated root mean square (RMS) of the as-grown sample is 0.388 nm. However, after the annealing, the morphology became square-like, which originates from the surficial atom reorganization. The top view top-viewed orthogonal edge perfectly represents the [0001] direction of the substrate [23]. Despite that the periodical square-piece replaced the original In order to explore the surface morphology of semipolar AlN before and after 5 h high-temperature annealing, the 2 × 2 µm 2 AFM measurements were carried out as shown in Figure 8. Before the annealing, the as-grown sample presented stripe-like morphology, which is vertical to the sapphire c axis direction, due to the orientation of AlN crystallization. According to previous studies, such an orientation-preferential growth mode has been observed by the TEM method [15,16], and it originates from the anisotropic lateral growth rate of the grains [22]. Moreover, such a striation morphology which is along the sapphire [0001] direction seems to be strongly related with the grown temperature [15], and our preparation condition just agreed with the qualification. The width and vertical distance of one wire-like structure are~20 nm and~1.4 nm, respectively. The calculated root mean square (RMS) of the as-grown sample is 0.388 nm. However, after the annealing, the morphology became square-like, which originates from the surficial atom reorganization. The top view top-viewed orthogonal edge perfectly represents the [0001] direction of the substrate [23]. Despite that the periodical square-piece replaced the original striation shape, the RMS slightly increased to 0.668 nm, and the height of the step was in the range of 1 to 3 nm. When the morphology was cross-sectionally cut along the sapphire [0001] direction, the angle between exposed surface and the horizontal plane was 50 • , indicating that the exposed surface plane was most probably (11-2-1), which is principally 49.34 • with the (11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22) plane. The AlN thickness is 500 nm. Figure 9 presents the morphology evolution upon increasing annealing duration by AFM. As the figure shows, in the beginning one hour, the morphology was largely reset. The thin ripples started to aggregate and form the step morphology, which was somehow step-bunching-like. The step was parallel with the [112 ̅ 0] direction of sapphire substrate, and the step height was 3.3 nm. Likewise, square-like edge started to appear somewhere as shown by the white mark, which reprints the lattice structure of the AlN film along the in-plane direction. Upon increasing the annealing duration up to 3 h as shown in Figure  9d, the morphology did not change. Even the step height was maintained as 2.65 and 3.72 nm for 2-h and 3-h annealing samples. However, when the annealing duration was up to 4 h, the zig-zag-like edge area spread with the step-height of 4.06 nm, as shown in Figure  9e. Until 5 h, the surface morphology totally became mosaic-like. It is worth noting that as the annealing duration increased, the RMSs were 0.316 nm, 1.02 nm, 0.776 nm, 0.958 nm, 0.949 nm, and 0.916 nm in the as-grown sample, 1-h, 2-h, 3-h, 4-h, and 5-h annealed samples, respectively. Thus, despite that the RMS was obviously reset by the annealing operation, the annealing duration did not seem contribute to the RMS value.
Actually, if only the morphology itself was focused on, it was really not smooth enough, and seems even impossible to be used in epitaxy. However, despite the morphology being step-like, the vertical distance of each step was ~3 nm, which is comparable with the step of conventional step-bunching morphology whose vertical distance is as large as several nanometers in previous device epitaxy [24]. Therefore, it acts as an ideal  Figure 9 presents the morphology evolution upon increasing annealing duration by AFM. As the figure shows, in the beginning one hour, the morphology was largely reset. The thin ripples started to aggregate and form the step morphology, which was somehow step-bunching-like. The step was parallel with the 1120 direction of sapphire substrate, and the step height was 3.3 nm. Likewise, square-like edge started to appear somewhere as shown by the white mark, which reprints the lattice structure of the AlN film along the in-plane direction. Upon increasing the annealing duration up to 3 h as shown in Figure 9d, the morphology did not change. Even the step height was maintained as 2.65 and 3.72 nm for 2-h and 3-h annealing samples. However, when the annealing duration was up to 4 h, the zig-zag-like edge area spread with the step-height of 4.06 nm, as shown in Figure 9e. Until 5 h, the surface morphology totally became mosaic-like. It is worth noting that as the annealing duration increased, the RMSs were 0.316 nm, 1.02 nm, 0.776 nm, 0.958 nm, 0.949 nm, and 0.916 nm in the as-grown sample, 1-h, 2-h, 3-h, 4-h, and 5-h annealed samples, respectively. Thus, despite that the RMS was obviously reset by the annealing operation, the annealing duration did not seem contribute to the RMS value.

Conclusions
In summary, we successfully obtained the epitaxial semipolar (11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22) AlN on nonpolar m-sapphire substrate by combining sputtering and high-temperature annealing. The high-temperature annealing treatment intensively improved the crystalline quality through reordering both lattice and stress. Through varying the AlN thickness, no obvious changing was observed in both crystalline quality and strain state, therefore negating the huge contribution from the sapphire-AlN interface during the annealing process. Upon increasing the annealing duration from 1 to 5 h, although the crystalline reorder process was carried out in the first annealing hour, the morphology gradually changed into square-like zig-zag morphology upon increasing the annealing duration up to 5 h. As a consequence, our results are of importance as basic information for subsequent semipolar device epitaxy.